{"paper_id":"b6eed1ea-4332-4de0-bf0c-c530779cdcc0","body_text":"1 \nDisclaimer\nThe ECR 2025 Book of Abstracts is published by the European Society of \nRadiology (ESR) and summarises the presentations accepted to be held \nat the European Congress of Radiology 2025 (Vienna, Austria, February 26 - \nMarch 2, 2025). \nAbstracts were submitted by the authors warranting that good \nscientific practice, copyrights and data privacy regulations have been observed \nand relevant conflicts of interest declared. \nAbstracts reflect the authors' opinions and knowledge. The ESR does not give \nany warranty about the accuracy or completeness of medical procedures, \ndiagnostic procedures or treatments contained in the material included in this \npublication. The views and opinions presented in ECR abstracts and \npresentations, including scientific, educational and professional matters, do not \nnecessarily reflect the views and opinions of the ESR. \nIn no event will the ESR be liable for any direct or indirect, special, incidental, \nconsequential, punitive or exemplary damages arising from the use of these \nabstracts. \nThe Book of Abstracts and all of its component elements are for general \neducational purposes for health care professionals only and must not take the \nplace of professional medical advice. Those seeking medical advice should \nalways consult their physician or other medical professional. \nIn preparing this publication, every effort has been made to provide the most \ncurrent, accurate, and clearly expressed information possible. Nevertheless, \ninadvertent errors in information can occur. The ESR is not responsible for \ntypographical errors, accuracy, completeness or timeliness of the information \ncontained in this publication. \nThe ECR 2025 Book of Abstracts is a supplement to Insights into Imaging \n(1869-4101) and published under Creative Commons Attribution 4.0 \nInternational License. \nInsights Imaging (2025) Vol 16 (Suppl 1)\nhttps://doi.org/10.1186/s13244-025-02003-8\n\n \n \nAbstract-based Programme \n \n 2  \n \n \n \n \n  \n \nECR 2025 \nAbstract-based Programme \n \nResearch Presentation Sessions (RPS) \nWednesday, February 26 .............. 3\nThursday, February 27 ................ 72\nFriday, February 28 ................... 134\nSaturday, March 1 ..................... 193\nSunday, March 2 ........................ 260\n\n \n \nAbstract-based Programme \n \n 3  \nWednesday \n \n  \nWednesday, February 26 \n\n \n \nAbstract-based Programme \n \n 4  \nWednesday \n08:00-09:30 Research Stage 1 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 101 \nWhat's going on in the pancreas? \n \nModerator \nC. Ewertsen; Copenhagen/DK  \n(caroline.ewertsen@dadlnet.dk) \nAuthor Disclosures:  \nCaroline Ewertsen: Speaker: Bracco 2023 \n \n \nImproved pancreatic imaging with photon-counting ct  \n*E. G. S. Brandt*¹, C. F. Müller¹, A. M. Ewald¹, Y.  Wirenfeldt Nielsen¹,  \nH. S. S. Thomsen¹, B. Ibragimov², M. Andersen¹; ¹He rlev/DK, ²Copenhagen/DK \n(Erikbr1980@gmail.com) \n \nPurpose or Learning Objective: The aim of this study is to investigate the \nimage quality of pancreatic late arterial (LA) and portovenous phase (PV) \npancreatic images from Photon-Counting CT (PCCT) sc anners in comparison \nto conventional CT (EID-CT). \nMethods or Background: We retrospectively identified 35 patients without \nsuspicion of pancreatic pathology scanned on both E ID-CT and PCCT in the \nperiod from October 2021 until December 2023. IV co ntrast was given \naccording to patient weight and both a late arteria l phase (LA) and a \nportovenous phase (PV) was performed. Image quality  was rated on a 5-point \nLikert-scale (from 1=nondiagnostic to 5=optimal). E leven different pancreatic \nparameters were scored by four radiology consultant s. One reader made \nquantitative measurements of density and noise. All  data analysis was \nperformed with RStudio, version 2022.07.1. Continuo us parameters were \ncompared with a paired t-test and mean image qualit y ratings with a Wilcoxon \nsigned rank test. \nResults or Findings: Image quality was rated significantly higher on PCC T for \nthe pancreatic parenchyma in the LA (3.87 vs 2.77, p<.001), the pancreatic \nparenchyma in the PV (3.31 vs 2.53, p<.001), pancre atic ducts (2.88 vs 2.62, \np<.001), SMA (4.10 vs 2.74, p<.001), celiac axis (4 .04 vs 2.70, p<.001) and \nportal vein (3.29 vs 2.52, p<.001). Noise levels we re significantly lower with \nPCCT with a mean reduction of 5.8 HU across all par ameters. DLP was \nsignificantly reduced with a 31.8% reduction (p< 0. 01) for the LA and 21.5% \n(p< 0.01) for the PV. \nConclusion: Image quality was significantly improved for all ev aluated \npancreatic and peripancreatic structures with PCCT.  Additionally, image noise \nand radiation dose were significantly reduced. The improved image quality with \nPCCT could potentially lead to improvements in the currently difficult \nevaluation of pancreatic diseases. \nLimitations: No limitations were identified. \nFunding for this study: Innovation Fund Denmark, Grant No. 1044-00015B. \nEthics committee - additional information: The study was approved by the \nnational ethics committee with the number: nvk22153 38. \nAuthor Disclosures:  \nMichael Andersen: Speaker: Different Vendor Seminar s, GE, Philips and \nSiemens \nAnne Marie Ewald: Nothing to disclose \nHenrik S. S. Thomsen: Nothing to disclose \nBulat Ibragimov: Nothing to disclose \nErik Gudmann Steuble Brandt: Nothing to disclose \nYousef Wirenfeldt Nielsen: Nothing to disclose \nChristoph Felix Müller: Nothing to disclose \n \n \nDevelopment and Validation of Contrast-enhanced CT- based Imaging \nIntratumor Heterogeneity of Pancreatic Ductal Adeno carcinoma \n*B. Zhao*, S. Ju; Nanjing/CN \n(zhaoben1207@163.com) \n \nPurpose or Learning Objective: To construct an imaging ITH (IITH) through \nradiomics methodology to effectively reflect the IT H of PDAC and explore its \nprognostic value. \nMethods or Background: This study enrolled 961 patients with pathologicall y \nconfirmed PDAC who had undergone preoperative contr ast-enhanced \ncomputed tomography (CT) in two cohorts. Firstly, T umor regions of interest \nwere automatically segmented in both arterial and v enous phase images. \nRadiomics features from these 2 phasea images were extracted based on \nPyRadiomics. Highly variable radiomic features with  median absolute deviation \n> 1, were selected to assess the ITH. Similarity Ne twork Fusion (SNF) was \nemployed to identify distinct imaging heterogeneity  phenotypes in the \ndiscovery cohort (Cohort 1), and the identical crit eria were applied to Cohort 2. \nKaplan-Meier analysis was utilized to investigate t he association between the \nidentified imaging phenotypes and overall survival (OS). \nResults or Findings: A total of 961 patients (mean age, 63.1 years ± 9.2  \nstandard deviation; 460 men) from 2 cohorts were en rolled. And 3378 \nradiomics features were extracted for each patient.  In the discovery cohort \n(Cohort 1, n = 637), 241 highly variable heterogene ity-related features were \nselected to identify IITH, and 283 patients were cl assified into a high-IITH \nsubgroup. Patients with high-IITH in the discovery cohort presented \nsignificantly poorer OS compared to those with low IITH (median time, 20.6 vs. \n37.1 months, P < 0.001). The identical criteria wer e applied to the validation \ncohort (Cohort 2, n = 324). Kaplan-Meier analysis a lso confirmed that patients \nwith high-IITH (n =149) had shorter OS (median time , 16.4 vs. 26.3 months,  \nP < 0.001). \nConclusion: We established a noninvasive radiomics method to ev aluate the \nITH of PDAC. Furthermore, we demonstrated the progn ostic power of IITH. \nLimitations: This study requires further multi-omics validation in the future. \nFunding for this study: NSFC, No. 82330060, 92059202, 823B2040, \n61821002 and 82372024) \nEthics committee - additional information: IEC for Clinical Research \nofZhongda Hospital, Affliated to Southeast Universi ty \nAuthor Disclosures:  \nBen Zhao: Nothing to disclose \nShenghong Ju: Nothing to disclose \n \n \nInterobserver agreement of pancreatic tumor size me asurement before \nand after neoadjuvant therapy: is MRI as reproducib le as CT? \n*A. Licha*, C. Touloupas, A. Delpla, A. Pouvelle, M . Zins; Paris/FR \n \nPurpose or Learning Objective: Assess inter-observer agreement of \npancreatic ductal adenocarcinoma (PDAC) tumor size measurement on CT \nand MRI, before and after neoadjuvant therapy (NAT)  \nMethods or Background: We reviewed all patients with a histological \ndiagnosis of PDAC at Paris Saint-Joseph Hospital, b etween 2010 and 2022, \nand who underwent CT and MRI, both before and after  NAT. Three \nindependent radiologists anonymously evaluated the large axial tumor axis on \n2 CT acquisitions and 6 MRI sequences. Inter-observ er agreement was \nassessed by intra-class correlation coefficients (I CCs) and by LOAM graphs \n(Bland & Altmann extension for multiple observers).  \nResults or Findings: The final population consisted of 50 patients. On C T \nexams, inter-observer agreement was excellent befor e NAT (ICC of 0.83 \n[0.73;0.90] at arterial phase and 0.84 [0.74;0.90] at portal-venous phase) and \ndecreased but remained good after NAT (ICC of 0.66 [0.52;0.78] at arterial \nphase and 0.65 [0.51;0.77] at portal-venous phase).  On MRI exams, inter-\nobserver agreement was moderate to good before NAT (best sequence being \nT1 at arterial phase with ICC of 0.67 [0.53;0.79]) and decreased becoming \nmoderate for all sequences after NAT (best sequence  being T1 at late phase \nwith ICC of 0.55 [0.37;0.71]). \nConclusion: Inter-observer agreement of PDAC great axe measurem ent is \nbetter on CT than on MRI, and decreases between pre -NAT and post-NAT \nimaging, both on CT and MRI. These results encourag e to keep on performing \ntechnically perfect CT scans, without questioning t he need for pre-operative \nhepato-pancreatic MRI in non-metastatic patients. \nLimitations: This is a monocentric retrospective study. \nFunding for this study: Inter-observer agreement of tumor size measurement \nis higher on CT than on MRI, both before and after NAT.  Inter-observer \nagreement of tumor size measurement decreases follo wing NAT, both on CT \nand MRI. \nEthics committee - additional information: This study was approved by our \norganization's Medical Research Ethics Group. \nAuthor Disclosures:  \nCaroline Touloupas: Nothing to disclose \nMarc Zins: Nothing to disclose \nArié Licha: Nothing to disclose \nAlexandre Delpla: Nothing to disclose \nArnaud Pouvelle: Nothing to disclose \n \n \nA Radiomics-Based Model for Predicting Lymph Node M etastasis of \nPancreatic Ductal Adenocarcinoma: A Multi-Center St udy \n*B. Zhao*, S. Ju; Nanjing/CN \n(zhaoben1207@163.com) \n \nPurpose or Learning Objective: To develop a radiomics model to predict \nlymph node metastasis (LNM) in patients with pancre atic ductal \nadenocarcinoma (PDAC) and assess its value for clin ical management. \nMethods or Background: Patients with pathologically confirmed PDAC were \nretrospectively enrolled from four centers and divi ded into a training (n = 192), \nvalidation (n = 82), testing (n = 100), and clinica l utilization cohort (n = 163).A \nradiomics model was constructed based on the arteri al phase of computed \n\n \n \nAbstract-based Programme \n \n 5  \nWednesday \ntomography (CT) for predicting LNM. The areas under  the curve (AUCs) were \nused to compare the performance between the radiomi cs model and other \nmodels. Subsequently, Kaplan-Meier analysis was use d to validate the model’s \nvalue for prognosis and therapy decisions. \nResults or Findings: A total of 437 patients (mean age, 63.1 years ± 9.2  \nstandard deviation; 253 men) were included. The rad iomics model \ndemonstrated AUCs of 0.84, 0.82, and 0.78 in the tr aining, validation, and \ntesting cohorts, respectively, superior to other mo dels (all P < 0.05). Besides, \nLNM predicted by the radiomics model was strongly a ssociated with overall \nsurvival (OS) (P < 0.001). Kaplan-Meier analysis al so demonstrated that \npatients with a high risk of LNM had a worse progno sis (all P < 0.05). \nFurthermore, patients who were dissected with ≥ 15 LNs had a longer OS than \nthose with fewer LNs dissected in the high-risk sub group predicted by the \nradiomics model in the clinical utilization cohort (P = 0.002). \nConclusion: The radiomics model demonstrated impressive perform ance in \npredicting LNM and prognosis, indicating its potent ial for. therapy decisions. \nLimitations: The model we developed should be validated in a pro spective \nstudy. \nFunding for this study: NSFC, No. 82330060, 92359304, 92059202, \n823B2040, 61821002 and 82372024 \nEthics committee - additional information: IEC for Clinical Research \nofZhongda Hospital, Affliated to Southeast Universi ty \nAuthor Disclosures:  \nBen Zhao: Nothing to disclose \nShenghong Ju: Nothing to disclose \n \n \nCT-Based Early Indicators of Severe Pancreatic Fist ula and Hemorrhage \nAfter Pancreatoduodenectomy \nD. Palumbo, *A. Campisi*, V. Andreasi, F. Prato, S.  Partelli, D. Tamburrino,  \nM. Falconi, F. De Cobelli; Milan/IT \n \nPurpose or Learning Objective: Postoperative pancreatic fistula (POPF) and \npostpancreatectomy hemorrhage (PPH) are major compl ications following \npancreatoduodenectomy (PD). Despite their clinical importance, no tool \ncurrently exists to predict their occurrence or sev erity. This study aims to \nidentify radiological characteristics that can aid in the early prediction and \nstratification of POPF and PPH. \nMethods or Background: We retrospectively reviewed 399 patients who \nunderwent PD at San Raffaele Hospital between Janua ry 2015 and December \n2021. Patients included had at least one contrast-e nhanced computed \ntomography (CE-CT) scan within 14 days post-surgery . Several radiological \nfeatures were systematically assessed, including pa ncreaticojejunostomy (PJ) \ndehiscence, PJ defects, fluid collections, perianas tomotic air bubbles, and \npancreatic remnant density. \nResults or Findings: Clinically relevant POPF occurred in 230 patients \n(57.9%), with 185 classified as grade B and 45 as g rade C. PPH occurred in 61 \npatients (15.3%). PJ dehiscence was significantly a ssociated with clinically \nrelevant POPF (31% vs. 22%, p = 0.035), and PJ defe cts were more extensive \nin patients with severe POPF (median 7 mm vs. 5 mm,  p = 0.001). Fluid \ncollections, particularly above the PJ site, were a lso linked to the development \nof severe POPF (p < 0.001). Additional markers, suc h as stump pancreatitis, \nperianastomotic air bubbles, and lower pancreatic r emnant density, were \nsignificantly more frequent in severe POPF cases. \nConclusion: These findings support the use of postoperative CT scans to \nprospectively identify patients at risk of developi ng severe POPF and PPH, \nallowing for better clinical management. \nLimitations: Retrospective study design. \nFunding for this study: None. \nEthics committee - additional information: Ethics committee approval \nnumber: 28/INT/2015 \nAuthor Disclosures:  \nDomenico Tamburrino: Nothing to disclose \nValentina Andreasi: Nothing to disclose \nDiego Palumbo: Nothing to disclose \nMassimo Falconi: Nothing to disclose \nAntonino Campisi: Nothing to disclose \nStefano Partelli: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \nFrancesco Prato: Nothing to disclose \n \n \nCorrelation between celiac axis stenosis and compli cations after \npancreatoduodenectomy \n*Y. Shu*, Y. Dai, J. Wei, Q. Xu; Nanjing/CN \n(18227237972@163.com) \n \nPurpose or Learning Objective: We aimed to explore the correlation between \nceliac axis stenosis and complications after pancre atoduodenectomy. \nMethods or Background: Patients who underwent pancreatoduodenectomy \nin our hospital pancreas center from January 2021 t o December 2023 were \nretrospectively collected. The stenosis rate of cel iac axis was measured on \npre-operation arterial phase imaging of routine enh anced CT, and graded the \nseverity of celiac trunk stenosis: no stenosis (< 3 0%), mild stenosis (30%-\n50%), and significant stenosis (≥50%). The incidence of postoperative \ncomplications was evaluated, and both univariate an d multivariate logistic \nregression analysis were conducted. \nResults or Findings: A total of 774 patients were included in the study,  205 \n(26.5%) had celiac axis stenosis: 144 (18.6%) with mild stenosis, and 61 \n(7.9%) with significant stenosis. Celiac axis steno sis was associated with \npancreatic fistula (p<0.001), postoperative bleedin g (p=0.033), and \npostoperative biliary leakage (p= 0.006). In multiv ariate logistic regression \nanalysis, mild stenosis of the celiac axis was an i ndependent risk factor for \npostoperative pancreatic fistula (OR 2.81, 95%CI 1. 82-4.33, p<0.001), and \nsignificant stenosis of the celiac axis was an inde pendent risk factor for \npostoperative biliary leakage (OR 4.91, 95%CI 1.27- 19.04 , p=0.021). \nConclusion: Celiac axis stenosis was associated with the risk o f complications \nafter pancreatoduodenectomy. Surgeons may need to p ay attention to the \ncondition of celiac axis stenosis before pancreatod uodenectomy. \nLimitations: Retrospective study. Single center. \nFunding for this study: There was no funding for this study. \nEthics committee - additional information: None \nAuthor Disclosures:  \nYuping Shu: Nothing to disclose \nQing Xu: Nothing to disclose \nYuran Dai: Nothing to disclose \nJishu Wei: Nothing to disclose \n \n \nCystic fluid non-invasive evaluation based on photo n-counting detector \nCT spectral imaging in patients with pancreatic cys tic lesions \n*I. Dudás*, B. Lovász, M. Benke, Á. Szücs, P. N. Ka posi-Novák, A. Szijártó,  \nP. Maurovich-Horvat, B. K. Budai; Budapest/HU \n \nPurpose or Learning Objective: Differentiation between pancreatic cystic \nlesions is a challenging task for clinicians. Spect ral imaging via photon-\ncounting detector CT (PCD-CT) scanners allows the r econstruction of virtual \nmonoenergetic images (VMI) enabling the measurement  of Hounsfield unit \n(HU) densities at different keV and the generation of spectral absorption \ncurves. Our study aimed to investigate whether muci nous and non-mucinous \npancreatic cystic lesions (PCL) have different spec tral absorption curves that \ncould help the differential diagnostics. \nMethods or Background: Our study included 74 patients with PCLs, 53 \npatients with mucinous cystic neoplasms and 21 with  non-mucinous cystic \npancreatic lesions diagnosed based on current pract ice guidelines. The \nspectral absorption curves were generated from the pancreatic-phase scans. \nThe average densities were measured on the 70keV (H U70keV) and 40keV \n(HU40keV) virtual monoenergetic images (VMIs), by p lacing 3-3 circular \nregions of interest in PCL’s cystic component parts  and the density differences \nwere calculated (HUdiff(40keV-70keV)). Kruskal-Wall is test with post-hoc \nDunn’s test was used for comparing the groups. The discrimination \nperformance was assessed by receiver operating char acteristic (ROC) curve \nanalysis. The intraobserver reproducibility and int erobserver reproducibility \nwere evaluated by the intraclass correlation coeffi cient (ICC). \nResults or Findings: On 70keV VMIs, no significant differences were foun d \nbetween the average densities of mucinous and non-m ucinous PCLs cystic \ncomponents, however, a significant difference was f ound in HUdiff(40keV-\n70keV) values (p<0.0001). The diagnostic performanc e of HUdiff(40keV-\n70keV) in differentiating between mucinous vs. non- mucinous PCLs had AUCs \nof 0.92 and 0.92 on the training and test datasets,  respectively, with a good \ninterobserver (ICC=0.82) and excellent intraobserve r reproducibility \n(ICC=0.94). \nConclusion: Spectral absorption curve assessment of cystic comp onents \ncould be a useful additional measurement to facilit ate the non-invasive \ndifferential diagnosis between mucinous and non-muc inous pancreatic cystic \nlesions. \nLimitations: This was a single-center study with a retrospective  study design. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The presented study was \napproved by the institutional ethics committee of o ur University (SE RKEB \n256/2023). \nAuthor Disclosures:  \nAttila Szijártó: Nothing to disclose \nPál N. Kaposi-Novák: Nothing to disclose \nPál Maurovich-Horvat: Nothing to disclose \nÁkos Szücs: Nothing to disclose \nIbolyka Dudás: Nothing to disclose \nMárton Benke: Nothing to disclose \nBorbála Lovász: Nothing to disclose \nBettina Katalin Budai: Nothing to disclose \n \n \n \n\n \n \nAbstract-based Programme \n \n 6  \nWednesday \nCT-Derived Body Composition Metrics Predict Severit y in Acute \nPancreatitis: A Post-Hoc Multicenter Study \n*R. Z. Borbély*¹, B. Teutsch¹, V. Vass¹, K. Márta¹,  B. Erőss¹, A. Vincze²,  \nA. Szentesi¹, P. Hegyi¹, N. Faluhelyi²; ¹Budapest/H U, ²Pécs/HU \n(drborbelyruben@gmail.com) \n \nPurpose or Learning Objective: Accurately predicting the severity of acute \npancreatitis (AP) remains a significant clinical ch allenge. CT scans, traditionally \nused for diagnosis, also provide valuable insights into body composition. \nHowever, the prognostic utility of CT-derived body composition metrics has \nbeen inconsistently reported. This study aimed to e valuate whether key CT \nbody composition metrics can effectively predict th e severity of AP. \nMethods or Background: A post-hoc analysis was conducted on a \nmulticenter study involving 437 AP patients who und erwent CT scans within \nthe first 24 hours of hospital admission. Measureme nts of visceral adipose \ntissue (VAT), subcutaneous adipose tissue (SAT), an d skeletal muscle area \n(SMA) were obtained at the third lumbar vertebra le vel. These areas were \nnormalized for patient height to calculate the Visc eral Adipose Tissue Index \n(VATI), Subcutaneous Adipose Tissue Index (SATI), a nd Skeletal Muscle \nIndex (SMI). Muscle radiodensity in Hounsfield Unit s (HU) was assessed to \ndetermine fatty infiltration. The fat-to-muscle vol ume ratio was also calculated. \nThese metrics were analyzed as potential predictors  of severe AP using \nreceiver operating characteristic (ROC) curves and area under the curve (AUC) \nvalues. Severity was determined using the Modified CT Severity Index \n(mCTSI). Statistical analyses were performed using IBM SPSS Statistics. \nResults or Findings: The fat-to-muscle ratio demonstrated the highest \npredictive accuracy for severe AP (AUC = 0.68), fol lowed by VATI (AUC = \n0.65). Other indices did not show significant predi ctive potential (AUC < 0.6). \nConclusion: CT-derived body composition metrics, particularly t he fat-to-\nmuscle ratio and VATI, are valuable predictors of s evere acute pancreatitis. \nIncorporating body composition analysis into routin e CT evaluations may \nenhance prognostic assessments for AP patients. \nLimitations: As a post-hoc analysis of retrospective data, the s tudy may be \nsubject to selection bias, potentially limiting the  generalizability of the findings. \nFunding for this study: Funding for Ruben Zsolt Borbély was supported by \nthe EKÖP-2024-239 New National Excellence Program o f the Ministry for \nCulture and Innovation from the source of the Natio nal Research Development \nand Innovation Fund. Funding for Brigitta Teutsch w as provided by the ÚNKP-\n22-3 New National Excellence Program of the Ministr y for Innovation and \nTechnology from the source of the National Research , Development and \nInnovation Fund (to BT - ÚNKP-22-3-IPTE-1693). Cent er costs were covered \nby the University of Pécs, the Momentum Grant of th e Hungarian Academy of \nSciences (LP2014-10/2014), and grants from the Nati onal Research, \nDevelopment, and Innovation Office (GINOP-2.3.2-15- 2016-00015, KH-\n125678). The funders had no influence on the study design, data collection, \nanalysis, or manuscript preparation. \nEthics committee - additional information: This post-hoc analysis is based \non a study that received ethical approval from the Scientific and Research \nEthics Committee of the Medical Research Council (I SRCTN63827758, \ndecision 55961-2/2016/EKU). \nAuthor Disclosures:  \nBálint Erőss: Nothing to disclose \nNándor Faluhelyi: Nothing to disclose \nPéter Hegyi: Nothing to disclose \nKatalin Márta: Nothing to disclose \nRuben Zsolt Borbély: Nothing to disclose \nAron Vincze: Nothing to disclose \nBrigitta Teutsch: Nothing to disclose  \nVivien Vass: Nothing to disclose \nAndrea Szentesi: Nothing to disclose \n \n \nRole of incidental pancreatic calcifications on com puted tomography as \nopportunistic biomarker for chronic pancreatitis \n*A. Pata*, F. Rizzetto, C. B. Monti, A. Vanzulli; M ilan/IT \n(annamaria.pata@unimi.it) \n \nPurpose or Learning Objective: To evaluate whether incidentally detected \npancreatic calcifications on computed tomography (C T) serve as a reliable \nbiomarker for chronic pancreatitis as defined by cl inical criteria. \nMethods or Background: We retrospectively reviewed CT scans from adult \npatients between 2014 and 2024, identifying cases w here \"pancreatic \ncalcifications\" were mentioned in the radiology rep ort. Patients with known \nhistory of pancreatitis or pancreatic surgery were excluded. For each patient, \nwe recorded pancreatic size, calcification characte ristics (number, size, and \nlocation), and other features of chronic pancreatit is, such as duct dilatation or \nintraductal calculi. Clinical data, including pain,  abdominal symptoms, and risk \nfactors for chronic pancreatitis, were also collect ed. \nResults or Findings: A total of 137 patients with incidental pancreatic \ncalcifications were identified. A small subset had coarse calcifications (n=9, \n7%), while the majority had both coarse and punctif orm calcifications (n=102, \n74%), with calcification numbers ranging from 7 to 50 in over half of the cases. \nIn 121 patients (88%), at least two pancreatic segm ents were involved, most \nfrequently the head (n=125, 91%) and the body (n=11 2, 81%). When coarse \ncalcifications were present alongside duct dilation , the duct caliber was \nsignificantly larger (6.7 mm vs 4.1 mm; p=0.026), w hile the tail was slightly \nsmaller (17 mm vs 19 mm, p=0.018), with no signific ant differences in the size \nof the head or body (p>0.198). Among patients with available clinical \ninformation (n=38), only 3 (11%) reported symptoms such as abdominal pain, \ndiarrhea, or bloating, and none (0%) reported weigh t loss. Elevated alcohol \nconsumption or smoking history was not associated w ith calcification type or \nlocation (p>0.186), pancreatic size, or duct dilati on (p>0.317). \nConclusion: Incidentally detected pancreatic calcifications are  not a reliable \nbiomarker for screening for chronic pancreatitis. \nLimitations: The main study limitation is its retrospective desi gn. \nFunding for this study: No fundings were received for this study \nEthics committee - additional information: Institutional Review Board \napproved the retrospective data collection in anony mous form \nAuthor Disclosures:  \nFrancesco Rizzetto: Author: nothing to disclose \nCaterina Beatrice Monti: Author: nothing to disclos e \nAngelo Vanzulli: Author: nothing to disclose \nAnnamaria Pata: Author: nothing to disclose \n \n \nThe diagnostic potential of unenhanced dual-layer s pectral CT \nquantitative parameters in diabetic pancreas \n*L. Ge*, Y. Li, Y. Gao, X. Zhang, X. Yu; Xi An/CN \n(1051394676@qq.com) \n \nPurpose or Learning Objective: To investigate pancreas characterizing for \ntype 2 diabetes mellitus (T2DM) using unenhanced du al-layer Spectral CT. \nMethods or Background: This retrospective study included patients who \nunderwent abdominal unenhanced dual-layer spectral CT between March 2023 \nand April 2024. The patients were divided into T2DM  group and control group. \nNine regions of interest (ROIs) were drawn (three f or head, three for body and \nthree for tail). Mean attenuation on conventional 1 20-kVp CT images (CTconv), \neffective atomic number maps (Z-eff), iodine densit y maps (ID), virtual non-\ncontrasted (VNC), and mean attenuation on virtual m onoenergetic images \n(VMIs) at 40-200keV were measured. The Mann -Whitne y U test was used to \ncompare the differences between the two groups. The  receiver operating curve \n(ROC) was used to evaluate the diagnostic efficacy of the above parameters. \nResults or Findings: A total of 84 patients, including 44 T2DM patients and \n40 controls, were evaluated. There was a statistica lly significant difference in \nthe CTconv (46.7 ± 5.9 HU vs. 50.7 ± 4.5 HU, p < 0.01), Z-eff (7.43 ± 0.05 vs. \n7.46 ± 0.03, p < 0.01), ID (103.9 ± 0.6 vs. 104.3 ± 0.4, p < 0.01), VNC (39.2 ± \n5.5 HU vs. 42.3 ± 4.2 HU, p < 0.01) and VMIs (62.1 ± 9.7 HU vs. 68.6 ± 6.2 HU \nat 40 keV, p < 0.001) between the T2DM group and th e control group. The \nAUCs of the CTconv, Z-eff, ID, VNC and 40 keV were 0.722,0.671,0.695, \n0.691 and 0.734. Histogram analysis found that the 10th percentile value of the \nabove parameters had higher diagnostic efficiency ( 0.757,0.736,0.702,0.734 \nand 0.805). The AUC of VMIs increased with decreasi ng monoenergeic levels. \nConclusion: The 10th percentile value of 40 keV was the best in dicator for \ndistinguishing T2DM patients from the controls. \nLimitations: Not applicable \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nXiao Yu: Nothing to disclose \nYishan Li: Nothing to disclose \nXiaoyue Zhang: Nothing to disclose  \nYanjun Gao: Nothing to disclose \nLiu Ge: Nothing to disclose \n \n \nAbdominal mapping with MOLLI and T2p-SSFP: can you do without the \ncardiac trigger? \nA. Lupi, P. Dardeshi, I. Toniolo, *G. Beggiato*, M.  Pizzi, D. Negro, E. Quaia,  \nA. Pepe; Padua/IT \n(giulia.beggiato.1@studenti.unipd.it) \n \nPurpose or Learning Objective: Quantitative abdominal imaging is \nincreasingly emerging in the radiology arena, drawi ng inspiration from cardiac \nmapping, which is significantly changing patient ma nagement. In fact, cardiac \nmapping sequences offer the possibility of obtainin g abdominal mapping \nvalues, although with long acquisition times, due t o the cardiac trigger. The aim \nof our study is to evaluate the differences between  abdominal mapping values \nobtained with and without cardiac trigger. \nMethods or Background: Ten patients who were candidates for the \nabdominal mapping study as part of a research proto col were included. MOLLI \nand T2p-SSFP sequences were acquired for T1 and T2 mapping, respectively, \nwith (T) and without (NT) cardiac trigger and with simulated trigger (ST, HR 60 \nbpm). T1/T2 mapping values were extracted using man ually traced regions of \n\n \n \nAbstract-based Programme \n \n 7  \nWednesday \ninterest (ROIs) at the hepatic, pancreatic and rena l cortical levels, using the \ncvi42 software. The results obtained were compared using paired t-tests. \nResults or Findings: Liver T1/T2 mapping showed statistically significan t \ndifferences between NT and ST sequences (p=0.003 an d 0.004, respectively), \nand T2 mapping values turned out to be significantl y different between NT and \nT acquisition also (p<0.001). Pancreas T1mapping va lues were different \nbetween NT vs T (p=0.016) and ST (p=0.037), while T 2mapping values were \ndifferent between NT and ST only (p=0.029). No diff erences between T and ST \nsequences were found in liver and pancreas T1/T2 ma pping (p>0.05). Renal \nT1/T2 mapping did not show significant differences among the three \nacquisition strategies. \nConclusion: Our results show that cardiac trigger should not be  avoided in \nliver and pancreas mapping with MOLLI and T2p-SSFP sequences, but in \norder to reduce acquisition time, a simulator could  be used. Further analysis on \nlarger sample and with standard ROIs are needed to confirm these data. \nLimitations: Sample size \nFunding for this study: n/a \nEthics committee - additional information: Azienda Ospedale Università \nPadova \nAuthor Disclosures:  \nAmalia Lupi: Nothing to disclose \nAlessia Pepe: Nothing to disclose \nGiulia Beggiato: Nothing to disclose \nMarco Pizzi: Nothing to disclose \nPajtim Dardeshi: Nothing to disclose \nDonato Negro: Nothing to disclose \nEmilio Quaia: Nothing to disclose \nIrene Toniolo: Nothing to disclose \n \n \n08:00-09:30 Research Stage 2 \nResearch Presentation Session: Cardiac \nRPS 103 \nApplications of cardiac CT \n \nModerator \nR. Vliegenthart; Groningen/NL  \n(r.vliegenthart@umcg.nl) \nAuthor Disclosures:  \nRozemarijn Vliegenthart: Advisory Board: Lifelines,  ICAN (Institute for \nCardiometabolism and Nutrition); Board Member: ESCR ; Grant Recipient: \nSiemens Healthineers (institutional research grant) ; Speaker: Siemens \nHealthineers, Bayer Healthcare, Wiley \n \n \nLong-term exposure to particulate and gaseous air p ollution and \ncoronary atherosclerotic disease assessed by cardia c CT \n*F. Castillo Aravena*¹, C. Desroche², S. Delaney³, R. Nethery³,  \nP. Thavendiranathan¹, H. Ross¹, K. Hanneman¹; ¹Toro nto, ON/CA,  \n²Kingston, ON/CA, ³Boston, MA/US \n(felipe.castilloaravena@uhn.ca) \n \nPurpose or Learning Objective: Both fine particulate matter (PM2.5) and \nnitrogen dioxide (NO2) are associated with cardiova scular mortality. However, \nthe underlying pathophysiological mechanisms are un clear. The purpose of \nthis study was to evaluate the relationship between  long-term exposure to \nthese air pollutants and extent of coronary artery disease. \nMethods or Background: Adult patients undergoing cardiac CT between \n2012-2023 were retrospectively evaluated. Coronary atherosclerosis was \nquantified using Agatston coronary artery calcium s cores (CACS). Long-term \nair pollution exposures were assessed as the averag e of daily direct \nmeasurements of PM2.5 and NO2 in the ten-year perio d prior to cardiac CT. \nMultivariable linear regression models were adjuste d for sex, age, year, \ndistance to monitoring station, and socioeconomic s tatus (neighborhood \nmedian household income and employment rate). \nResults or Findings: 11,140 patients were included (52% male, mean age \n59±11 years). Median 10-year exposure to PM2.5 was 7.5 (range 4.3–9.2) \nμg/m3 and NO2 was 13.4 (range 3.2-17.8) parts per bi llion (ppb). Each 1 \nµg/m3 increase in ten-year PM2.5 exposure was assoc iated with 23.2 higher \nCACS (β-coefficient 23.2, 95%CI, 5.3-41.0, P=0.011) in una djusted analysis \nand 19.2 higher CACS (β-coefficient 19.2, 95%CI, 0.7-37.7, P=0.042) in \nmultivariable analysis. Each 1 ppb increase in ten- year NO2 exposure was \nassociated with 5.0 higher CACS (β-coefficient 5.0, 95%CI, 1.9-8.2, P=0.002) \nin unadjusted analysis; however, this association w as attenuated in \nmultivariable analysis (β-coefficient 1.4, 95%CI, -1.7-4.4, P=0.38). \nConclusion: Higher long-term exposure to fine particulate (PM2. 5) air pollution \nis associated with higher extent of coronary athero sclerotic disease. The \nrelationship with NO2 was not significant in adjust ed analysis. These results \nhighlight the potential for CT to detect the sequel a of long-term air pollution. \nLimitations: CACS does not quantify non-calcified plaque and fur ther study is \nneeded to evaluate relationships with total plaque burden. \nFunding for this study: None \nEthics committee - additional information: University Health Network \n(CAPCR 24-5344) \nAuthor Disclosures:  \nPaaladinesh Thavendiranathan: Nothing to disclose \nRachel Nethery: Nothing to disclose \nKate Hanneman: Nothing to disclose \nHeather Ross: Nothing to disclose \nChloe Desroche: Nothing to disclose \nFelipe Castillo Aravena: Nothing to disclose \nScott Delaney: Nothing to disclose \n \n \nVideo Assisted Informed Consent in Cardiac Imaging:  Influence on \nPatient Anxiety during CT – The VAICICI-trial \n*R. Gohmann*, S. Mettke, C. F. Lücke, C. D. Kriegho ff, M. Gutberlet; \nLeipzig/DE \n(robin.gohmann@helios-gesundheit.de) \n \nPurpose or Learning Objective: CT is a non-invasive tool for the diagnosis of \ncoronary artery disease (CAD) and preoperative plan ning. However, cardiac \nCT (cCT) can elicit anxiety, potentially impacting patient compliance and \nultimately image quality. This study investigates w hether video-assisted \ninformed consent in cardiac imaging (VAICICI) reduc es patient anxiety during \ncCT and enhances patient understanding of the exami nation process. \nMethods or Background: This prospective, randomized, controlled trial \nenrolled 205 patients scheduled for cCT. Patients w ere randomized into three \ngroups: Control (n=69), Video I (n=67), and Video I I (n=69). Video I was an \neducational video with visuals, subtitles, and voic eover explaining the \nexamination. Video II presented only voiceover and subtitles without visuals. \nAll patients received a standard physician consulta tion. Anxiety and patient \nsatisfaction were measured using visual analog scal es immediately before and \nafter the examination. Statistical analyses include d group-comparisons and \nmultivariate-analysis to examine the influence of d emographic and anamnestic \nvariables. \nResults or Findings: Both Video I and II significantly improved patient \nunderstanding, and satisfaction compared to the con trol group (p<0.05). \nPatients with Video II reported the informed consen t form as more important for \nunderstanding than those in the control group (p=0. 023). Satisfaction was \nhigher after watching any video (p=0.020) with sign ificant difference between \nVideo I and II. Anxiety levels did not differ betwe en the groups, though female \npatients (p=0.008) and those having suspected CAD r eported higher pre-\nexamination-anxiety. Overall, 10 baseline demograph ics were found to be \npartially explanatory to the response and independe ntly statistically significant, \ne.g. age and previous CT/MRI-experience. \nConclusion: VAICICI improved patient satisfaction and understan ding of the \ncCT examination. However, its impact on reported an xiety reduction was \nlimited. The findings suggest that VAICICI enhances  the informed consent \nprocess. Its influence on compliance and thus image  quality remains to be \ninvestigated. \nLimitations: The Single-center Design And The Relatively High Fa miliarity \nWith Imaging Among The Study Cohort May Limit The G eneralizability Of The \nResults. \nFunding for this study: None. \nEthics committee - additional information: This Study Was Approved By \nThe Local Ethics Committee (Reference-No.:172/22-ek ). Written Informed \nConsent Was Waived. \nAuthor Disclosures:  \nRobin Gohmann: Nothing to disclose \nChristian Friedrich Lücke: Nothing to disclose \nMatthias Gutberlet: Nothing to disclose \nChristian Dominik Krieghoff: Nothing to disclose \nSophia Mettke: Nothing to disclose \n \n \nImpact of Cardiac Computed Tomography Angiography t o prevent major \nadverse cardiovascular events in patients undergoin g diagnostic work-up \nfor orthotopic liver transplant \n*F. Santoro*, B. La Delfa, D. Tore, C. Guarnaccia, R. Faletti, G. A. Strazzarino, \nC. Gaetani, A. Depaoli, P. Fonio; Turin/IT \n(federicasantoro1996@gmail.com) \n \nPurpose or Learning Objective: To evaluate the impact of Cardiac CT \nAngiography (CCTA) in cardiovascular risk assessmen t to prevent major \nadverse cardiovascular events (MACE) in patients un dergoing diagnostic work-\nup before orthotopic liver transplant (OLT). \n\n \n \nAbstract-based Programme \n \n 8  \nWednesday \nMethods or Background: Monocentric retrospective study on 140 patients \nwith intermediate to high risk of CAD who underwent  CCTA during pre-OLT \ndiagnostic work-up at our Institution from March 20 21 to October 2024. All \nexams were performed using prospective ECG-gated si ngle heartbeat axial \nacquisition (0.28 s gantry rotation time, kV and mA  set depending on patient \nBMI, ECG window 40-80% of R-R cycle) with a whole-h eart coverage CT \nscanner (Revolution CT, GE, USA). 95 patients were classified with non-critical \nCAD at CCTA. 23 patients with suspicion of critical  CAD at CCTA underwent \ninvasive coronary angiography (ICA). Sensitivity, s pecificity, positive predicting \nvalue (PPV) and negative predictive value (NPV) wer e calculated. \nResults or Findings: In the 95 patients with negative CCTA no one develo ped \nMACEs or cardiovascular related complications while  on waiting list, nor during \nor after OLT; 23 patients with critical disease (CA D-RADS 4-5) at CCTA \nsubsequently underwent ICA. In 14 cases stenosis wa s confirmed at ICA and \npatients underwent percutaneous coronary interventi on (PCI) with Drug Eluting \nStent (DES). CCTA in this group of patients had 100 % sensitivity, 91.3% \nspecificity, 100% negative predicting value and 60. 9% positive predicting \nvalue. \nConclusion: CCTA has an excellent diagnostic accuracy for cardi ovascular \nrisk stratification in the setting of pre-OLT work- up and it may have a role in \npreventing MACEs or other cardiovascular events in asymptomatic patients \nwith intermediate to high risk of CAD. \nLimitations: Monocentric study, small simple size. \nFunding for this study: Nothing to discloure \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nGiulio Antonino Strazzarino: Nothing to disclose \nAlessandro Depaoli: Nothing to disclose  \nRiccardo Faletti: Nothing to disclose \nCarla Guarnaccia: Nothing to disclose \nBenedetta La Delfa: Nothing to disclose \nFederica Santoro: Nothing to disclose \nDavide Tore: Nothing to disclose \nPaolo Fonio: Nothing to disclose \nClara Gaetani: Nothing to disclose \n \n \nMatters of the heart in stroke - acute phase cardia c CT in intracranial \nlarge vessel occlusion stroke for the identificatio n of high-risk imaging \nbiomarkers indicative of a cardioembolic cause \n*K. Mostafa*, C. Wolf, C. Wünsche, S. Krutmann, M. Both, O. Jansen,  \nH. Seoudy, P. Langguth; Kiel/DE \n(mostafa.karim86@gmail.com) \n \nPurpose or Learning Objective: The purpose of this study is the exploration \nof imaging biomarkers on acute phase cardiac CT tha t may suggest a \ncardioembolic etiology in patients with large vesse l occlusion (LVO) stroke in \npatients without intracardial thrombi or atrial fib rillation in an effort to further \nspecify and weigh the known major and minor cardioe mbolic risk factors. \nMethods or Background: A total of 317 patients with LVO stroke and acute-\nsetting one-step cardiac CT imaging examination wer e retrospectively \nidentified and included in this study. Images were assessed for a total of 15 \nspecific imaging findings according to known minor and major cardioembolic \nrisk factors. Final etiology (TOAST) of LVO stroke was determined by \ninterdisciplinary consensus after full clinical wor kup. Multivariate regression \nanalysis was performed to identify cardiac imaging findings associated with a \ncardioembolic etiology. \nResults or Findings: Overall, 221 (70%) of the LVO strokes were found to  \nhave a cardioembolic etiology by interdisciplinary consensus. After correction \nfor atrial fibrillation and intracardiac thrombi, m ultivariate regression analysis \ndefined dilatative cardiomyopathy (adjusted odds-ra tio (AOR) 37.9), right-to-left \nshunt (AOR 21.8), valvular implants (AOR 23.7), typ e II and III thrombotic \naortic arch (AOR 8.1) and visible myocardial scars (AOR 6.8) as risk factors for \na cardioembolic etiology (AUC 0.87, p < 0.05 for al l factors respectively). \nConclusion: In patients with LVO stroke without atrial fibrilla tion or intracardiac \nthrombi on acute phase cardiac CT imaging, the pres ence of dilatative \ncardiomyopathy, right-to-left shunt, valvular impla nts, type II and II thrombotic \naortic arch and visible myocardial scars on acute p hase CT imaging findings is \nsignificantly associated with a cardioembolic strok e etiology. \nLimitations: This is a retrospective study with its associated l imitations. The \nnumber of patients was low due to the single centre  design. \nFunding for this study: None. \nEthics committee - additional information: Ethikkommission der Christian \nAlbrechts Universität Kiel \n \n \n \n \n \n \n \n \nAuthor Disclosures:  \nCosima Wünsche: Nothing to disclose \nCarmen Wolf: Nothing to disclose \nPatrick Langguth: Nothing to disclose \nMarcus Both: Nothing to disclose \nKarim Mostafa: Nothing to disclose \nSarah Krutmann: Nothing to disclose \nOlav Jansen: Nothing to disclose \nHatim Seoudy: Nothing to disclose \n \n \nThe pericoronary adipose tissue attenuation in CT s trongly depends on \nkernels and iterative reconstructions \nC. Lisi¹, *K. Klambauer*², L. J. Moser², V. Mergen² , R. Manka², T. Flohr²,  \nM. Eberhard², H. Alkadhi²; ¹Milan/IT, ²Zürich/CH \n \nPurpose or Learning Objective: To investigate the influence of kernels and \niterative reconstructions on pericoronary adipose t issue (PCAT) attenuation in \ncoronary CT angiography (CCTA) \nMethods or Background: Twenty subjects (16 females; median age 52 years \n(IQR 48-61)) with atypical chest pain and low risk of coronary artery disease \n(CAD) who were otherwise healthy and without eviden ce of CAD in photon-\ncounting detector CCTA were included. In each subje ct images were \nreconstructed with a quantitative smooth (Qr36) and  three vascular kernels of \nincreasing sharpness levels (Bv36, Bv44, Bv56). Qua ntum iterative \nreconstruction (QIR) was either switched-off (QIR o ff) or was used with \nstrengths 2 and 4. The fat-attenuation-index (FAI) of the PCAT surrounding the \nright coronary artery was calculated in each datase t. Histograms of FAI \nmeasurements were created. Intra- and inter-reader agreement were \ndetermined. A CT edge-phantom was used to determine  the edge-spread-\nfunction (ESF) for the same datasets \nResults or Findings: Intra- and inter-reader agreement of FAI were excel lent \n(ICCs=0.99 and 0.98, respectively). Significant dif ferences in FAI were \nobserved depending on the kernel and iterative reco nstruction strength level \n(each, p < 0.001), with inter-individual variation up to 34HU. FAI showed also \nconsiderable intra-individual variation (average FA I difference 19HU, maximal \nintra-individual difference 33HU), also depending o n kernels and iterative \nreconstruction levels. The ESFs showed a reduced ra nge of edge-smoothing \nwith increasing kernel sharpness, causing FAI decre ase. Histogram analyses \nrevealed a narrower peak of PCAT values with increa sing iterative \nreconstruction levels, causing FAI increase \nConclusion: PCAT attenuation determined with CCTA heavily depen ds on \nkernels and iterative reconstruction levels both wi thin and across subjects. \nStandardization of CT reconstruction parameters is mandatory for FAI studies \nto enable meaningful interpretations \nLimitations: Single centre study, including only healthy patient s. Single vendor \nscanner and software analysis, \nFunding for this study: No funding was received for this study \nEthics committee - additional information: No additional information needed \nAuthor Disclosures:  \nThomas Flohr: Nothing to disclose \nCostanza Lisi: Nothing to disclose \nVictor Mergen: Nothing to disclose \nMatthias Eberhard: Nothing to disclose \nLukas Jakob Moser: Nothing to disclose \nRobert Manka: Nothing to disclose \nKonstantin Klambauer: Nothing to disclose \nHatem Alkadhi: Nothing to disclose \n \n \nInter-Observer Agreement of the Coronary Artery Dis ease-Reporting and \nData System (CAD-RADS) 2.0 \n*J. H. Lund*, J. Erley, G. Adam, E. Tahir, I. Molwi tz, M. Meyer; Hamburg/DE \n(lundjonas@gmx.de) \n \nPurpose or Learning Objective: In 2022, the Coronary Artery Disease-\nReporting and Data System (CAD- RADSTM) was updated  to standardize \ncoronary CT angiography (CCTA) reports, requiring h igh inter-observer \nreproducibility. This study aims to assess the inte r-observer agreement of \nCAD-RADS 2.0. \nMethods or Background: Patients who underwent CCTA between 2022 and \n2024 using a 3rd-generation-dual-source-CT were ind ependently evaluated by \nthree readers with varying levels of experience (1,  3, and 12 years). CAD-\nRADS 2.0 was used to assess visual grading of plaqu e burden (P1 = mild to \nP4 = extensive), stenosis degree (CAD-RADS 0 = no s tenosis to CAD-RADS 5 \n= total occlusion), and modifiers (HRP for high-ris k plaque features, E for \nexceptions, S for stents, G for grafts, and N for n on-evaluable studies). Inter-\nobserver agreement was measured using intraclass co rrelation coefficients \n(ICC). \n \n \n\n \n \nAbstract-based Programme \n \n 9  \nWednesday \nResults or Findings: 100 patients (29% female, age 63 ± 12 years) with a \nmedian Agatston score of 267 were included. Observe rs 1, 2, and 3 rated 71%, \n60%, and 64% of patients, respectively, as CAD-RADS  3 or above. Inter-\nobserver agreement for plaque burden grading (ICC:0 .92, 95% CI:0.88–0.94, p \n< 0.001) and stenosis degree (ICC:0.88, 95% CI:0.83 –0.92, p < 0.001) on a \nper-patient level was excellent. The agreement was also excellent on a per-\nvessel basis, with the highest for the left anterio r descending artery (ICC:0.90) \nand the lowest for the left main artery (ICC:0.86).  Agreement on modifiers was \npoor (ICC:0.06, p = 0.410). \nConclusion: CAD-RADS 2.0 demonstrates excellent inter-observer agreement \nfor plaque burden and stenosis grading, but agreeme nt on modifier use is low, \nlikely due to limited use or uncertainty in their a pplication in routine practice. \nLimitations: Limited assessment of the modifiers. Experienced ba sed-bias \ndue to limited amount of observer, thus the real-wo rld clinical experience might \nnot be entirely reflected. \nFunding for this study: Nothing to disclose. \nEthics committee - additional information: This study was accepted by the \nethics committee of the University Medial Center Ha mburg-Eppendorf (UKE). \nAuthor Disclosures:  \nGerhard Adam: Nothing to disclose \nIsabel Molwitz: Nothing to disclose \nMathias Meyer: Nothing to disclose \nJennifer Erley: Nothing to disclose \nEnver Tahir: Nothing to disclose \nJonas H. Lund: Nothing to disclose \n \n \nPreliminary Experience of 60-kVp Tube Voltage Combi ned with Deep \nLearning Reconstruction Algorithm in Coronary CT An giography \nX. Wu¹, S. Jiang², Y. Zou², *T. Wang*², G. Zhang², F. Huang¹, P. Liu¹, W. Sun¹, \nW. He¹; ¹Changsha/CN, ²Shanghai/CN \n(tiantian.wang@cri-united-imaging.com) \n \nPurpose or Learning Objective: To explore the clinical value of 60-kVp \ncoronary CT angiography (CCTA) combined with DEep L earning Trained \nAlgorithm (DELTA). \nMethods or Background: Thirty-nine patients (20 male, 58.77 ± 16.26 years, \n22.87 ± 3.71 kg/m^2) with suspected coronary artery disease (CAD) were \nprospectively enrolled. Each underwent both low- (6 0 kVp, 28 ml contrast \nmedium at 2.5 ml/s) and routine-dose CCTA (100 kVp,  44 ml contrast medium \nat 4.0 ml/s) on a 320-row scanner within 2 weeks. T he routine-dose data were \nreconstructed using hybrid iterative reconstruction  (RD-HIR) and served as the \nreference standard. Low-dose data were reconstructe d using both HIR (LD-\nHIR) and DELTA (LD-DELTA). Coronary stenosis in the  right coronary artery \n(RCA), left anterior descending (LAD), and left cir cumflex (LCX) was assessed \nusing CAD-Reporting and Data System (CAD-RADS) scor es. The diagnostic \nperformance of LD-HIR and LD-DELTA in distinguishin g moderate (CAD-\nRADS<3) to severe (CAD-RADS≥3) stenosis was analyzed via receiver \noperating characteristic analysis. Signal-noise-rat io (SNR) and contrast-noise-\nratio (CNR) on each vessel were also compared. \nResults or Findings: The low-dose CCTA reduced radiation dose by 85.8% \ncompared to the routine-dose acquisition (0.55 ± 0. 09 mSv vs. 3.86 ± 1.25 \nmSv, p<0.001). In distinguishing moderate to severe  stenosis, LD-DELTA \ndemonstrated superior diagnostic performance compar ed to LD-HIR, with area \nunder the curve (AUC) being 1.00 (95% CI, 0.91-1.00 ) versus 0.87 (95% CI, \n0.72-0.96) in the RCA, and 1.00 (95% CI, 0.91-1.00)  versus 0.78 (95% CI, \n0.62-0.90) in the LCX. However, no difference was f ound in the LAD, with both \nshowing an AUC of 0.98 (95% CI, 0.88-1.00). Additio nally, LD-DELTA \ndemonstrated higher SNRs and CNRs compared to LD-HI R (all p<0.001). \nConclusion: The 60-kVp low-dose CCTA acquisition with DELTA sig nificantly \nreduces radiation dose while maintaining diagnostic  performance for assessing \ncoronary stenosis. \nLimitations: Not applicable. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was approved by the \nlocal institutional review board. \nAuthor Disclosures:  \nPeng Liu: Nothing to disclose \nTiantian Wang: Investigator: at Central Research In stitute, United Imaging \nHealthcare \nYixuan Zou: Investigator: Central Research Institut e, United Imaging \nHealthcare \nFeng Huang: Nothing to disclose \nGuozhi Zhang: Investigator: Central Research Instit ute, United Imaging \nHealthcare \nWeiling He: Nothing to disclose \nWenjie Sun: Nothing to disclose \nSenyang Jiang: Investigator: Central Research Insti tute, United Imaging \nHealthcare \nXi Wu: Nothing to disclose \n \n \nQuantitative assessment of early changes in myocard ial extracellular \nvolume during postoperative adjuvant chemotherapy i n breast cancer \npatients by dual-layer spectral detector CT \n*H. Wu*, Y. Huang, W. Deng, Y. Wang, Q. Xiao, Y. Gu ; Shanghai/CN \n(wuhonglinsimple@yeah.net) \n \nPurpose or Learning Objective: This study utilized dual-layer spectral \ndetector CT (DLCT) technology to simultaneously ass ess cardiac function and \nmyocardial myocardial extracellular volume (ECV) , characterizing the changes \nin parameters early during breast cancer therapy. \nMethods or Background: Forty female breast cancer patients who underwent \npostoperative adjuvant chemotherapy were prospectiv ely enrolled and \nunderwent baseline and 3-month-postchemotherapy car diac CT (CCT) and \nultrasound cardiography (UCG). Global ECV of the le ft ventricle (LV) were \nmeasured based on an iodine map of the late enhance ment phase of DLCT. \nChanges in cardiac function parameters and global E CV from baseline to the \n3-month follow-up were analyzed. Correlation coeffi cients between the \nchanges in cardiac function parameters and global E CV were calculated. \nResults or Findings: LV ejection fraction by UCG (UCG-LVEF) and by CCT \n(CCT-LVEF) did not significantly change between bas eline and 3 months. \nHeart rate (HR) increased over 3 months of follow-u p. After normalization to \nbody surface area (BSA), cardiac output (CCT-CO ind exed) and LV late \n(active) filling volume (LVLFV indexed) significant ly increased (P<0.01), while \nLV early (passive) filling volume (LVEFV indexed) a nd LVEFV/LVLFV \ndecreased significantly at the 3-month follow-up (P <0.05). Global ECV were \nelevated significantly at 3 months (25.4±2.4 vs. 27.3±2.7, P<0.01). Although \nchanges in global ECV were not associated with chan ges in LVEFs, global \nECV change were moderately correlated with changes in LV end-diastolic \nvolume / BSA (CCT-LVEDV indexed) (r=0.52, P<0.01), LV stroke volume / \nBSA (CCT-LVSV indexed) (r=0.56, P<0.01), CCT-CO ind exed (r=0.40, P=0.01) \nand LVEFV indexed (r=0.41, P<0.01) . \nConclusion: CCT-derived ECV can be used to evaluate myocardial changes \nin the early stage of chemotherapy before LVEF sign ificantly decreases. The \nincreases in global ECV were not correlated with LV EFs. The changes in \nmyocardial global ECV were moderately correlated wi th cardiac function \nparameters. \nLimitations: The small sample. \nFunding for this study: None \nEthics committee - additional information: The study was approved by the \nEthics Committee of Fudan University Shanghai Cance r Center, and written \ninformed consent was obtained from all subjects in the study. \nAuthor Disclosures:  \nHonglin Wu: Nothing to disclose \nYan Huang: Nothing to disclose \nYajia Gu: Nothing to disclose \nYu Wang: Nothing to disclose \nQin Xiao: Nothing to disclose \nWeiwei Deng: Nothing to disclose \n \n \nCardiac calcifications detected on planning CT are major predictors of \nlong-term cardiotoxicity after radiotherapy for bre ast cancer \nK. B. Dimayuga, *A. Belardo*, L. Perna, A. Fodor, P . Mangili, A. Del Vecchio, \nN. Di Muzio, C. Fiorino; Milan/IT \n(belardo.alfonso@hsr.it) \n \nPurpose or Learning Objective: Breast cancer (BC) patients undergoing \nradiotherapy (RT) may experience long-term cardioto xicity. In modern series, \ndelivering low dose to the heart, non-dosimetry pre dictors are emerging. The \npurpose was to test if cardiac calcifications (CAC)  at planning CT, suggested \nas potential predictors, are associated with long-t erm cardiac events. \nMethods or Background: Planning CT and clinical information of 1172 \nconsecutive patients treated at our hospital (2009- 2017) were available \n(right:569, left:603). The heart of all patients wa s automatically segmented \nusing a previously validated AI-based tool (MIM Pro tegé & MIM assistant) and \nthe mean heart dose (MHD) was assessed. CAC were au tomatically extracted \nby applying a home-made, validated, Python script e xtracting the Agatson \nscore (AS) and the CAC overall volume. Their associ ation with the risk of \ncardiac events was tested by logistic regression, i ncluding the potential \ncombined effect of MHD and available clinical param eters. \nResults or Findings: With a median follow-up of 8 years (range: 5-15), 3 2 \npatients experienced cardiac events. AS/CAC volumes  were the most \nsignificant predictors (p<0.0001), with similar per formances. Age, laterality \n(left/right), concomitant chemotherapy, obesity and  hypertension were also \nsignificant at univariate analysis. MHD encoded usi ng the best cut-off (1Gy, \nmostly representing laterality) was also predictive . The best multivariate model \ncombined MHD>1Gy, age and CAC volume (AUC=0.79, p<0 .0001, calibration \nplot: m=1.506, q=-0.007), being CAC volume the stro ngest predictor (OR: \n1.0008/mm3, p<0.0001). \nConclusion: CAC load was the most important factor in cardiac r isk \nstratification after BC RT in a modern series. \n\n \n \nAbstract-based Programme \n \n 10  \nWednesday \nLimitations: Events are not recovered from a registry. Then, the  risk of \nmissing events is not negligible. \nFunding for this study: None \nEthics committee - additional information: All respected. \nAuthor Disclosures:  \nLucia Perna: Nothing to disclose \nNadia Di Muzio: Nothing to disclose \nClaudio Fiorino: Nothing to disclose \nAlfonso Belardo: Nothing to disclose  \nAntonella Del Vecchio: Nothing to disclose \nPaola Mangili: Nothing to disclose \nKerby Bjorn Dimayuga: Nothing to disclose \nAndrei Fodor: Nothing to disclose \n \n \nOne-Scan Acquisition of Coronary CT Angiography and  CT Aortography \nUsing Photon Counting Detector CT \n*H. Kato*, S. Araki, S. Nakamura, A. Yamazaki, N. K ato, Y. Ichikawa,  \nH. Sakuma, K. Kitagawa; Tsu/JP \n \nPurpose or Learning Objective: For patients with aortic diseases, Coronary \nCT angiography (CCTA) is performed for preoperative  evaluation of coronary \nartery disease (CAD), often in combination with CT aortography (CTAO) within \nthe same examination. However, the shared imaging f ield in both CCTA and \nCTAO results in overlapping radiation exposure to t he thoracic region. \nRecently, a dual-source photon-counting detector CT  (PCD-CT) has emerged, \nwith its capability to perform high-pitch helical s canning to enable one-scan \nacquisition of CCTA and CTAO. This study aimed to c ompare the radiation \ndose and image quality of one-scan CCTA and CTAO im aging with those of \nseparate CCTA and CTAO imaging. \nMethods or Background: This study included 40 patients who underwent \nCCTA and CTAO for preoperative CAD screening for ao rtic disease: 22 had \nseparate, 18 had one-scan CCTA and CTAO. A low tube  potential of 70 or 90 \nkVp was used for all patients. CT dose-length produ ct (DLP) data for CCTA \nand CTAO was collected. Image quality for each coro nary segment and the \naorta was assessed using a four-point scale (excell ent, good, fair, non-\ndiagnostic). \nResults or Findings: The mean DLP was significantly lower with the one-s can \nCCTA and CTAO protocol (160.5±36.5 mGy*cm) compared  to the separate \nscans (716.0±203.9 mGy*cm) (P<0.001). There was no significant difference \nbetween the separate and one-scan protocols in the assessment of image \nquality for CCTA (P=0.17) and CTAO (P=0.92), and in  both protocols, over \n95% segments/cases in the CCTA and CCAO images were  rated as having \ngood or excellent image quality. \nConclusion: PCD-CT, with high-pitch helical scanning and low tu be voltage, \nenabled one-scan acquisition of CCTA and CTAO, sign ificantly reducing \nradiation exposure while maintaining high image qua lity. \nLimitations: There is no reference standard such as coronary ang iography. \nFunding for this study: No funding was provised for this study. \nEthics committee - additional information: Clinical Research Ethics Review \nCommittee of Mie University Hospital (approval No. H2019-207) \nAuthor Disclosures:  \nKakuya Kitagawa: Nothing to disclose \nHajime Sakuma: Nothing to disclose \nSuguru Araki: Nothing to disclose \nYasutaka Ichikawa: Nothing to disclose \nNoriyuki Kato: Nothing to disclose \nHiroaki Kato: Nothing to disclose  \nAkio Yamazaki: Nothing to disclose \nSatoshi Nakamura: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n08:00-09:30 Research Stage 3 \nResearch Presentation Session: Oncologic \nImaging \nRPS 116 \nStaging, metastases and response \nassessment \n \nModerator \nR. Perez-Lopez; Barcelona/ES  \n \n \nUndifferentiated pleomorphic sarcoma: Building an e ffective multi-\nparametric MRI (mpMRI) predictive treatment respons e model to replace \nRECIST \n*R. F. Valenzuela*, B. Amini, E. Duran-Sierra, J. E . Madewell, M. Antony,  \nC. M. Costelloe, W. Murphy; Houston, TX/US \n(raulvp@estigia.cl) \n \nPurpose or Learning Objective: Undifferentiated pleomorphic sarcoma (UPS) \nis the largest soft-tissue sarcoma subgroup. Post-t herapeutically, UPS \ndemonstrates hemosiderin deposition, fibrosis, and calcification. This study \naimed to establish the clinical value of multiparam etric MRI (mpMRI) for \npredicting UPS response. \nMethods or Background: An IRB-approved retrospective study included 33 \nextremity UPS patients with pre-operative mpMRI, in cluding diffusion-weighted \nimaging (DWI), contrast-enhanced susceptibility-wei ghted imaging (CE-SWI), \nand perfusion-weighted imaging with dynamic contras t-enhancement \n(PWI/DCE), and surgical resection February 2021-May  2023. Lesions were \nvisually classified on CE-SWI into one of 6 morphol ogy patterns. On PWI/DCE, \nlesions were classified into one of 6 patterns, and  time-intensity curves (TICs) \nwere classified as types I-V. Patients were divided  into three groups based on \nthe percentage of pathology-assessed treatment effe ct (PATE) in the surgical \nspecimen: Responders (>=90% PATE, n=16), partial-re sponders (31-89% \nPATE, n=10), and non-responders (<=30% PATE, n=7). Receiver operating \ncharacteristic (ROC) analysis of classification mod els based on CE-SWI and \nPWI/DCE patterns and TICs compared responders vs. p artial/non-responders. \nResults or Findings: At post-radiation therapy (PRT), a CE-SWI Complete \nRing pattern was observed in 71% of responders (p=7 .71x10-6). On PWI/DCE \nimages, 79% of responders displayed a Capsular patt ern (p=1.49x10-7), and \n100% demonstrated a TIC-type II (p=8.32x10-7). RECI ST could not separate \nresponders from partial/non-responders; all demonst rated 100% stability at \nPRT and pseudoprogression at PC. ROC analysis compa ring responders \n(n=14) vs. partial/non-responders (n=16) at PRT sho wed that the model \ncombining the PWI/DCE TIC-type II, PWI/DCE Capsular  pattern and CE-SWI \nComplete Ring pattern yielded the highest classific ation performance \n(AUC=0.99). \nConclusion: mpMRI-derived features can help assess UPS treatmen t \nresponse. Observing a pre-operative PWI/DCE TIC-typ e II, PWI/DCE Capsular \npattern, and CE-SWI Complete Ring pattern can poten tially predict \nsuccessfully treated UPS patients with >=90% PATE, outperforming RECIST. \nLimitations: Limitations include a small sample (n=33) and manua l, time-\nconsuming tumor VOI segmentation. \nFunding for this study: The John S. Dunn, Sr. Distinguished Chair in \nDiagnostic Imaging. \nM.R Evelyn Hudson Foundation Endowed Professorship.  \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nElvis Duran-Sierra: Nothing to disclose \nBehrang Amini: Nothing to disclose \nColleen M Costelloe: Nothing to disclose \nRaul Fernando Valenzuela: Nothing to disclose \nWilliam Murphy: Nothing to disclose \nMathew Antony: Nothing to disclose \nJohn Edward Madewell: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 11  \nWednesday \nRedefining radiologic responses in high-risk soft-t issue sarcomas treated \nwith neoadjuvant chemotherapy. Final results of ISG -STS 1001, a \nrandomized clinical trial \n*A. Vanzulli*, R. Vigorito, C. Buonomenna, P. Verde rio, S. Pasquali,  \nP. G. Casali, C. Morosi, S. Stacchiotti, A. Gronchi ; Milan/IT \n(andrea.vanzulli@unimi.it) \n \nPurpose or Learning Objective: We report the results of the pre-planned \nsecondary analysis of radiologic responses (RR) of ISG-STS 1001, a \nrandomized clinical trial comparing anthracycline +  ifosfamide (AI) vs. \nhistology-tailored (HT) neoadjuvant chemotherapy fo r primary localized high-\nrisk soft-tissue sarcomas of the extremities/trunk wall. \nMethods or Background: Patients with undifferentiated pleomorphic sarcoma \n(UPS), leiomyosarcoma (LMS), malignant peripheral n erve sheath tumor, \nsynovial sarcoma or myxoid liposarcoma (MLPS) were randomized, whereas \npatients with myxofibrosarcoma, pleomorphic liposar coma, pleomorphic \nrhabdomyosarcoma or unclassified sarcoma were alloc ated in the \nobservational arm (O) and treated with AI. Patients  with UPS, LMS or MLPS \nneeding concurrent preoperative radiotherapy were i ncluded in O. We \nevaluated associations between: Disease-Free Surviv al (DFS)/Overall Survival \n(OS) and centrally reviewed RR, assessed with RECIS T 1.1 and as percent \ndimensional variation (D; both dichotomized and con tinuous); DFS/OS and \nhistology; RR and histology. \nResults or Findings: 435 patients were included (287 randomized, 148 \nobserved). The analysis of RR comprised 236 patient s (154 randomized, 82 \nobserved) with measurable disease and available for  central review. RECIST \nbest responses were: 28 (11.9%) partial response (P R), 195 (82.6%) stable \ndisease (SD) and 13 (5.5%) progressive disease (PD) . RECIST significantly \ncorrelated with DFS (PD-vs-PR: HR 8.18, 95% CI 2.96 -22.58; SD-vs-PR: HR \n2.96, 95% CI 1.30-6.75) and OS (PD-vs-PR: HR 12.61,  95% CI 3.40-46.84; \nSD-vs-PR: HR 4.24, 95% CI 1.34-13.47). The median v alue of D was -1.6%. \nPatients with D > -1.6% had worse clinical outcomes  than those with D < -1.6% \n(DFS: HR 1.73, 95% CI 1.19-2.50; OS: HR 1.86, 95% C I 1.21-2.86). D in \ncontinuous scale inversely correlated with DFS (HR 1.53, 95% CI 1.25-1.87) \nand OS (HR 1.78, 95% CI 1.41-2.25). \nConclusion: Dimensional variation in continuous scale predicted  the \nproportional efficacy of treatment irrespective of tumor histology. \nLimitations: Nothing to disclose. \nFunding for this study: Pharmamar® provided trabectedin for the HG-MLPS \ncohort. The study was partially funded through a Eu ropean Union grant \n(EUROSARC FP7 278472). In addition, the French site s were supported by \nNETSARC, LYRICAN (LYRICAN [INCA-DGOS-INSERM 12563])  and \nDEPGYN (RHU4). \nEthics committee - additional information: The trial protocol and all \namendments were approved by the independent ethics committee at each trial \ncenter. \nAuthor Disclosures:  \nSandro Pasquali: Nothing to disclose \nPaolo Giovanni Casali: Nothing to disclose \nPaolo Verderio: Nothing to disclose \nAlessandro Gronchi: Nothing to disclose \nAndrea Vanzulli: Nothing to disclose \nCarlo Morosi: Nothing to disclose \nCiriaco Buonomenna: Nothing to disclose \nRaffaella Vigorito: Nothing to disclose \nSilvia Stacchiotti: Nothing to disclose \n \n \nCan a fast T2-Dixon sequence surpass the time obsta cle of whole-body \nMRI in the evaluation of skeletal metastases? \nN. Magdi, *M. Elmansy*, M. Elhawary, A. Sultan; Man soura/EG \n \nPurpose or Learning Objective: Our study was conducted to elucidate the \nrole of the T2-Dixon sequence as a rapid alternativ e to the standard Whole-\nbody magnetic resonance imaging (WB-MRI) protocol w ith the assessment of \nits diagnostic accuracy and comparability to the es tablished methodology. \nMethods or Background: This prospective study included 30 patients with \nprimary solid malignancies who underwent WB-MRI. Th e sequences obtained \nwere T1WI, STIR, and T2-Dixon (fat-only and water-o nly images). Skeletal \nmetastases were evaluated in each sequence. Results  were compared \nbetween the T1-STIR combination and T2-Dixon fat an d water reconstructions. \nResults or Findings: The sensitivity of fat and water reconstructions fr om a \nsingle T2-Dixon in the detection of lytic skeletal metastases was marginally \nsuperior to a combination of T1WI and STIR sequence s (0-7%). Detection of \nmixed lesions demonstrated equally high sensitivity  in both protocols. Sclerotic \nmetastases detection in WB-MRI showed low sensitivi ty in both protocols. \nHowever, specificity surpassed 95% for all lesion t ypes in both protocols. \nOverall image quality was favored (in 87-90% of pat ients ) in T2-Dixon images. \nThe overall estimated acquisition timing using T2-D ixon appeared to be \napproximately half that of the standard T1-STIR com bination. \nConclusion: WB-MRI using T2-Dixon fat and water reconstructions  showed \nsimilar accuracy to T1WI and STIR combination in th e evaluation of skeletal \nmetastases in patients with primary solid cancers w ith significantly shorter \nacquisition time. \nLimitations: Few skull and humeri lesions with limited assessmen t of the \nsclerotic lesions due to high fals negative results . \nLack of quantitative analysis of signal to noise an d contrast to noise ratio. \nFunding for this study: No funding \nEthics committee - additional information: This study was approved by the \nResearch Ethics Committee of the Faculty of Medicin e at Mansoura University \nin Egypt on 10 /1 /2023; reference number of approv al: MS.22.12.2250 \nAuthor Disclosures:  \nAmina Sultan: Nothing to disclose \nNoha Magdi: Nothing to disclose \nMohammed Elhawary: Nothing to disclose \nMostafa Elmansy: Nothing to disclose \n \n \nImpact of reader experience on reader agreement for  whole-body MRI \nstaging of oesophageal cancer \n*P. Chapellier*¹, S. W. Soo², O. Westerland², A. Gr een², S. Gourtsoyianni³,  \nV. Goh²; ¹Lausanne/CH, ²London/UK, ³Athens/GR \n \nPurpose or Learning Objective: Whole-body MRI (WB-MRI) may be valuable \nalternative to standard imaging pathways for stagin g. We evaluated how \nreader experience impacts agreement for tumour-node -metastasis (TNM) \nstaging of oesophageal cancer. \nMethods or Background: Following ethical approval, prospective patients \nunderwent WB-MRI (T2-weighted, diffusion-weighted, T1-weighted post-\ncontrast) alongside standard imaging (contrast-enha nced CT, 18F-FDG \nPET/CT ± EUS). WB-MRI was staged using AJCC TNMv8 b y four readers from \ndifferent countries with different levels of experi ence: subspecialty vs. non-\nsubspecialty trained; in-training vs. staff radiolo gists. Inter-reader agreement \nwas assessed using kappa statistics. For each reade r, agreement with a \nreference standard of final tumour board stage, sur gical pathology and clinical \nfollow up was obtained. \nResults or Findings: 29/30 (97%) patients had adenocarcinoma; 25/30 (83% ) \nhad ≥T3 stage; 27/30 (90%) had locoregional lymphadenopa thy; 12/30 (40%) \nwere metastatic. 22/30 (74%) received chemotherapy only; 8/30 (27%) had \nsurgery, 63% with neoadjuvant treatment. Compared t o reference standard, \nagreement for T- and N-stage was highest for the tw o gastrointestinal-trained \nradiologists (T-stage: κ =0.516, κ =0.824; N-stage: κ =0.434, κ =0.589, \nrespectively). Agreement for M-stage was highest fo r the oncology-trained \nradiologist (κ =0.795). Detection of lung metastases was limited on MRI. Inter-\nreader agreement was also highest for gastrointesti nal-trained radiologists (T-\nstage: κ = 0.624; N-stage: κ = 0.822). Agreement across TNM staging was \nlowest with the in-training radiologist who had no exposure to WB-MRI staging. \nConclusion: WB-MRI could be valuable alternative for initial TN M staging of \noesophageal cancer, but reliable interpretation app ears to be related to \nsubspecialty experience and level of training. \nLimitations: WB-MRI was a relatively long acquisition, and not s uit all patients. \nNodal disease was not just based on size measuremen t. MRI is limited for \nsome sites eg.lung \nFunding for this study: This project was supported by the National Institut e \nfor Health and Care Research (MIHR) Biomedical Rese arch Centre at Guy’s & \nSt Thomas’ Hospitals and King’s College London. \nEthics committee - additional information: The study was approved by the \nResearch Ethics Committee (IRAS ID 107508, 12/LO/17 54). \nAuthor Disclosures:  \nOlwen Westerland: Nothing to disclose \nAdrian Green: Nothing to disclose \nSofia Gourtsoyianni: Nothing to disclose  \nPauline Chapellier: Nothing to disclose \nVicky Goh: Nothing to disclose \nSuet Woon Soo: Nothing to disclose \n \n \nObserver variability and reproducibility of bone ma rrow metastasis \nbiomarkers on MRI \n*C. Sattin*¹, C. Pizzi¹, M. Kosmin², W. McGuire³, A . Makris², N. J. Taylor²,  \nG. Petralia¹, A. R. R. Padhani³; ¹Milan/IT, ²London /UK, ³Northwood/UK \n(caterina.sattin@unimi.it) \n \nPurpose or Learning Objective: To document inter- and intra-observer \nvariability and test-retest reproducibility of quan titative MRI metastasis \nbiomarkers. \nMethods or Background: Whole-body MRI was performed in women with \nbone-predominant metastatic breast cancer on a 1.5T  MRI system. Paired \nreproducibility scans were done in 14 patients afte r repositioning. Observer \nvariability was assessed in 10 patients after a two -week washout period. Two \ntrained radiologists delineated up to 5 bone metast ases per patient. \nQuantitative biomarkers (lesion sizes, apparent dif fusion coefficient (ADC), \nrelative fat fraction (rFF%) and b-900 s/mm2 signal -to-muscle ratio [SMR]) for \neach lesion were averaged per patient. Data logn tr ansformed after testing for \n\n \n \nAbstract-based Programme \n \n 12  \nWednesday \nnormality. Within-patient coefficient of variation (wCV) and variance ratio were \ncalculated. The repeatability (R-value; mean 95% co nfidence) for a single \npatient was calculated. \nResults or Findings: Reproducibility analysis (2 radiologist consensus; 14 \npatients; 41 lesions): Variance ratios were >15 for  all biomarkers. Size had \nwCV of 3.5% and R-value of 9.5%; SMR had wCV of 7.4 % and R-value of \n19.8%; ADC had wCV of 2% and R-value of 5.2%; rFF% had wCV of 9.4% and \nR-value of 24.9%. Inter- and intra-observer variabi lity (for 2 independent \nradiologists; 10 patients; different lesions) were similar (ICC >0.8) indicating \nconsistent reader performance. \nConclusion: Higher reproducibility test variance ratios and int erclass \ncorrelations of clinically relevant biomarkers indi cate the reliability of \nradiological assessments regardless of the observer . Changes in aggregated \nlesion size of >3mm, SMR of > 1.4 au, ADC of >53 µm 2/s and rFF% of > 2.7% \ncan be used as bone tumour response biomarkers in w omen with metastatic \nbreast cancer. \nLimitations: This abstract represents a preliminary part of a la rger study that \naims to evaluate whether changes in quantitative an d qualitative WB-MRI \nbiomarkers could be predictive of outcomes in patie nts with bone-predominant \nMBC on SACT. \nFunding for this study: This work was funded by the Paul Strickland Scanner  \nCentre Charity (UK registered charity number 298867 ) and the Fighting Breast \nCancer (UK registered charity number 1091882). \nEthics committee - additional information: All procedures permorfed in this \nstudy (ClinicalTrials.gov identifier: NCT03266744) followed the ethical \nstandards of the National Health Service Health res earch Authority East of \nEngland-Cambridge East Research Committee and the 1 964 Helsinki \nDeclaration and its later amendments or comparable ethical standards. \nAuthor Disclosures:  \nN. Jane Taylor: Nothing to disclose \nMichael Kosmin: Nothing to disclose \nGiuseppe Petralia: Nothing to disclose \nWill McGuire: Nothing to disclose \nCaterina Pizzi: Nothing to disclose \nAnwar R. R. Padhani: Nothing to disclose \nCaterina Sattin: Nothing to disclose \nAndreas Makris: Nothing to disclose \n \n \nSpectral CT imaging for assessment of metastases in  melanoma patients: \nMulti-reader evaluation \n*C. Nelles*, P. Rauen, T. M. Dratsch, D. Maintz, J.  Kottlors, N. Große Hokamp, \nD. Zopfs, T. Persigehl, S. Lennartz; Cologne/DE \n \nPurpose or Learning Objective: To investigate the sensitivity, specificity and \nqualitative assessment of spectral image reconstruc tions for metastases in \nmelanoma patients in a large-scale, multi-reader ev aluation. \nMethods or Background: In total, 308 patients with melanoma, 95 patients \nwith metastases and a control group of 213 patients  without metastases, who \nunderwent staging CT of the chest and abdomen on a dual-layer dual-energy \nCT system (dlDECT) were retrospectively included. C onventional images (CI), \niodine overlays (IO) and virtual monoenergetic imag es at 40 keV (VMI40keV) \nwere reconstructed. Six radiologists (three experie nced, three less \nexperienced) evaluated all cases in a CI-based sess ion and a session based \non a combination of CI, IO and VMI40keV. Readers we re asked to binarily \ndetermine presence of metastases in specific tissue s and to indicate diagnostic \ncertainty and lesion delineation on 5-point Likert scales. \nResults or Findings: Sensitivity for detection of metastases in skeletal  muscle \nand peritoneum was significantly higher for the spe ctral assessment (for \nskeletal muscle 70% vs. 61%; for peritoneum 76% vs.  62%, both: p < 0.05). \nFor subcutaneous metastases, there was a significan t increase in specificity \n(92% vs. 89%, p < 0.05), however accompanied with a  significant decrease in \nsensitivity (79% vs. 85%, p < 0.05). Diagnostic cer tainty was rated significantly \nhigher for spectral images than CI for 100% (6/6) o f the assessed tissues, \nwhereas improvements in lesion delineation were not ed for skeletal muscle, \nsubcutaneous tissue and pancreas. \nConclusion: In melanoma patients, the benefit of dlDECT-derived  spectral \nreconstructions depends on the assessed tissue. Whi le assessment of skeletal \nmuscle and peritoneal metastases was significantly improved, low or absent \niodine uptake of subcutaneous lesions led to false negatives and a consecutive \ndecrease in sensitivity. \nLimitations: -Retrospective, monocenter study design -The result s may not be \ngeneralisable to all different DECT platforms \nFunding for this study: Funding was provided by the German Research \nFoundation (DFG, FI 773/15-1). \nEthics committee - additional information: After reviewing the study design, \nthe institutional review board waived the need for informed patient consent for \nthis retrospective single-center study. \n \n \n \n \nAuthor Disclosures:  \nSimon Lennartz: Speaker: Amboss GmbH Author: Amboss  GmbH \nNils Große Hokamp: Consultant: Bristol-Myers Squibb ; BeiGene Speaker: \nPhilips Healthcare; Amboss GmbH Research/Grant Supp ort: Philips \nHealthcare \nDavid Zopfs: Research/Grant Support: Philips Health care \nDavid Maintz: Nothing to disclose \nChristian Nelles: Nothing to disclose \nPhilip Rauen: Nothing to disclose \nJonathan Kottlors: Nothing to disclose \nThomas Markus Dratsch: Nothing to disclose \nThorsten Persigehl: Nothing to disclose \n \n \nAccuracy of DOTATATE PET CT versus DOTATATE PET MR in the \nAssessment of Multifocal Small Bowel Neuroendocrine  Neoplasm \n*A. Keane*¹, H. Takahashi², C. Zhang³, C. Thiels², P. Navin²; ¹Dublin/IE, \n²Rochester, MN/US, ³Pheonix, AZ/US \n(annakeane@rcsi.ie) \n \nPurpose or Learning Objective: Small bowel neuroendocrine neoplasms \n(sbNENs) provide challenges in diagnosis and manage ment. Current gold \nstandard treatment is open resection with manual sm all bowel palpation as 45-\n50% of sbNEN have synchronous lesions. Advancements  in nuclear medicine \nhave improved preoperative assessment, potentially negating manual \npalpation and allowing for laparoscopic approach. D OTATATE PET CT (DPCT) \nand PET MR can identify somatostatin receptors, fre quently expressed on \nsbNENs. We aim to assess the accuracy of DPCT versu s PET MR in \ndetermining multifocality of sbNENs. \nMethods or Background: Multicentre retrospective analysis was performed \non patients with sbNEN who underwent open small bow el resection between \nJanuary 2016 to August 2022 and had either preopera tive DPCT or PET MR. \nBlinded retrospective review of images for small bo wel lesions was performed \nby two fellowship-trained radiologists and compared  to postoperative pathology \nreports. Final radiology diagnosis was attained by consensus agreement. \nDescriptive statistics, sensitivity, specificity, p ositive predictive value (PPV), \nnegative predictive value (NPV), accuracy, and area  under the receiver \noperator characteristic curve (AUROC) of DPCT and P ET MR were compared. \nResults or Findings: Seventy-eight patients met inclusion criteria. Fift y-six \n(71.8%) had preoperative DPCT, twenty-two (28.2%) h ad PET MR. Forty-eight \npatients (61.5%) had multiple (> 1)sbNENs on final pathology. Thirty-six \n(75.0%) were identified on preoperative imaging. DP CT and PET MR \ndemonstrated sensitivity of 85.0% and 58.0% respect ively, specificity of 81.0% \nand 60.0%, PPV of 87.9% and 63.6%, NPV of 77.3% and  54.6%, and accuracy \nof 83.6% and 59.0% for the presence of multifocal d isease. AUROC was 0.8 \nfor DPCT and 0.5 for PET MR. \nConclusion: DPCT demonstrated higher accuracy in identifying mu ltifocal \nsbNEN versus PET MR. However, as the NPV is insuffi cient, we conclude that \npreoperative DPCT should not replace open explorati on. \nLimitations: Retrospective study, underpowered due to rarity of sbNENs. \nFunding for this study: Nil \nEthics committee - additional information: IRB (Institutional Review Board) \nAuthor Disclosures:  \nCornelius Thiels: Nothing to disclose \nHiroaki Takahashi: Nothing to disclose \nAnna Keane: Nothing to disclose \nChi Zhang: Nothing to disclose \nPatrick Navin: Nothing to disclose \n \n \nDiscord Dilemmas in Lung Cancer Clinical Trials: Na vigating Reader \nVariability in Response Assessment \n*H. Beaumont*¹, R. Gill², N. Faye¹, A. Iannessi¹; ¹ Valbonne/FR,  \n²Boston, MA/US \n(hubertbeaumont@hotmail.com) \n \nPurpose or Learning Objective: In lung cancer trials, blinded independent \ncentral response assessment with double reads is ch allenging and prone to \ninterobserver variability. We analyzed the patterns  of discordance in reporting \nProgressive Disease (PD) and performed a root cause  analysis. \nMethods or Background: We retrospectively analyzed data from five clinical  \ntrials evaluating 1932 lung cancer patients treated  with targeted and immune \ntherapies, read by 17 central readers. Progressive Disease was defined based \non RECIST 1.1 criteria. The RECIST components were the Sum of tumors \nDiameter (SOD), the unequivocal progression of the non-Target Lesions (NTL) \nand the detection of New Lesion (NL). We analyzed t he RECIST components \naccording to 1) Concordant/discordant PD detection;  2) Positive Predictive \nValue (PPV) of declaring PD; 3) Offset versus singl e reader detection. \nResults or Findings: Discordance in PD was observed in 39.2% (675/1718) of \npatients, with adjudication of PD for 62.5% (422/67 5) based on 44.8% (95%CI: \n39.9, 49.8) new lesion detection, 28.3% (95%CI: 24. 0, 33.0) significant \nincrease of SOD and 12.6% (95%CI: 9.5, 16.2) unequi vocal non-target \n\n \n \nAbstract-based Programme \n \n 13  \nWednesday \nprogression. For 54.2% of concordant PD, at least o ne reader involved more \nthan one RECIST component. The PPV for increased SO D was 0.59, rising to \n0.89 when multiple RECIST components were involved.  In 49.2% of discrepant \ncases, PD was reported with a delay of one cycle in  majority of cases (80%). \nConfirmation rate for NLs in the lungs was lowest ( 40.6%) and new nodal \nlesions was highest (88.4%). \nConclusion: Discordance among trained central readers in lung c ancer trials \nis common. New lesion detection is pivotal in the d etection of PD, also the \nmain cause of discordances. Involving multiple RECI ST components improves \nthe reliability of assessments. When relying on NL only, the detection of extra \npulmonary lesion is more reliable. \nLimitations: No limitations \nFunding for this study: No funding \nEthics committee - additional information: Median Technologies institutional \nethics committee approved the study, informed conse nt was not required for \nthis retrospective analysis \nAuthor Disclosures:  \nNathalie Faye: Consultant: Median Technologies \nMadam Ritu Gill: Nothing to disclose \nAntoine Iannessi: Employee: Median Technologies \nHubert Beaumont: Employee: Median Technologies \n \n \nDetection of abdominal metastases in brain tumor pa tients following \nventriculoperitoneal shunting \n*N. Plakhotina*, A. V. Smirnova, K. Boiko, V. Bikul ov, P. Ivanov;  \nSaint-Petersburg/RU \n(plahotinadezhda@gmail.com) \n \nPurpose or Learning Objective: To develop a diagnostic algorithm and \nidentify imaging patterns for metastatic abdominal lesions in children with brain \ntumors following shunt placement. \nMethods or Background: Clinical cases of shunt-associated metastasis of \nprimary brain tumors in children, identified throug h CT, MRI, and laparoscopy, \nwith morphological confirmation. \nResults or Findings: Acute abdominal pain developed in children with \nembryonal tumors (2 patients with medulloblastoma, 1 patient with ATRT) and \n1 child with ependymoma in remission, as well as du ring chemotherapy. CT \nscans revealed multiple nodules of varying sizes wi thin the heterogeneous \nadipose tissue of the abdominal cavity. MRI identif ied multiple solid isointense \ntumors on both T1- and T2-WI, exhibiting significan t contrast enhancement and \ndiffusion restriction. Tumor sizes ranged from 3 to  15 mm. A notable feature \nwas the predominant spread in the interintestinal s paces, as well as the \nsubdiaphragmatic and subhepatic regions, which comp licates visualization. In \none of the cases, imaging did not yield conclusive results, and metastases \nwere confirmed only at autopsy. \nConclusion: Although shunt-associated metastasis in CNS tumors is \nextremely rare, it carries a very poor prognosis. D ynamic monitoring of children \nwith brain tumor with a shunt system should include  abdominal examinations \n(ultrasound or MRI). The development of an acute ab domen requires urgent \nevaluation to rule out tumor presence, including la paroscopy if indicated. \nReports of such cases suggest that ventriculoperito neal shunting for occlusive \nhydrocephalus should be considered only as a last r esort when temporary \nexternal drainage is not feasible. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nKonstantin Boiko: Nothing to disclose \nVyacheslav Bikulov: Nothing to disclose \nNadezhda Plakhotina: Nothing to disclose \nPavel Ivanov: Nothing to disclose \nAlina Vyacheslavovna Smirnova: Nothing to disclose \n \n \nImaging treatment response in High Grade Serous Ova rian Cancer: \nMetabolic imaging vs Cell death imaging \n*M. L. Chia*, K. Brindle; Cambridge/UK \n \nPurpose or Learning Objective: Given the poor therapy response noted in \nadvanced stage High Grade Serous Ovarian Cancer (HG SOC) patients and \nthe lack of fast and reliable treatment response mo nitoring methods, there is a \nneed to predict treatment response earlier. This pr oject investigated the \npotential of metabolic and cell death imaging techn iques to detect early \ntreatment response to standard-of-care chemotherapy  in HGSOC patients. \nMethods or Background: HGSOC cells, derived from the ascites of stage 3-4 \nHGSOC patients, were maintained as patient derived organoids(PDO) and \nimplanted into mice subcutaneously. The resulting t umours were imaged with \nvarious imaging techniques. Metabolic imaging techn iques included MRS \n(hyperpolarized [1-13C]pyruvate metabolism) and PET (measurements of 2-\ndeoxy-2-[fluorine-18]fluoro-D-glucose uptake). Cell  death imaging techniques \nincluded diffusion-weighted 1H MRI (DWI) and 2H MRS I measurements of \n[2,3-2H2]fumarate metabolism. PDO 2(carboplatin sen sitive) and PDO \n5(Carboplatin resistant) tumour models were treated  with i.v. Carboplatin \n(50mg/kg) or drug vehicle weekly, with imaging at b aseline and weekly \nthereafter. \nResults or Findings: Both metabolic imaging techniques were successful i n \ndiscriminating responding from non-responding tumou rs to Carboplatin before \nthere was a change in tumour volume. The techniques  for detecting cell death \nwere not as sensitive for detecting treatment respo nse, which may reflect a \nslow accumulation of dead cells post treatment, a l ack of knowledge of when \nthe rate of cell death increases post treatment and  immune clearance of dead \ncells. \nConclusion: Imaging with hyperpolarized [1-13C]pyruvate has the  potential to \nbe used in the clinic to detect the early treatment  response in HGSOC patients. \nLimitations: We only tested Carboplatin but other chemotherapies  or \ncombination treatment regimens would also be import ant for investigation as \nthey might produce a greater and faster increase in  cell death, possibly \nallowing cell death detection techniques to be more  successful. This will be \npart of future work. \nFunding for this study: Cancer Research UK Cambridge Institute Core \nfunding \nEthics committee - additional information: Na \nAuthor Disclosures:  \nMing Li Chia: Nothing to disclose \nKevin Brindle: Nothing to disclose \n \n \n08:00-09:30 Research Stage 4 \nResearch Presentation Session: \nMusculoskeletal \nRPS 110 \nAI, radiomics and other technologies \nsupporting MSK diagnostics \n \nModerator \nV. Mascarenhas; Lisbon/PT  \n(vmascarenhas@me.com) \n \n \nAI-Driven SuperResolution reconstruction for high-q uality, fast MR \nimaging of the lumbar spine: enhanced image clarity  for pathology \ndetection \n*R. Hahnfeldt*¹, R. A. Terzis¹, T. M. Dratsch¹, J. Bremm¹, P. Rauen¹, K. Weiss², \nD. Maintz¹, G. Bratke¹, A-I. Iuga¹; ¹Cologne/DE, ²H amburg/DE \n \nPurpose or Learning Objective: The aim of this study was to investigate \nwhether a 2D MRI lumbar spine protocol with an AI-b ased SuperResolution \nreconstruction method meets the requirements for cl inical diagnostic purposes. \nMethods or Background: In this retrospective study, 25 patients underwent \nMRI examinations of the lumbar spine using a 1,5T M RI scanner (Philips \nIngenia 1.5T, Best, NL). The MRI protocol included three sagittal sequences \n(STIR, T1 TSE, T2 TSE), and an axial T2 TSE sequenc e. The images were \nacquired in both standard and low resolution. Both the clinical standard \n(Compressed SENSE (CS)) and the new AI-based SuperR esolution \nreconstruction method (SuperRes-AI) were applied. F our experienced readers \n(two radiologists and two orthopedic surgeons) eval uated the sequences for \npathologies (bone marrow edema, neuroforaminal sten osis, disc herniation). \nResults or Findings: The acquisition time for the clinical standard sequ ences \nwas 11 minutes and 5 seconds. In contrast, the acqu isition time for the low \nresolution SuperRes-AI sequences was 7 minutes and 37 seconds (31% scan \ntime reduction). A generalized estimating equations  (GEE) analysis revealed \nno significant differences in the sensitivity for d etecting edema between reader \ngroups and reconstruction algorithms (all p>0.99). Bonferroni-corrected post-\nhoc tests in a GEE analysis revealed significantly higher sensitivity for \ndetecting neuroforaminal stenosis with AI-powered r econstruction compared to \nconventional algorithms among radiologists (p=0.001 ), with no other significant \ndifferences observed. \nConclusion: The new AI-based SuperResolution reconstruction of low-\nresolution 2D MRI sequences of the lumbar spine all ows for a reduction in \nacquisition time of approximately 31% without compr omising diagnostic quality, \nshowing significantly higher sensitivity for detect ing neuroforaminal stenosis. \nThe AI-based SuperResolution method improves MRI ef ficiency by significantly \nreducing scan times without compromising image qual ity, potentially enhancing \nsensitivity in pathology detection, offering advant ages for patient comfort and \nclinical workflow. \nLimitations: Not applicable. \n\n \n \nAbstract-based Programme \n \n 14  \nWednesday \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number DRKS00024156 \nAuthor Disclosures:  \nRobert Hahnfeldt: Nothing to disclose \nAndra-Iza Iuga: Nothing to disclose \nDavid Maintz: Nothing to disclose \nPhilip Rauen: Nothing to disclose \nJohannes Bremm: Nothing to disclose \nThomas Markus Dratsch: Nothing to disclose \nKilian Weiss: Employee: Philips GmbH Market DACH \nRobert Angelo Terzis: Nothing to disclose \nGrischa Bratke: Nothing to disclose \n \n \nAI-based Detection of Postoperative Abnormalities F ollowing Lumbar \nFusion Surgery in Spine Radiographs \nM. Kim¹, *J. Song*¹, K. Sung², E. Oh¹; ¹Seoul/KR, ² Los Angeles, CA/US \n(jmsong@promedius.ai) \n \nPurpose or Learning Objective: The purpose of this study is to develop a \ndeep learning-based system to detect postoperative abnormalities in spine \nradiographs following lumbar fusion surgery. This s ystem aims to assist \nradiologists by detecting postoperative abnormaliti es. \nMethods or Background: A total of 1,505 spine radiographs from 85 patients  \nwho underwent lumbar fusion surgery were collected at a secondary \nhealthcare facility between February 2018 and Janua ry 2022. These \nradiographs, taken post-operation and during follow -up visits, included \nanteroposterior, lateral, flexion, and extension vi ews. Annotations for \nperiprosthetic loosening, cage subsidence, and comp ression fracture were \nperformed by a musculoskeletal radiologist, and ver ified with CT scans. The \nCo-DETR model was trained on a subset of 634 radiog raphs from 74 patients \nwith 726 annotations. The class distribution includ ed 58, 24, and 17 patients \nyielding 278, 215, and 168 images respectively, wit h each image averaging \n1.10 annotations. Initial training was conducted on  a public dataset (FracAtlas), \nfollowed by transfer learning to enhance detection of postoperative \nabnormalities. Negative samples were included to bo ost training efficiency, and \nmodel performance was evaluated using mean Average Precision (mAP). \nResults or Findings: Periprosthetic loosening achieved an mAP score of \n0.601 with 0.5 IoU threshold. The mAP score for eac h class of periprosthetic \nloosening, cage subsidence, and compression fractur e were 0.565, 0.667, \n0.572, respectively. \nConclusion: The study demonstrates the potential of detecting p ostoperative \nabnormalities in spine radiographs after lumbar fus ion surgery using deep \nlearning. The results indicate a foundational poten tial for enhancing diagnostic \ncapabilities in clinical settings. The potential of  this approach to improve early \ndetection of complications could lead to more timel y interventions and better \npatient outcomes. \nLimitations: Further validation is required to optimize its perf ormance, \nparticularly to support radiologists in settings wi th limited access to specialists. \nFunding for this study: Not applicable \nEthics committee - additional information: IRB No. 2022-08-018 \nAuthor Disclosures:  \nJeongmin Song: Nothing to disclose \nMinjee Kim: Nothing to disclose \nKyunghyun Sung: Nothing to disclose \nEunsun Oh: Nothing to disclose \n \n \nPost-operative X-rays radiomics-based machine learn ing to predict two-\nyear clinical outcome in patients with lumbar spine  arthrodesis \n*I. C. Pizza*¹, M. Pedullà², S. Fusco², F. Serpi², D. Albano³, C. Messina²,  \nS. Gitto², L. M. Sconfienza²; ¹Eboli/IT, ²Milan/IT,  ³Cefalu'/IT \n(irene.pizza@unimi.it) \n \nPurpose or Learning Objective: The aim of this study is to predict two-year \nclinical outcome in patients with lumbar spine arth rodesis using machine \nlearning and radiomics based on post-operative X-ra ys. \nMethods or Background: This retrospective study was performed at a tertiar y \northopaedic centre and included 162 patients with l umbar spine arthrodesis, \npost-operative X-rays available for analysis and mi nimum follow-up of two \nyears. Clinical follow-up was evaluated at two year s using Oswestry Disability \nIndex (ODI): ODI≤20 indicated good clinical outcome (n=90), ODI>20 i ndicated \npoor clinical outcome (n=72). All X-rays were manua lly segmented by drawing \nrectangular regions of interest including the arthr odesis and one adjacent non-\noperated vertebra on both proximal and distal sides . Radiomic features were \nextracted. After feature selection and class balanc ing, machine learning (three \nensembles of Random Forest classifiers) was trained , validated using nested \n10-fold cross-validation and tested. \nResults or Findings: After training and cross-validation, in the test da taset \nmachine learning showed ROC-AUC (%) of 74 (majority  vote), 72.9** (mean) \n[confidence interval 69-76.7], accuracy (%) of 68 ( majority vote), 67.7** (mean) \n[65.9-69.5], sensitivity (%) of 60 (majority vote),  60.6** (mean) [52.7-68.6], \nspecificity (%) of 74 (majority vote), 73.3** (mean ) [67.8-78.9], PPV (%) of 65 \n(majority vote), 64.6** (mean) [62-67.1], and NPV ( %) of 70 (majority vote), \n70** (mean) [67.1-72.9] (*p<0.05, **p<0.005). \nConclusion: Radiomics-based machine learning may assist clinici ans in \npredicting clinical outcome of patients with lumbar  spine arthrodesis based on \npost-operative X-rays, thus modifying physical reha bilitation and therapeutic \nstrategies accordingly. \nLimitations: Retrospective study. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Approved by Local Ethics \nCommittee (RETRORAD protocol) \nAuthor Disclosures:  \nIrene Carmen Pizza: Nothing to disclose \nStefano Fusco: Nothing to disclose \nCarmelo Messina: Nothing to disclose \nSalvatore Gitto: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nMartina Pedullà: Nothing to disclose \nFrancesca Serpi: Nothing to disclose \nDomenico Albano: Nothing to disclose \n \n \nAI based thoracolumbar and sacral spine fracture de tection for computed \ntomography \n*J-B. Pialat*¹, D. Gicquel¹, A-K. Golla², C. Bürger ², C. Lorenz², M. Villien¹,  \nS. Gouttard¹, A. Vlachomitrou¹, T. Klinder²; ¹Lyon/ FR, ²Hamburg/DE \n(jean-baptiste.pialat@chu-lyon.fr) \n \nPurpose or Learning Objective: AI algorithms which detect vertebral \nfractures generate limited classifications which on ly identify vertebral body \nfracture. We propose a thoracolumbar and sacral spi ne fracture detection \nalgorithm able to identify individual fracture loca tions in both the vertebral body \nand the posterior arch . It segments the entire spi ne, extracts spine-aligned sub \nvolumes and detects spinal fractures using a convol utional neural network. \nMethods or Background: 195 CT scans from polytraumatized patients were \ncollected in a single-center retrospective clinical  study. Dataset was split into \ntraining (n=145) and validation (n=50) sets. Accura cy for identification of injury \nlocation within the body was assessed in the valida tion set using a Free \nResponse ROC (FROC) curve and performance at the ve rtebral body level \nwas measured using a Receiver Operating Characteris tic (ROC) curve. A \nsubsequent test set including 173 patients ( fractu red N=109, non fractured N= \n64) was analyzed with the same algorithm. Performan ce was assessed \nsimilarly using FROC curve. \nResults or Findings: The algorithm detected 87.3% of the 775 spinal frac ture \nlocations of the validation set using a false posit ive threshold of 5 per case. It \ndetected 92.4% of the fractured vertebrae. 249 fals e positives were detected, \nmost of which were easily rejected upon review by r adiologists. 26 false \nnegatives were found, most of which were transverse  process fractures. There \nwere 7 vertebral body fractures; all were single en dplate stable fractures. In the \ntest set, the algorithm detected 88.6% of the 255 f ractures using a false \npositive threshold of 5 per case. \nConclusion: We have developed and validated a deep learning alg orithm \nwhich determines location of fractures in the whole  vertebra with reasonable \naccuracy. \nLimitations: This as to be tested prospectively in routine emerg ency condition \nto assess the gain in time / sensitivity \nFunding for this study: Collaborative study between Hospices Civils de Lyon  \nand Philips using GOPI research fundings \nEthics committee - additional information: Approved by local ethics \ncommitee \nAuthor Disclosures:  \nTobias Klinder: Employee: Philips Healthcare \nMarjorie Villien: Employee: Philips Healthcare \nChristian Lorenz: Employee: Philips Healthcare \nDavid Gicquel: Research/Grant Support: Philips Heal thcare \nSylvain Gouttard: Nothing to disclose \nAnna Vlachomitrou: Employee: Philips Healthcare \nJean-Baptiste Pialat: Research/Grant Support: Phili ps Heathcare \nChristian Bürger: Employee: Philips Healthcare \nAlena-Kathrin Golla: Employee: Philips Healthcare \n \n \nThe new frontier of MRI: virtual dissection with 3D PDw sequence. A pilot \nstudy on ATiFL anatomy \n*G. Del Gaudio*¹, G. Vuurberg², M. Dalmau-Pastor³, G. Kerkhoffs⁴, M. Maas⁴; \n¹Rome/IT, ²Weesp/NL, ³Barcelona/ES, ⁴Amsterdam/NL \n(g.d.gaudio@gmail.com) \n \nPurpose or Learning Objective: In the literature, there is much conflicting \ndata regarding the anatomy of the anterior tibiofib ular ligament (ATiFL), even in \nstudies with anatomical specimens. Therefore, this study aims to reassess the \n\n \n \nAbstract-based Programme \n \n 15  \nWednesday \nanatomy of this ligament using MRI with a high-reso lution isotropic 3D-PDw \nsequence. \nMethods or Background: From February to May 2024, 72 MRI scans (3T) of \nthe ankle were performed at Amsterdam UMC. The incl usion criterion was \npatients over 16 years of age. The exclusion criter ia were: absence of a 3D-\nPDw scan, ATiFL trauma or surgery, congenital anoma lies, metallic or \nmovement artifacts. The 43 3D-PDw valid scans, allo wed for aligning the \nplanes along the individual bundles of the ligament . \nResults or Findings: The high spatial resolution (0,23mm) of 3DPDw allow ed \nthe identification of three bundles: superior, inte rmediate and inferior. \nRegarding dimensions the superior is the thickest a nd widest (mean \n2.68x9.28mm) and the inferior the longest (mean 15. 45mm). Regarding \norientation (axial from the fibula to the tibia) th e superior and inferior have a \ntransverse orientation, while the intermediate is o riented backward. Regarding \nthe shape, they are fanned in 97.7%, 71.7%, and 25. 6% respectively, while \nthey are band-like in the remaining cases. We did n ot identify any anatomical \nvariance regarding the number of bundles. \nConclusion: The use of volumetric isotropic sequences as the 3D -PDw, can \nbe a very useful tool for the anatomical study of l igamentous structures in the \nabsence of available anatomical specimens. Understa nding the exact anatomy \nof this structure is crucial for managing both acut e and chronic traumatic \npathologies (impingement, overuse), especially in y oung patients and athletes. \nLimitations: Sample size and lack of anatomical specimen compari son. \nFunding for this study: No funding. \nEthics committee - additional information: This study received a waiver by \nthe ethical committee according to local rules and regulations. \nAuthor Disclosures:  \nGiovanni Del Gaudio: Nothing to disclose \nGwendolyn Vuurberg: Nothing to disclose  \nMiquel Dalmau-Pastor: Nothing to disclose \nGinom M.J. Kerkhoffs: Nothing to disclose \nMario Maas: Nothing to disclose \n \n \nMRI biomarker assessment of Duchenne muscular dystr ophy disease \nprogression: a 12-month longitudinal study \n*Y. Song*, H. Xu, R. Xu, K. Xu; Chengdu/CN \n(babysong020@163.com) \n \nPurpose or Learning Objective: To evaluate the disease progression in \npatients with Duchenne muscular dystrophy (DMD) by using multi-modal \nquantitative magnetic resonance imaging (qMRI), and  comparing the \nresponsiveness of these imaging indicators with the  clinical function scales. \nMethods or Background: 130 DMD patients were enrolled and underwent \nMRI examination of hip muscles to determine fat fra ction (FF) and longitudinal \nrelaxation time (T1). All participants returned for  follow-up at an average of 12 \nmonths. According to the baseline North Star Ambula tory Assessment (NSAA) \nscore, all patients were divided into three subgrou ps: mild (76-100 score), \nmoderate (51-75 score) and severe (0-50 score) func tional decline. \nStandardized response mean (SRM) was used as the re sponsiveness to the \ndisease progression, and the responsiveness of qMRI  and clinical function \nscales to the disease progression in different DMD stages was compared. \nSRM>0.8 is considered as a high response to disease  progression. \nResults or Findings: The overall SRM of MRI biomarkers is higher than th at \nof the clinical function scales. For mild group, FF  of adductors and abductors \nhave higher responsiveness, with SRM of 0.816 and 1 .043, respectively. For \nmoderate group, FF of all muscle groups have a high  responsiveness, and the \nSRM are between 1.004 and 1.606. For severe group, T1 of abductors and FF \nof all muscle groups have high responsiveness, and SRM are between 0.867 \nand 1.633. However, the SRM of the clinical functio n scales for patients with \ndifferent disease stages are all less than 0.8. \nConclusion: The sensitivity of MRI biomarkers to DMD disease pr ogression is \nhigher than that of clinical function scales, espec ially the FF of gluteal muscles \nis more sensitive to disease progression, and the s ensitivity indicators are \ndifferent in different disease stages. \nLimitations: This study didn‘t discuss whether patients received  steroid \ntherapy. \nFunding for this study: National Natural Science Foundation of China \n(82271981) \nEthics committee - additional information: ChiCTR1800018340 \nAuthor Disclosures:  \nKe Xu: Nothing to disclose \nHuayan Xu: Nothing to disclose \nYu Song: Nothing to disclose \nRong Xu: Nothing to disclose \n \n \n \n \n \n \nHip Imaging: Radiation-free 3D models based on 3D M RI of the hip joint \nfor children with Slipped capital femoral epiphysis  \nT. D. Lerch, *T. Kaim*, K. Ziebarth, M. K. Meier, J . D. Busch; Bern/CH \n(tilman.kaim@bluewin.ch) \n \nPurpose or Learning Objective: Slipped capital femoral epiphyses (SCFE) is \na common pediatric hip disease with the risk of ost eoarthritis and impingement \ndeformities, and 3D models could be useful for pati ent-specific analysis. \nTherefore, magnetic resonance imaging (MRI) bone se gmentation was \ninvestigated. \nMethods or Background: A retrospective IRB-approved study involving 23 \nsymptomatic pediatric patients (23 hips) with SCFE was performed. All patients \nunderwent preoperative hip MR with pelvic axial hig h-resolution images (T1 \nVIBE DIXON images). Slice thickness was 1.2 mm. Mea n age was 12 ± 2 \nyears. All patients underwent surgical treatment. M anual and automatic MRI-\nbased bone segmentation was compared. automatic bon e segmentation was \nperformed by machine learning algorithm, a previous ly used and validated \nconvolutional neural network trained for adult pelv is bone segmentaiton was \nadapted to pelvis of children. \nResults or Findings: Manual MRI-based bone segmentation was feasible (al l \npatients, 100%, duration 4-5 hours per case). Dice coefficient was calculated to \nassess differences between manual and automatic bon e segmentation, Dice \ncoefficient was 82% for the pelvis and 88% for prox imal femur. Precision was \n80% for the pelvis and 94% for proximal femur. \nConclusion: MRI-based 3D models were feasible for SCFE patients . Three-\ndimensional models could be useful for SCFE patient s for preoperative 3D \nprinting and deformity analysis. This could aid for  patient-specific diagnosis, \ntreatment decisions, and preoperative planning. MRI -based 3D models are \nradiation-free and could be used instead of CT-base d 3D models in the future \nfor computer-assisted 3D simulation of surgery. \nLimitations: MRI is expensive and access is limited \nFunding for this study: None \nEthics committee - additional information: IRB approval was obtained \nAuthor Disclosures:  \nKai Ziebarth: Nothing to disclose \nTilman Kaim: Nothing to disclose \nTill Dominic Lerch: Nothing to disclose \nMalin Kristin Meier: Nothing to disclose \nJasmin D. Busch: Nothing to disclose \n \n \nCould a single isotropic 3D sequence replace a mult isequence knee MRI \nin the new era of deep learning reconstruction? \n*E. Nikolova*¹, J. Kroschke¹, C. Obermüller¹, F. Ze cca², K. Pawlus¹, T. Rauer¹, \nF. Ensle¹; ¹Zurich/CH, ²Cagliari/IT \n \nPurpose or Learning Objective: To assess whether a single isotropic 3D \nproton-density-weighted fat-saturated (PDFS) sequen ce could replace a \nstandard 2D multisequence MRI protocol for comprehe nsive examination of \nthe knee using deep learning reconstruction (DLR). \nMethods or Background: In this retrospective study, 95 consecutive patient s \n> 18 years without history of prior knee surgery un dergoing MRI knee \nexamination at the same 1.5 Tesla scanner between M ay 2023 and July 2024 \nwere included. Standard MRI protocol with DLR consi sted of a 3D PDFS \nsequence and five 2D fast-spin-echo sequences in va rious orientations. Two \nradiologists separately evaluated the 3D sequence i n all three planes and the \n2D sequences, assessing pathologies of bone, cartil age, menisci and \nligaments for all joint compartments, and overall i mage quality, diagnostic \nconfidence and artifacts. Wilcoxon signed-rank test  was used to compare \nLikert scale gradings, McNemar’s test for binary gr ades. Interreader agreement \nwas assessed with Cohen’s kappa. \nResults or Findings: There was no significant difference between protoco ls \nregarding assessment of medial(MC) and lateral comp artment(LC) meniscus, \n(MC) and patellofemoral(PF) cartilage, medial and l ateral collateral ligament, \nanterior and posterior cruciate ligament, MC and PF  bone marrow \nedema(BME), and fractures in all compartments (p>0. 05). Significant \ndifferences were shown in assessment of LC cartilag e (p=0.002) and LC BME \n(p=0.04). Image quality and artifacts did not demon strate significant \ndifferences. Diagnostic confidence was significantl y higher for the 2D \nprotocol(p=0.023). Interreader agreement overall wa s substantial for the 3D-\nPDFS(k=0.67) and 2D protocol (k=0.66). \nConclusion: Our results suggests comparable performance between  a single \n3D-PDFS and a multisequence 2D protocol using DLR f or comprehensive \nassessment of knee structures, except for LC cartil age and BME. With DLR-\npowered image enhancement, 3D-PDFS might be able to  partly replace 2D \nsequences for time-efficient knee MRI in the future . \nLimitations: Retrospective study design. No arthroscopic referen ce standard. \nFunding for this study: This research received no financial support. \nEthics committee - additional information: Not applicable \n \n \n \n\n \n \nAbstract-based Programme \n \n 16  \nWednesday \nAuthor Disclosures:  \nElizabet Nikolova: Nothing to disclose \nFabio Zecca: Nothing to disclose  \nFalko Ensle: Nothing to disclose \nKarolina Pawlus: Nothing to disclose \nJonas Kroschke: Nothing to disclose \nCarina Obermüller: Nothing to disclose \nThomas Rauer: Nothing to disclose \n \n \nAssessment of proximal tibial fractures with 3D FRA CTURE (fast field \necho resembling a CT using restricted echo-spacing)  MRI – \nIntraindividual comparison with computed tomography  \n*I. Ristow*¹, S. Zhang², C. Riedel¹, A. Lenz¹, M. K rause¹, G. Adam¹,  \nP. Bannas¹, F. O. Henes¹, L. Well¹; ¹Hamburg/DE, ²B est/NL \n \nPurpose or Learning Objective: To evaluate the feasibility and diagnostic \nperformance of a 3D FRACTURE (fast field echo resem bling a CT using \nrestricted echo-spacing) MRI sequence for the detec tion and classification of \nproximal tibial fractures compared with CT. \nMethods or Background: We retrospectively included 126 patients (85 male; \n39.6±14.5 years) from two centers following acute k nee injury. Patients \nunderwent knee MRI at 3T including FRACTURE-MRI. Ad ditional CT was \nperformed in patients with tibial fractures (32.5%;  n=41) as the reference \nstandard for fracture classification. Two radiologi sts independently evaluated \nFRACTURE-MRI for the presence of fractures and clas sified them according to \nAO/OTA, Schatzker, and the 10-segment classificatio n. Diagnostic \nperformance of FRACTURE-MRI was assessed using cros stabulations. Inter-\nreader agreement was estimated using Krippendorff’s  alpha. Image quality was \ngraded on a five-point scale (5=excellent; 1=inadeq uate definition of fracture \nlines and fracture displacement) and assessed using  estimated marginal \nmeans. \nResults or Findings: Fractures were detected by FRACTURE-MRI with a \nsensitivity of 91.5% (83.2–96.5%) and a specificity  of 97.1% (93.3–99.0%). \nRegarding fracture classification, diagnostic perfo rmances were slightly lower, \nwith the 10-segment classification yielding the bes t sensitivity of 85.7% (81.4–\n89.3%) and specificity of 97.4% (96.6–98.0%), and t he Schatzker classification \nyielding the lowest sensitivity of 78.2% (67.4–86.8 %) and specificity of 97.7% \n(94.1–99.4%). Inter-reader agreement across the who le cohort was excellent \n(Krippendorff’s alpha 0.89–0.96) and when consideri ng only patients with \nfractures, good to acceptable (0.48–0.91). Image qu ality was rated good \n(estimated marginal mean 4.3 (4.1–4.4)). \nConclusion: FRACTURE-MRI is feasible at 3T enabling accurate de lineation \nof fracture lines for precise diagnosis and classif ication of proximal tibial \nfractures. \nLimitations: Future studies need to address in a comparative int ra-individual \nsetting whether the diagnostic performance of FRACT URE-MRI is better or \nequivalent to other CT-like bone imaging techniques , such as UTE/ZTE, GRE, \nor SWI, for fracture detection. \nFunding for this study: N/A \nEthics committee - additional information: The retrospective study was \napproved by the local institutional review board (Ä rztekammer Hamburg). \nAuthor Disclosures:  \nGerhard Adam: Nothing to disclose \nChristoph Riedel: Nothing to disclose \nMatthias Krause: Nothing to disclose \nAlexander Lenz: Nothing to disclose \nPeter Bannas: Nothing to disclose \nFrank Oliver Henes: Nothing to disclose \nShuo Zhang: Employee: Philips \nInka Ristow: Nothing to disclose \nLennart Well: Nothing to disclose \n \n \nqBone: a quantitative software for the semi-automat ed extraction of bone \nmicroarchitecture metrics in vivo using Photon-coun ting-detector CT and \nArtificial Intelligence \n*A. Ferrero*, J. Thorne, A. O. El Sadaney, K. Rajen dran, C. Mccollough,  \nF. Baffour; Rochester, MN/US \n(Ferrero.andrea@mayo.edu) \n \nPurpose or Learning Objective: Pathologies affecting bone health impact \nboth mineral density (vBMD) and morphometric charac teristics (thickness (Th) \nand spacing (Sp)) of trabecular (Tb) and cortical ( Ct) bone. This work \nintroduces a semi-automated software, qBone, which quantifies bone \nmorphometry from in vivo CT scans of the extremitie s and the vertebral spine. \nMethods or Background: protocols were optimized for extremity and spine \nexams using a commercial photon-counting-detector ( PCD) CT system. A \ndedicated CNN algorithm was trained to reduce image  noise of the spine CT \nexams while maintaining high resolution details. Ad aptive segmentation \nalgorithms automatically delineated Ct and Tb compa rtments allowing \nquantification of Th, Sp and vBMD for each. To vali date the software’s \naccuracy, 10 cadaveric wrists were scanned on HRpQC T and PCD-CT. A 3D-\nprinted bone model (Ct.Th=2mm, Tb.Th=0.3mm and Tb.S p=0.75mm) was \nused to assess the CNN denoising performance across  different patient sizes. \nFinally, qBone was applied in vivo to multiple pros pective cohorts for wrist and \nspine. \nResults or Findings: optimized PCD-CT protocols for wrist (70kV, 12mGy, \n<0.1mSv) and spine (120kV, 40mGy, 8mSv) yielded <0. 15mm in-plane \nresolution. Validation with cadaveric wrists and th e 3D-printed bone model \ndemonstrated excellent agreement in Ct.Th and Tb.Sp  metrics. CNN denoising \nsignificantly improved trabecular morphometry accur acy in the spine for small \nand medium patient sizes. In vivo measurements (wri st, N=50; spine, N=14) for \neach metric (Tb.Th=0.3-0.45mm, Tb.Sp=0.6-1.05mm, Ct .Th=0.5-1.58mm, \nCt.vBMD=450-600mg/cm3) were consistent with literat ure values. \nConclusion: qBone facilitates semi-automated quantification of bone \nmorphometry from high resolution CT data, providing  a comprehensive \nassessment of bone health in vivo beyond traditiona l mineral density. \nLimitations: Comparisons with microCT are needed to validate met rics for \nvertebral bones. Additionally, qBone does not lever age spectral information to \nestimate vBMD in the spine. \nFunding for this study: NIH R21ar084126-01a1 \nEthics committee - additional information: IRB 23_005308, PI: Baffour \nAuthor Disclosures:  \nAhmed O. El Sadaney: Nothing to disclose \nFrancis Baffour: Nothing to disclose \nAndrea Ferrero: Nothing to disclose  \nJamison Thorne: Nothing to disclose \nKishore Rajendran: Nothing to disclose \nCynthia Mccollough: Nothing to disclose \n \n \n10:00-11:00 Research Stage 1 \nResearch Presentation Session: Head and \nNeck \nRPS 208 \nKey insights in soft tissue neck imaging \n \nModerator \nE. Loney; Halifax/UK  \n(elizabeth_loney@hotmail.co.uk) \nAuthor Disclosures:  \nElizabeth Loney: Consultant: DMC Healthcare Ltd \n \n \nDetection of MRI edema patterns in patients with ac ute neck infections: \na prospective blinded multidisciplinary and multice nter interobserver \nhuman performance evaluation \n*J-P. T. Vierula*, J. Velhonoja, A. Sirén, J. Nurmi nen, M. J. Nyman, K. Mattila, \nJ. Hirvonen; Turku/FI \n \nPurpose or Learning Objective: In patients with acute neck infections, MRI \nshows reactive edema patterns that predict disease severity: retropharyngeal \nedema (RPE) and mediastinal edema (ME). How well ra diologists and \nclinicians with diverse backgrounds and neck MRI ex perience can detect these \nedema patterns is unknown. \nMethods or Background: This prospective, blinded, multidisciplinary, \nmulticenter interobserver study evaluated human per formance in detecting \nRPE and ME from axial in-phase and water T2-weighte d Dixon images. \nReaders (N=28, including radiologists, neuroradiolo gists, radiology residents, \nhead and neck surgeons) from all five university ho spitals in Finland were \nbriefly trained and assessed the presence of RPE an d ME (yes/no) and rated \ntheir confidence (1-5) blinded to clinical data. Em ergency MRI images were \nobtained from 60 patients with acute neck infection s. Ten patients appeared \ntwice to assess intraobserver variability. Sensitiv ity, specificity, accuracy, and \ninterobserver agreement were assessed. \nResults or Findings: Overall sensitivity, specificity, and accuracy were  0.89, \n0.81, and 0.85 for RPE and 0.85, 0.81, and 0.82 for  ME. ME accuracy \ncorrelated with confidence (p=0.002), whereas RPE a ccuracy did not \n(p=0.580). Radiologists achieved higher RPE sensiti vity (p=0.01), RPE \naccuracy (p=0.04), ME sensitivity (p=0.01), and ME accuracy (p=0.02) than \nclinicians, whereas other group comparisons were no t significant. High \nconfidence was found for RPE (4.3) and ME (4.1). Ov erall, interobserver kappa \nwas 0.61 (substantial) for RPE and 0.52 (moderate) for ME, with radiologists \nshowing higher agreement than clinicians. Median in traobserver accuracy was \n90% for both RPE and ME. \n\n \n \nAbstract-based Programme \n \n 17  \nWednesday \nConclusion: We show high diagnostic accuracy and substantial in terobserver \nagreement for detecting clinically relevant reactiv e edema patterns on MRI in \npatients with acute neck infections. These results encourage using these \nbiomarkers in clinical practice. \nLimitations: Limited availability of emergency MRI. \nFunding for this study: This study was financially supported by the Sigrid \nJusélius Foundation. The funders had no role in stu dy design, data collection \nand analysis, decision to publish, or preparation o f the manuscript. \nEthics committee - additional information: A waiver for patient consent was \nnot sought because it is not required by the nation al legislature for \nretrospective studies of existing data. \nAuthor Disclosures:  \nAapo Sirén: Nothing to disclose \nJarno Velhonoja: Nothing to disclose \nJussi Hirvonen: Nothing to disclose \nJari-Pekka Tapani Vierula: Nothing to disclose \nMikko Juhani Nyman: Nothing to disclose \nKimmo Mattila: Nothing to disclose \nJanne Nurminen: Nothing to disclose \n \n \nEnhanced Survival Prediction in Nasopharyngeal Carc inoma Through \nIntegrated Peritumoral and Intratumoral Radiomics \n*S. Khongwirotphan*, A. Prayongrat, S. Kitpanit, D.  Kannarunimit,  \nC. Chakkabat, V. Shotelersuk, S. Sriswasdi, C. Lert butsayanukul,  \nY. Rakvongthai; Bangkok/TH \n(sararas.k@chula.ac.th) \n \nPurpose or Learning Objective: Accurately prediction of overall survival (OS) \nis essential for optimizing treatment strategies in  nasopharyngeal carcinoma \n(NPC), thereby improving patient outcomes. This stu dy aimed to enhance OS \nprediction by integrating radiomic features from bo th intra- and peritumoral \nareas, offering a novel biomarker approach beyond t raditional clinical features. \nMethods or Background: We analyzed 251 NPC patients treated with \nchemoradiotherapy between 2010 and 2019, all follow ed for a minimum of \nthree years. Radiomics features were extracted from  the gross tumor volume \n(GTV) contours and a 3-mm peritumoral area of the p re-treatment CT images \nusing PyRadiomics v3.0.1. The robustness and predic tive power of radiomic \nfeatures were assessed by intraclass correlation an d univariate Cox \nregression. Selected radiomics features were combin ed with clinical data (age, \ngender, T-stage, N-stage) for a multivariate analys is. Cox regression models \nwere optimized with recursive feature elimination a nd 20 repetitions of five-fold \ncross-validation, reserving 20% of dataset for mode l testing. \nResults or Findings: Addition of peritumoral radiomic features significa ntly (P \n< 0.05) improved survival predictions (C-index: 0.7 87±0.067 validation; 0.669 \ntest), over intratumoral only (C-index: 0.755±0.063 validation; 0.626 test set). \nIntegrating clinical data with intra- and peritumor al radiomics yielded the best \nmodel, with a C-index of 0.832±0.052 in validation and 0.727 in test set, that \noutperformed (P < 0.05) the model with clinical and  intratumoral features (C-\nindex of 0.769±0.066 and 0.705, respectively). The baseline clinical model \nyielded a C-index of 0.703±0.100 in validation and 0.618 in test set. \nConclusion: Integrating radiomic features from both intra- and peritumoral \nareas significantly improved OS predictions in NPC,  surpassing traditional \napproaches that utilize only clinical and intratumo ral radiomics. This could lead \nto more personalized treatment strategies, potentia lly improving patient \noutcomes. \nLimitations: External validation is recommended for this single- center \nretrospective study before clinical use. \nFunding for this study: This research project is supported by National \nResearch Council of Thailand (NRCT) and grants for development of new \nfaculty staff, Ratchadaphiseksomphot Fund, Chulalon gkorn University \nEthics committee - additional information: The study received ethics \napproval from the Institutional Review Board of the  Faculty of Medicine, \nChulalongkorn University, Thailand (IRB no. 0630/66 ). \nAuthor Disclosures:  \nYothin Rakvongthai: Nothing to disclose \nSira Sriswasdi: Nothing to disclose \nAnussara Prayongrat: Nothing to disclose \nDanita Kannarunimit: Nothing to disclose \nChakkapong Chakkabat: Nothing to disclose \nVorasuk Shotelersuk: Nothing to disclose \nChawalit Lertbutsayanukul: Nothing to disclose \nSararas Khongwirotphan: Nothing to disclose \nSarin Kitpanit: Nothing to disclose \n \n \n \n \n \n \n \nPredictive potential of dynamic contrast-enhanced M RI and plasma-\nderived angiogenic factors for response to concurre nt \nchemoradiotherapy in human papillomavirus-negative oropharyngeal \ncancer \n*A. Longo*, P. Hudler, P. Strojan, G. Plavc, L. Ume k, K. Surlan Popovic; \nLjubljana/SI \n(aljalongo@gmail.com) \n \nPurpose or Learning Objective: Dynamic contrast-enhanced magnetic \nresonance imaging (DCE-MRI) can assess tumour vascu larity, which depends \non the process of angiogenesis and affects tumour r esponse to treatment. Our \nstudy explored the associations between DCE-MRI par ameters and the \nexpression of plasma angiogenic factors in human pa pilloma virus (HPV)-\nnegative oropharyngeal cancer, as well as their pre dictive value for response \nto concurrent chemoradiotherapy (cCRT). \nMethods or Background: 25 patients with locally advanced HPV-negative \noropharyngeal carcinoma were prospectively enrolled  in the study. DCE-MRI \nand blood plasma sampling were conducted before cCR T, after receiving a \nradiation dose of 20 Gy, and after the completion o f cCRT. Perfusion \nparameters ktrans, kep, Ve, initial area under the curve (iAUC) and plasma \nexpression levels of angiogenic factors (vascular e ndothelial growth factor \n[VEGF], connective tissue growth factor [CTGF], pla telet-derived growth factor \n[PDGF]-AB, angiogenin [ANG], endostatin [END] and t hrombospondin-1 \n[THBS1]) were measured at each time-point. Patients  were stratified into \nresponders and non-responders based on clinical eva luation. Differences and \ncorrelations between measures were used to generate  prognostic models for \nresponse prediction. \nResults or Findings: Higher perfusion parameter ktrans and higher plasma  \nVEGF levels successfully discriminated responders f rom non-responders \nacross all measured time-points, whereas higher iAU C and higher plasma \nPDGF-AB levels were also discriminative at selected  time points. Using early \nintra-treatment measurements of ktrans and VEGF, a predictive model was \ncreated with cut-off values of 0.259 min-1 for ktra ns and 62.5 pg/mL for plasma \nVEGF. \nConclusion: Early intra-treatment DCE-MRI parameter ktrans and plasma \nVEGF levels may be valuable early predictors of res ponse to cCRT in HPV-\nnegative oropharyngeal cancer. \nLimitations: The small sample size and absence of healthy contro ls for \nangiogenic factors limit our findings. Perfusion pa rameter values in DCE-MRI \ncan vary with different post-processing software, s o comparisons between \nstudies should be made with caution. \nFunding for this study: This research was funded by the Slovenian Research \nand Innovation Agency (ARIS), grant number P3-0307.  \nEthics committee - additional information: The study was approved by the \nNational Medical Ethics Committee of the Republic o f Slovenia (No. 0120-\n247/2019/4, 12 June 2019) and the Committee for Med ical Ethics of the \nInstitute of Oncology Ljubljana (OI: 28.5.2019, ERI DEK-0064/2019). Written \ninformed consent was obtained from all patients. \nAuthor Disclosures:  \nLan Umek: Nothing to disclose \nGaber Plavc: Nothing to disclose \nKatarina Surlan Popovic: Nothing to disclose \nPetra Hudler: Nothing to disclose \nPrimož Strojan: Nothing to disclose \nAlja Longo: Nothing to disclose \n \n \nThe impact of quantification of circulating tumor H PV-DNA in the clinical \nand surgical management of patients with oropharyng eal squamous cell \ncarcinoma \n*S. Ruggiero*¹, S. Lucchese², V. Dolcetti¹, S. Marz i¹, A. Vidiri¹; ¹Rome/IT, \n²Naples/IT \n(s.ruggiero94@gmail.com) \n \nPurpose or Learning Objective: To conduct a multifactorial assessment in \npatients with oropharyngeal squamous cell carcinoma  (OPSCC) that includes: \n1) clinical characteristics; 2) detection of the am ount of circulating tumor HPV-\nDNA in plasma (ctHPVDNA); 3) MRI-based volumetric a nalysis of both the \nprimary tumor and lateral cervical lymph node metas tases. \nMethods or Background: A prospective study was conducted on patients with \nOPSCC. As controls, patients with suspected HPV-neg ative OPSCC were \nused. Both the primary tumor and lateral cervical l ymph node metastases were \nmanually delineated, slice by slice, by two expert head and neck radiologists \non T2-weighted axial images at diagnosis. In patien ts with multiple lymph node \nmetastases, the total volume was obtained by summin g the volume of each \nlymph node. This measurement was correlated with th e number of ctHPVDNA \ncopies using the Mann-Whitney test. \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 18  \nWednesday \nResults or Findings: A total of 95 patients were included, of which 58 ( 61%) \nwere p16+/HPV16+, 3 (3%) were p16-/HPV33+, 23 (24%)  were p16-/HPV-, 1 \n(1%) was p16-/HPV16+, and 10 (10%) were p16+/HPV-. No association was \nfound between the number of ctHPVDNA copies and the  primary tumor \nvolume. However, significant correlations emerged w ith the volume of lymph \nnode metastases (Rho = 0.42, p = 0.004) and with th e combined volume of the \nlymph node metastases and primary tumor (Rho = 0.51 , p < 0.001). \nConclusion: ctHPVDNA is a promising biomarker that could potent ially \neliminate the need for solid biopsy for diagnosis. The data demonstrate an \nexcellent correlation between p16+/HPV DNA testing,  ctHPVDNA, and the \nvolume of the primary tumor and lymph node metastas es. This suggests that \nliquid biopsy could be useful in identifying the su bgroup of patients with better \noncological outcomes. \nLimitations: Principal limitation of the study is the number of enrolled patients, \nbut it was statistically appropriate. \nFunding for this study: Partially supported by LILT 2020-21 Program and \nMAECI-Call for Joint Project Proposals Italy-Brazil  # BR22GR03. \nEthics committee - additional information: All enrolled patients signed an \ninformed consent to the protocol approved by the IR CCS Regina Elena \nNational Cancer Institute, Istituti Fisioterapici O spitalieri, Institu- tional Review \nBoard (RS1647/22). \nAuthor Disclosures:  \nSergio Ruggiero: Nothing to disclose  \nSimona Marzi: Nothing to disclose  \nSonia Lucchese: Nothing to disclose \nVincenzo Dolcetti: Nothing to disclose  \nAntonello Vidiri: Nothing to disclose \n \n \nThe utility of intraoral ultrasonography in differe ntial diagnosis of benign \nand malignant oral mucosal lesions \n*R. Abdalla-Aslan*¹, D. E. Gaitini¹, A. Rchmiel¹, G . Merhav¹, S. Akrish¹,  \nM. Javitt², D. Shilo¹, O. Emodi¹, N. Beck-Razi¹; ¹H aifa/IL, ²Miami/US \n(ragdaa@gmail.com) \n \nPurpose or Learning Objective: Simple yet reliable methods for \ndifferentiating between benign and malignant soft t issue tumours of the oral \ncavity are currently lacking. Our primary aim was t o assess the correlation \nbetween pre-operative intraoral ultrasound (US) var iables and malignancy of \noral lesions. \nMethods or Background: This is a cross-sectional prospective study of \nconsecutive patients over the age of 18 years from both genders, with a clinical \ndiagnosis of a soft tissue lesion in the oral mucos a. Within a 2-weeks interval, \npatients who are scheduled for biopsy and histopath ological examination \nunderwent high-resolution intraoral US obtained usi ng a 7-15 MHz-L15-7io-\nlinear-ultrasound-transducer-‘hockey stick-probe’ o n a Philips-Epiq-5, 7 \nmachines (Philips Medical, Netherlands). Sonographi c variables included: size \nin 3-dimensions, echogenicity, presence of cystic a reas, presence of \ncalcifications, margins and vascularity. The sonogr aphic findings were \ncompared with histopathology. \nResults or Findings: Full data was available for 52 patients with 53 tum ors. \nIncluded were 24 females and 28 males, with a mean age 60.11±16.7 years \n[range 18-90]. Following histopathological results,  22 patients with 23 tumors \nwere diagnosed with squamous cell carcinoma (SCC), 3 patients with \ndysplasia (1 mild, 1 moderate and 1 severe), and th e remaining 26 patients \nwith benign lesions.  \nSonographic variables of maximal diameter (a cutoff  of 12 mm with 0.76 \nsensitivity and 0.77 specificity), margins (ill-defined, p<0.001) and vascularity \n(high or type III/IV, p=0.002) proved to be significantly correlated to SCC group, \ncompared to benign lesions group. \nConclusion: The utility of intraoral US in the differential dia gnosis of benign \nand SCC tumors in the oral cavity is demonstrated b y this prospective clinical \nstudy, using sonographic features of maximal diamet er, margins and \nvascularity. \nLimitations: Small sample size and operator dependent technique.  \nFunding for this study: None \nEthics committee - additional information: Ehics committee of Rambam \nHealth Care Campus reference RMB-19-0596. \nAuthor Disclosures:  \nAdi Rchmiel: Nothing to disclose \nSharon Akrish: Nothing to disclose \nMarcia Javitt: Nothing to disclose \nOmri Emodi: Nothing to disclose \nGoni Merhav: Nothing to disclose \nDiana E. Gaitini: Nothing to disclose \nDekel Shilo: Nothing to disclose \nNira Beck-Razi: Nothing to disclose \nRagda Abdalla-Aslan: Nothing to disclose \n \n \n \n \nNeoadjuvant radiochemotherapy in patients with loca lly advanced oral-\ncavity tumour: Response-predictive radiological ima ging features \n*I. Burck*, A. Gleich, R. Winkelmann, M. Fleischman n, E. Herrmann,  \nJ-E. Scholtz, P. Thönissen, T. Vogl, D. Pinto Dos S antos; Frankfurt/DE \n \nPurpose or Learning Objective: To explore radiological MR imaging features \nto predict response to neoadjuvant radiochemotherap y in patients with locally \nadvanced oral-cavity tumour. \nMethods or Background: We included 30 patients (15 women, mean age \n60±10 years) with oral cavity cancer (stage IVa) who underwent neoadjuvant \nradiochemotherapy (RTX) before surgery. MRI scans w ere performed before \nRTX, 15 days after its initiation and preoperativel y. Two radiologists \nretrospectively evaluated the images for overall tu mour signal intensity (SI), SI \nchange over time, and tumour extent using a Likert scale. Quantitative analysis \nwas performed for the absolute SI of the tumor in A DC-, DWI-, and T2-\nweighted sequences normalized to the spinal cord. T umour volume (TV) was \ncalculated manually in a contrast-enhanced T1 seque nce. Differences and \nratios of ADC, DWI and T2-SI and TV were calculated . Patients with a stage \npT1 or T0 were classified as responders, all others  as non-responders. \nResults or Findings: ADC-SIs at 2nd and 3rd MRI differed significantly \nbetween responders and non-responders (p = 0.010 an d p = 0.013), as did the \nratio between baseline and preoperative DWI-SIs (p = 0.041) and the \ndifference between normalized baseline and preopera tive ADC-SIs (p = 0.049). \nNon-responders showed an increase in TV at the 2nd MRI, while responders \nshowed a significant decrease in TV, so the calcula ted percentage decrease \nand ratio are significant markers of response progr ession. \nConclusion: Diffusion weighted imaging parameters as well as tu mour \nvolumetry may predict response to neoadjuvant radio chemotherapy in oral \ncavity cancer and may be benefical for image guided  treatment potentially be \nused to guide treatment or extent of surgery in the se patients. \nLimitations: Limitations of this study include small sample size , single-center \nand retrospective study design. \nFunding for this study: None. \nEthics committee - additional information: Ethics approval was obtained by \nthe institutional review board (approvals number 20 8/12). \nAuthor Disclosures:  \nRia Winkelmann: Nothing to disclose \nThomas Vogl: Nothing to disclose \nJan-Erik Scholtz: Nothing to disclose \nDaniel Pinto Dos Santos: Nothing to disclose \nPhilipp Thönissen: Nothing to disclose \nAlexander Gleich: Nothing to disclose \nEva Herrmann: Nothing to disclose \nMaximilian Fleischmann: Nothing to disclose \nIris Burck: Nothing to disclose \n \n \nImpact of Deep Learning-Based Image Reconstructions  in Head and Neck \nMRI \n*F. Albisinni*¹, C. Carbone¹, C. Zacchi¹, M. Ravane lli¹, D. Farina¹,  \nB. Van Deberge²; ¹Brescia/IT, ²Leuven/BE \n(f.albisinni@unibs.it) \n \nPurpose or Learning Objective: This study aimed to assess the impact of \ndeep learning (DL)-based reconstructions on T2 imag e quality in head and \nneck MRI. Additionally, potential time savings from  using DL in various \nsequences were evaluated. \nMethods or Background: Three sequences were compared: (A) TSE T2 \nwithout DL, with three signal averages and an acqui sition time of 2'35''; (B) \nTSE T2 with DL at intermediate strength, with two s ignal averages and an \nacquisition time of 1'25''; and (C) TSE T2 with DL at maximal strength, with one \nsignal average and acquisition time of 43''. Images  from 52 patients were \nrandomly and blindly evaluated by three radiologist s with varying levels of \nexperience using MR scanners from different vendors . For each patient, two \nimages were analyzed at the level of the nasopharyn x and oral cavity. Four \ncategories were assessed: overall image quality, ar tifacts, edge sharpness, \nand noise, each rated on a 3-point Likert scale. Ad ditionally, radiologists were \ntasked with identifying the correct sequence for ea ch image. The two main \nendpoints evaluated were inter-rater reproducibilit y and comparison of the \nsequences \nResults or Findings: A total of 636 images were rated by three radiologi sts. \nInter-rated reproducibility was poor across all cat egories. The sequences type \nwas correctly identified in only 44% cases. The ove rall quality scores for \nsequence A, B and C were 2.63, 2.52, and 2.52 respe ctively (p=0.055). \nArtifacts scores were 2.61, 2.48, and 2.55 (p=0.07) ; edge sharpness scores \nwere 2.57, 2.39, and 2.43 (p=0.02, with significant  differences between \nsequences A and B); and noise scores were 2.51, 2.4 1, and 2.41 (p=0.1) \nConclusion: The performances of the three sequences were simila r overall. \nDL-based sequences for head and neck MRI were shown  effective and offered \nsignificant time savings, enabling potential ultra- fast imaging protocols \nLimitations: Small sample \nFunding for this study: No funding was provided for this study \n\n \n \nAbstract-based Programme \n \n 19  \nWednesday \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nDavide Farina: Nothing to disclose  \nBaptiste Van Deberge: Nothing to disclose \nFlavia Albisinni: Nothing to disclose  \nMarco Ravanelli: Nothing to disclose  \nChiara Zacchi: Nothing to disclose  \nCarmela Carbone: Nothing to disclose \n \n \n10:00-11:00 Research Stage 2 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 205 \nMeta-level topics in AI: cost-effectiveness, \nnon-interpretive use-cases and evidence \n \nModerator \nE. Neri; Pisa/IT  \n(emanuele.neri@med.unipi.it) \n \n \nEarly health technology assessment for an artificia l intelligence tool to \ndetect incidental pulmonary embolisms on computed t omography \n*E. H. M. Kemper*¹, K. Redekop¹, F. Vos², M. Ijzerm an¹, M. P. A. Starmans¹,  \nJ. J. Visser¹; ¹Rotterdam/NL, ²Delft/NL \n(e.h.m.kemper@erasmusmc.nl) \n \nPurpose or Learning Objective: Incidental pulmonary embolisms (IPE) on \ncomputed tomography (CT) are missed in up to 70% of  cases. While artificial \nintelligence (AI) tools for IPE detection exist, an  evaluation on if and how these \ntools can provide actual value, e.g., fit patients and end-users needs (i.e., \nradiologists), have never been performed. The aim o f this early health \ntechnology assessment (eHTA) is to determine the re quirements for a value-\nbased AI tool for IPE detection on CT. \nMethods or Background: A comprehensive eHTA process for radiology-AI \nwas proposed and conducted for IPE. A literature se arch, structured \ninterviews, focus group, and evaluation meetings we re performed with the \nidentified stakeholders to define criteria and scen arios for a multiple criteria \ndecision analysis (MCDA). A representative survey w as developed and \ncirculated to weigh the importance of the criteria and assess performances of \nfour possible AI designs. MCDA analysis on the surv ey help quantify the value \nrequirements. \nResults or Findings: Consultations with radiologists, treating physician s, \npatients, radiology technologists, AI specialists, legal experts, and ethicists \nresulted in 14 sub-criteria and five main criteria;  patient impact, model \nperformance, physician support, environmental impac t, and costs. Preliminary \noutcomes indicate that a short follow-up time for d iagnosing IPE is more \nimportant than a high sensitivity for IPE detection . \nConclusion: A value-based AI tool for IPE detection should be f ocused on \ntriage to reduce the impact of the diagnosis of IPE  on the patient, mainly \nbecause delay of diagnosis can result in progressio n of the IPE and \npreventable stress for the patient, while an improv ed detection rate is \nconsidered to result in significant overtreatment. \nLimitations: The scope of this analysis has been within Europe. Outcomes \nmight not be applicable elsewhere. \nFunding for this study: E.H.M.K., K.R., M.P.A.S., F.V, and J.J.V. \nacknowledge funding by LSH-TKI (Health~Holland Dutc h Top Sector Life \nSciences and Health) 23024 \nEthics committee - additional information: No applicable \nAuthor Disclosures:  \nMaarten Ijzerman: Nothing to disclose \nKen Redekop: Nothing to disclose \nErik Hermanus Marcellinus Kemper: Nothing to disclo se \nMartijn Pieter Anton Starmans: Nothing to disclose \nJacob Johannes Visser: Advisory Board: Contextflow \nFrans Vos: Nothing to disclose \n \n \n \n \n \n \nPotential costs and benefits of AI for fracture det ection in cervical spine \nCT scans at hospital level \n*G. Van Den Wittenboer*¹, I. M. Nijholt¹, M. Maas²,  M. F. Boomsma¹; \n¹Zwolle/NL, ²Amsterdam/NL \n \nPurpose or Learning Objective: Aim of this study was to assess healthcare \ncosts at the hospital level for patients screened f or cervical spine (CS) fractures \nusing CT, and to estimate costs and benefits of inc orporating artificial \nintelligence (AI) to detect CS fractures in clinica l practice. \nMethods or Background: Diagnostic accuracy of on-duty radiologists and AI \nin detecting CS fractures on CT scans from a retros pective database (n=2321, \n≥18 years, 2007-2014) was compared with a reference standard. Healthcare \ncosts for patients were inventoried up to 7 months after their emergency \ndepartment visit. Total and average costs per patie nt based on the radiologist \ndiagnosis were calculated for four categories: true  positive, true negative, false \npositive, and false negative. Finally, a scenario-a nalysis was conducted to \nestimate the diagnostic accuracy of radiologists co mbined with AI, and the \ncorresponding total healthcare costs per diagnostic  category. \nResults or Findings: Radiologists identified 193 out of 219 scans with \nfractures and 2085 out of 2102 scans without fractu res, whereas AI identified \n177 out of 219 fractures and 2065 out of 2102 scans  without fractures. AI \nidentified 23 fractures missed by the radiologists and correctly classified 16 \nnon-fracture scans that had been misclassified as f ractures by the radiologists. \nThis resulted in a potential sensitivity of 216/219  (98%) and specificity of \n2101/2102 (>99%) for the combined radiologist-AI ap proach. On average, \n€5,978 less was spent per missed fracture. The tota l cost for the AI-assisted \nscenario was €61,132 (0.3%) higher than for radiolo gists alone. \nConclusion: In this scenario-analysis, the use of AI appears to  increase \nhospital costs by 0.3% due to more accurate diagnos es. A next step could be \nto complement these results with non- hospital cost s and quality-adjusted life \nyears to further investigate the cost-effectiveness  of this AI. \nLimitations: No limitations were identified. \nFunding for this study: The radiology department of the Isala received a \ngrant from AIDOC Medical to have a third party (THI NC, Utrecht, the \nNetherlands) that is specialized in early health te chnology assessments, \nperform the analyses for this study. AIDOC medical had no role in the data \nanalyses itself. Neither AIDOC Medical nor THINC ha d a role in data collection \nor drafting of the abstract. \nEthics committee - additional information: The study uses retrospective \ndata. \nAuthor Disclosures:  \nMartijn Franklin Boomsma: Other: The Department of Radiology, Isala, has \nestablished a strategic partnership with Aidoc Medi cal. However, Aidoc \nMedical had no decisive role in data collection, da ta analysis nor data \ninterpretation. \nIngrid M. Nijholt: Other: The Department of Radiolo gy, Isala, has established a \nstrategic partnership with Aidoc Medical. However, Aidoc Medical had no \ndecisive role in data collection, data analysis nor  data interpretation. \nMario Maas: Nothing to disclose \nGaby Van Den Wittenboer: Other: The Department of R adiology, Isala, has \nestablished a strategic partnership with Aidoc Medi cal. However, Aidoc \nMedical had no decisive role in data collection, da ta analysis nor data \ninterpretation. \n \n \nCost-effectiveness of AI-assisted digital mammograp hy – results from a \nSwedish model-based analysis \n*P. Gialias*, J. Lyth, M. Kristoffersen Wiberg, T. Bjerner, M. Husberg,  \nL. Bernfort, H. Gustafsson, L-Å. Levin; Linköping/S E \n(pantelis.gialias@liu.se) \n \nPurpose or Learning Objective: To evaluate the cost-effectiveness of AI-\nassisted biennial digital mammography (AI-DM) in co mparison to conventional \ndigital mammography (cDM) with double reading of sc reening mammograms \n(screening interval ages 40-74). \nMethods or Background: We used a Markov decision analytic model with a \nlife-time horizon. The analysis was conducted from a healthcare perspective. \nModel parameters were based on Swedish registry dat a and published \nrandomized AI-DM studies. The model estimates the c osts and quality-\nadjusted life-years (QALYs) related to mammography and breast cancer. \nMammography-related costs were collected from the u niversity hospital in \nLinköping. Stage-specific cancer cost,QALY-weights were obtained from the \nliterature. Scenario analyses were performed with d ifferent screening \nstrategies. \nResults or Findings: Per 1000 individuals AI-DM gained 10.8 QALYs \ncompared to cDM. The costs per 1000 individuals wer e USD 3,752,278 and \nUSD 3,816,443 for AI-DM and cDM, respectively. AI-D M resulted in a cost \nsaving of USD 64 165 which makes it a dominant stra tegy. The isolated \nscreening costs were slightly higher in the used AI -DM setting, USD 597, but \nthis was offset by reduced lifetime costs of cancer  treatment. A screening \nstrategy with AI plus one radiologist for all exami nations saves USD 9128 \n\n \n \nAbstract-based Programme \n \n 20  \nWednesday \nscreening costs compared to cDM, however the QALYs gained were \ndecreased to 8.8. \nConclusion: AI-DM is cost saving in our setting and generates m ore quality-\nadjusted life-years. One of the add-on benefits is the possibility to free \nradiological time to other clinical work. These ben efits could be further \nimproved by changing the AI-DM triaging strategy. \nLimitations: We based AI parameters in the model mainly on two S wedish \nrandomized trials and cancer data from the populati on-based cancer registry \nfrom Sweden. However, cost data are highly dependen t on the Swedish health \ncare system and the generalizability to other healt h care systems might be \nlimited. \nFunding for this study: None \nEthics committee - additional information: Not ethics committee approval \nwas need for this study \nAuthor Disclosures:  \nMaria Kristoffersen Wiberg: Nothing to disclose \nMagnus Husberg: Nothing to disclose \nTomas Bjerner: Nothing to disclose \nHåkan Gustafsson: Nothing to disclose \nJohan Lyth: Nothing to disclose \nLars Bernfort: Nothing to disclose \nLars-Åke Levin: Nothing to disclose \nPantelis Gialias: Nothing to disclose \n \n \nAI Tools to Reduce Claims and Compensation Payments  of Missed \nFractures on Radiographs: A Potential Game Changer?  \nM. Tordjman¹, L. Gracia¹, E. Guillo¹, R. Amar¹, J. Ventre¹, *N-E. Regnard*²,  \nR. Y. Carlier¹, J-L. Marmorat¹, J-D. Laredo¹; ¹Pari s/FR, ²Lieusaint/FR \n(noreddine.regnard@gleamer.ai) \n \nPurpose or Learning Objective: To evaluate the potential of BoneView, an AI \ntool for fracture detection on radiographs, in clai ms files of missed fractures \nwhich led to financial compensation. \nMethods or Background: This retrospective study included all the files of \npatients who submitted a claim and had financial co mpensation for missed \nfractures on radiographs from January 2013 to Decem ber 2019 in the 38 \nuniversity hospitals of the Greater Paris area Hosp itals (APHP, France). Of the \n29 patients who claimed files, 26 were finally incl uded (3 were not available in \nthe system). For each patient with a claim, 5 patie nts with radiographs from the \nsame anatomical areas (with or without fracture) we re included from \nconsecutive patients who had radiographs at a unive rsity hospital in 2022. Two \nreaders (one fellow in musculoskeletal radiology an d one expert radiologist in \nmusculoskeletal imaging with more than 20 years of experience) read the \nradiographs, blinded from which patients had missed  fractures. \nResults or Findings: 156 patients were included (26 patients with missed  \nfractures and 130 « control » patients). The AI sof tware was able to detect \n80.7% of fractures (21/26) for the patients who fil ed claims for missed \nfractures. The sensitivity of readers was also impr oved with AI for these \npatients: the junior reader had a sensitivity of 61 .5% without AI and 69.2% with \nAI and the expert reader had a sensitivity of 73.1%  without AI and 84.6% with \nAI. The total of potentially avoided financial comp ensation would have been \n265.314 euros. \nConclusion: The sensitivity of the two readers is improved with  AI in a cohort \nof patients with missed fractures who submitted cla ims and had financial \ncompensations. AI was able to detect most of these fractures. \nLimitations: A limitation was the small number of claims files. \nFunding for this study: There was no funding for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nRaphael Amar: Nothing to disclose \nNor-Eddine Regnard: Founder: Chief Medical Officer of Gleamer \nJean-Luc Marmorat: Nothing to disclose \nMickael Tordjman: Nothing to disclose \nJean-Denis Laredo: Employee: Gleamer \nRobert Yves Carlier: Nothing to disclose \nJeanne Ventre: Employee: Gleamer \nEnora Guillo: Nothing to disclose \nLaure Gracia: Nothing to disclose \n \n \nOverlooked and underpowered: a meta-research study addressing \nsample size in radiomics research \n*J. Zhong*¹, J. Lu², Y. Xing¹, Y. Hu¹, D. Ding¹, X.  Liu¹, S. Dai¹, H. Zhang¹,  \nW. Yao¹; ¹Shanghai/CN, ²Stanford, CA/US \n(wal_zjy@163.com) \n \nPurpose or Learning Objective: To investigate how studies determine the \nsample size when developing radiomics models, and w hether it is sufficient. \nMethods or Background: We identified radiomics studies published from \nJanuary to December 2023 on seven leading peer-revi ewed radiological \njournals owned by European Society of Radiology and  Radiological Society of \nNorth America. We reviewed the sample size justific ation methods, and actual \nsample size used. We calculated the minimum sample size according to 3 \ncriteria proposed by Riley et al, and compared the estimated and the actual \nsample size used. We investigated which characteris tics factors were \nassociated with the sufficient sample size. \nResults or Findings: We included 116 studies. 11/116 studies justified t he \nsample size, in which 6/11 performed a priori sampl e size calculation. The \nmean ± standard deviation (SD), median (first and third quartile, Q1, Q3) of \ntotal sample size of models are 451 ± 871, 223 (130 , 463), and those of \nsample size for training are 292 ± 676, 150 (90, 288). The mean ± SD, median \n(Q1, Q3) of difference between the total sample siz e and minimum sample size \naccording to Riley et al criterion 3 are 120 ± 888, -100 (-216, 183), and those of \ndifference between the sample size for training and  minimum sample size \naccording to Riley et al all 3 criteria are -386 ± 1264, -268 (-427, -157). The \nmodel testing method and specialty of topic were as sociated with sufficient \nsample size. \nConclusion: Radiomics models are often designed without sample size \njustification, as a consequence many models are too  small to avoid overfitting, \nnoise, and outliers. It should be encouraged to jus tify, perform and report \nsample size calculations when developing radiomics models. \nLimitations: The limitation of the study is limited number of le ading peer-\nreviewed radiological journals. \nFunding for this study: Funding was provided by National Natural Science \nFoundation of China (82302183, 82471935, 82271934),  Yangfan Project of \nScience and Technology Commission of Shanghai Munic ipality \n(22YF1442400), Research Found of Health Commission of Changing District, \nShanghai Municipality (2023QN01), Laboratory Open F und of Key Technology \nand Materials in Minimally Invasive Spine Surgery ( 2024JZWC-ZDA03, \n2024JZWC-YBA07), and Research Fund of Tongren Hospi tal, Shanghai Jiao \nTong University School of Medicine (TRKYRC-XX202204 , TRYJ2021JC06, \nTRYXJH18, TRYXJH28). \nEthics committee - additional information: The study is a meta-research \nstudy with a protocol available on OSF (https://osf .io/pbukc/), and no human \nparticipants or animals were included in the study.  \nAuthor Disclosures:  \nDefang Ding: Nothing to disclose \nYue Xing: Nothing to disclose \nHuan Zhang: Nothing to disclose \nJingyu Zhong: Board Member: Dr. Jingyu Zhong acknow ledges his position as \na member of the Musculoskeletal section of the Scie ntific Editorial Board of \nEuropean Radiology, a member of Scientific Editoria l Board of BMC Medical \nImaging, and a guest editor of the collection “AI i n radiology: revolutionizing \nmedical imaging and interpretation” of BMC Artifici al Intelligence. \nJunjie Lu: Nothing to disclose \nXianwei Liu: Nothing to disclose \nYangfan Hu: Nothing to disclose \nWeiwu Yao: Nothing to disclose \nShun Dai: Nothing to disclose \n \n \nEvolution of commercially available artificial inte lligence in radiology:  \na follow-up on peer-reviewed evidence of 179 produc ts \nN. Antonissen¹, *I. B. Houben*², O. Tryfonos³, M. D e Rooij¹,  \nK. G. Van Leeuwen⁴; ¹Nijmegen/NL, ²Zwolle/NL, ³Amsterdam/NL, ⁴De Bilt/NL \n \nPurpose or Learning Objective: To investigate changes in peer-reviewed \nevidence on commercially available radiologic artif icial intelligence (AI) \nproducts from 2020 to 2023. \nMethods or Background: A comprehensive review of the literature published \nbetween January 2015 and March 2023 of CE certified  radiological AI products \n(according to www.healthairegister.com) was perform ed. Complying with the \nprevious systematic review, this follow-up study ca tegorized the publications \naccording to the hierarchical model of efficacy: fr om technical and diagnostic \naccuracy (levels 1 and 2) to impacts on clinical de cision-making and patient \noutcomes (level 3-5) or socio-economic impact (leve l 6). \nResults or Findings: By March 2023, 91 vendors were identified, offering  a \ntotal of 179 products, with 120 of these (67%) havi ng peer-reviewed evidence, \ncompared to 36% in 2020. In 2023, there were 662 pu blications on these 120 \nproducts, compared to 237 publications on 36 produc ts in 2020. An increase \n(22 to 25%) was found in publications focusing on t echnical or potential clinical \nefficacy. The majority of publications described th e diagnostic accuracy of the \nproduct (level 2), although relatively showing a de crease (55 to 52%). For the \nhigher levels of efficacy (level 3-6) the respectiv e contribution to the total \nremained the same as 2020 (23%). \nConclusion: While there is an increase in the amount of publica tions \nvalidating AI products, the majority of publication s continue to describe the \nlower levels of efficacy. This suggests that even t hough the field has been \nmaturing, we still have limited knowledge and evide nce of the clinical impact of \nAI products in radiology. \nLimitations: Several products have a high number of publications , which may \ncause them to be overrepresented in the total. \nFunding for this study: No funding was received for this study. \n\n \n \nAbstract-based Programme \n \n 21  \nWednesday \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nIgnas Bernardus Houben: Nothing to disclose \nKicky Gerhilde Van Leeuwen: Owner: Romion Health Fo under: Health AI \nRegister \nNoa Antonissen: Nothing to disclose \nOlga Tryfonos: Nothing to disclose \nMaarten De Rooij: Research/Grant Support: Siemens H ealthineers \n \n \nFlexible Deep Learning MR Image Enhancement with Pe rformance \nMonitoring \nZ. Zhou, C. Arnold, H. Gandhi, P. Gulaka, A. Shanka ranarayanan,  \n*S. Pasumarthi Venkata*; Menlo Park, CA/US \n(srivathsa@subtlemedical.com) \n \nPurpose or Learning Objective: Deep learning (DL) MR image enhancement \nallows scan time reduction while maintaining the di agnostic quality. However, \nits performance may deteriorate over time. This stu dy aims to develop an \nadaptive image enhancement DL model and investigate s a non-reference-\nbased metric without human annotation for performan ce monitoring. \nMethods or Background: A single DL model with a ConvNeXt backbone was \ntrained on 3027 paired MR data. High-quality images  were enhanced by a \ncommercial algorithm as targets. Low-quality input images were acquired with \nvarious acceleration methods (0-80%) for model to l earn adaptive \nenhancement. The trained DL model was evaluated on another diverse set of \n205 cases. Line profiles and region-of-interests (R OIs) were manually labeled \nfor each case. The slope/gradient was extracted fro m line profiles to measure \nimage sharpness, and signal-to-noise ratio (SNR) wa s derived from ROIs to \nevaluate noise level. In addition, gradient entropy  (GE) as a non-reference-\nbased metric (lower GE higher quality) was compared  with line/ROI based \nmetrics. \nResults or Findings: Compared to inputs, over 90% of model outputs \nachieved 45% SNR increase and 8% sharpness increase . On average, SNR \nand sharpness were improved by 73% and 27%, respect ively. GE measured \non outputs was reduced by 0.5% for 95% of test case s. For test cases with \n>0.5% GE reduction, the Pearson correlation of the relative change between \nGE and SNR is -0.333 (p < 0.05), and between GE and  sharpness is 0.214 (p \n< 0.05), showing a weak but significant correlation  between GE and annotated \nimage quality (IQ) metrics. \nConclusion: The developed DL model can adaptively improve IQ su pporting \nflexible protocol acceleration. Its strong denoisin g also enables MR scans with \nhigher acceleration/resolution. In addition, gradie nt entropy can be simply \ndeployed for performance monitoring and mitigate th e risk of mis-interpretation. \nLimitations: Not applicable \nFunding for this study: NIH SBIR grant (R44MH135725) \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nSrivathsa Pasumarthi Venkata: Employee: Subtle Medi cal Inc \nCampbell Arnold: Employee: Subtle Medical Inc \nPraveen Gulaka: Employee: Subtle Medical Inc \nAjit Shankaranarayanan: Employee: Subtle Medical In c \nHarsh Gandhi: Employee: Subtle Medical Inc \nZechen Zhou: Employee: Subtle Medical Inc \n \n \nRadiologist-Guided Active Learning for Medical Imag e Segmentation: \nMoving Beyond the Dice Score to Clinically Relevant  Targets \n*B. Föllmer*, V. Serafimoski, K. Schulze, F. Biavat i, M. Bosserdt, M. Dewey; \nBerlin/DE \n(bernhard.foellmer@charite.de) \n \nPurpose or Learning Objective: Deep learning models for medical image \nsegmentation typically require extensive pixel-wise  annotations, which are \ncostly and time-consuming. Active learning can miti gate this challenge by \nlabeling only the most informative (i.e., uncertain ) cases in multiple annotation \nand training rounds. However, conventional active l earning methods do not \naccount for clinically relevant segmentation target s. This study introduces a \nradiologist-in-the-loop approach for targeted activ e learning, to optimize model \nperformance beyond standard metrics like the Dice s core, focusing on clinically \nsignificant segmentation objectives. \nMethods or Background: We propose a targeted active learning framework \nconsisting of four iterative steps: (1) Automated i dentification of uncertain \ncases for review by the radiologist, (2) Radiologis t selection of cases relevant \nto predefined clinical segmentation targets, (3) Co mbined selection of \nuncertain and clinically relevant cases, and (4) Ef ficient partial annotation and \nmodel retraining. We applied this approach to multi -class segmentation of \ncoronary arteries using the SCCT 18-segment model, evaluating it on CTAs \nfrom 300 patients of the DISCHARGE (NCT02400229) an d CAD-Man trials. \nInitial model training was conducted using standard  active learning, followed by \ntargeted active learning with three predefined obje ctives: (1) segmentation of \nrare vessels (e.g., Ramus Intermedius), (2) segment ation of thin vessels (e.g., \nR-PDA, R-PLB), and (3) segmentation of heavily calc ified segments. \nResults or Findings: Our framework demonstrated improved segmentation \nperformance and time-efficiency over standard activ e learning for the three \npredefined targets (rare vessels, thin vessels, and  calcified segments. \nConclusion: The proposed targeted active learning framework ena bles more \ntime-efficient, radiologist-guided model training f ocused on clinically relevant \nsegmentation targets, improving performance beyond conventional accuracy \nmetrics like the Dice score. \nLimitations: This framework was evaluated exclusively on coronar y artery \nsegmentation in cardiac CT, with only three segment ation targets considered. \nBroader validation is needed for other anatomical s tructures and imaging \nmodalities. \nFunding for this study: This work was funded by the German Research \nFoundation through the graduate program BIOQIC (GRK 2260, project-ID: \n289347353) and the DISCHARGE project (603266-2, HEA LTH-2012.2.4.-2) \nfunded by the FP7 Program of the European Commissio n. \nEthics committee - additional information: This study does not require any \napproval of the ethics committee. \nAuthor Disclosures:  \nKenrick Schulze: Nothing to disclose \nMarc Dewey: Other: Hands-on cardiac CT courses (www .ct-kurs.de) Other: \nEuropean Society of Radiology (ESR) Publications Ch air (2022-2025); the \nopinions expressed in this abstract/presentation ar e the author’s own and do \nnot represent the view of ESR Patent Holder: Patent  on fractal analysis of \nperfusion imaging (jointly with Florian Michallek, EPO 2022 EP3350773A1, and \nUSPTO 2021 10,991,109, approved) Author: Cardiac CT  (Springer Nature) \nEquipment Support Recipient: Siemens, General Elect ric, Philips, Canon Grant \nRecipient: EU (EC-GA 603266 in HEALTH.2013.2.4.2-2)  DFG (DE 1361/14-1, \nDE 1361/18-1, BIOQIC GRK 2260/1, Radiomics DE 1361/ 19-1 [428222922] \nand 20-1 [428223139] in SPP 2177/1), GUIDE-IT (DE 1 361/24-1), Berlin \nUniversity Alliance (GC_SC_PC 27), Berlin Institute  of Health (Digital Health \nAccelerator). \nFederico Biavati: Nothing to disclose \nMaria Bosserdt: Research/Grant Support: Received fu nding from EU-FP7 \nFramework Program (DISCHARGE EU FP EC-GA 603266). \nVladimir Serafimoski: Nothing to disclose \nBernhard Föllmer: Nothing to disclose \n \n \n10:00-11:00 Research Stage 3 \nResearch Presentation Session: Vascular \nRPS 215 \nAdvances in peripheral imaging \n \nModerator \nE. Claus; Leuven/BE  \n \n \nA Novel Human Amputated Limb Model for Advancing Pe ripheral Artery \nDisease Research and Device Testing \n*J. Csőre*, A. Crichton, B. Benfor, C. Karmonik, T. L. Roy ; Houston, TX/US \n(csore.judit@gmail.com) \n \nPurpose or Learning Objective: Traditional animal models often fail to \ncapture the complexity of peripheral artery disease  (PAD) lesions, leading to a \ngap between preclinical and clinical research in pe rcutaneous vascular \ninterventions (PVI). To address this, we developed a human amputated limb \nmodel combined with a proprietary MRI-histology pro tocol for detailed plaque \ncharacterization and simulation of PVI procedures, assessing lesion-specific \ndevice impact on the vessel wall. \nMethods or Background: Amputated limbs from end-stage PAD patients \nwere scanned using 3T or 7T MRI, incorporating Ultr ashort Echo Time and T2-\nweighted sequences to differentiate hard (collagen/ calcium) and soft \n(fat/thrombus/smooth muscle) plaque components. PVI  procedures were \nsimulated in a hybrid operating room, targeting ide ntified lesions. Device \ntesting included balloon angioplasty, lithotripsy, atherectomy, drug-coated \nballoons, and novel wires/catheters. Vessel impact was evaluated \nintraprocedurally using intravascular ultrasound, f ollowed by post-procedure \nmicro-CT and 9.4T MRI. Histopathological analysis w as performed with \nMovat’s and H&E stains. \nResults or Findings: A total of 70 amputated limbs were collected, yield ing \n133 target lesions and 2500 histologic cross-sectio ns. Key findings include: 1. \nValidation of the MRI protocol and human amputated limb model. 2. Successful \ntesting of vessel preparation devices, showing plaq ue disruption and dissection \nin calcified lesions. 3. Correlation of chronic tot al occlusion crossing success \n\n \n \nAbstract-based Programme \n \n 22  \nWednesday \nwith pre-intervention MRI histology scoring. 4. Ide ntification of calcified lesions \nas barriers to effective drug delivery. 5. Collabor ation with industry for device \ndevelopment and testing. \nConclusion: This human cadaveric model offers a unique platform  for PAD \nresearch, providing detailed insights into plaque m orphology and PVI device \nperformance. By correlating plaque characteristics with procedural outcomes, it \nenables precise device testing and fosters innovati on in vascular interventions. \nThis model informs clinical decision-making, enhanc es new technology design, \nand guides personalized treatment strategies for PA D patients. \nLimitations: Single-center study, small cohort \nFunding for this study: Jerold B. Katz Academy of Translational Science \n(project ID 15790002, recipient: Trisha Roy); Ameri can Heart Association \nTransformational Award (project ID: 17590004, recip ient: Trisha Roy); National \nInstitutes of Health Research Project grant (R01) ( award ID: R01HL174587, \nrecipient: Trisha Roy) \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board under study ID PRO000272 58. \nAuthor Disclosures:  \nAlexander Crichton: Nothing to disclose \nBright Benfor: Nothing to disclose \nJudit Csőre: Nothing to disclose \nChristof Karmonik: Nothing to disclose \nTrisha L. Roy: Research/Grant Support: Baylis Medic al Technologies, Boston \nScientific, Light Matter Interaction Founder: Magel lan Biomedical Inc. \n \n \nApplication and significance of precise CTA scannin g technology in the \nassessment of lower extremity arterial diseases \n*J. Xing*, H. Yu, L. Zhu; Shang Hai/CN \n(13636300561@139.com) \n \nPurpose or Learning Objective: Objective: This study seeks to investigate \nthe differences in image quality and radiation expo sure between an advanced \nprecision scanning technique and traditional scanni ng methods in 320-slice \ncomputed tomography angiography (CTA) of the lower limb \nMethods or Background: Methods: A cohort of 89 patients with suspected \nlower limb a arteryial disease, who underwent CTA e xamination at our \ninstitution, were randomly allocated to either grou p A or group B. In group \nA,low-dose testing was first used. TS was obtained by subtracting the time to \npeak of the dorsalis pedis artery T2 and the time t o peak of the main \nabdominal artery T1;the formal scan began at T1 + 4  seconds,and the scan \nwas completed after adjusting the entire acquisitio n time to TS by the variable \npitch method.Group B was scanned using standard pit ch.The statistical \nanalysis involved the assessment of image quality,r adiation dose, and contrast \nagent dosage. \nResults or Findings: Results:Both subjective and objective evaluations \ndemonstrated superior image quality of lower extrem ity artery in group A(all \nP<0.001).The subjective score for group A demonstra ted a significant 21% \nincrease compared to that of group B, particularly in the assessment of ankle \nand dorsum images (4.32±0.79 vs. 3.57±0.94).In terms of patients' radiation \ndose and contrast agent dosage, group A exhibited a  16.23% reduction in \nradiation dose and a 12.28% reduction in contrast a gent dosage compared to \ngroup B, respectively (both P< 0.001). \nConclusion: Conclusion: The implementation of VHP technology in lower \nextremity artery CTA scanning facilitates enhanced visualization of distal blood \nvessels and improves overall image quality, meanwhi le effectively reducing \nradiation exposure and contrast agent consumption, which presents substantial \nclinical value. \nLimitations: The sample size is relatively small and warrants ex pansion for \nfurther validation of the derived conclusions. \nFunding for this study: Young Scientists Fund of the National Natural \nScience Foundation of China (82302188) \nEthics committee - additional information: No:2019tjdx123 \nAuthor Disclosures:  \nJun Xing: Nothing to disclose \nHong Yu: Nothing to disclose \nLin Zhu: Nothing to disclose \n \n \nRun-off CT angiography with a patient-tailored post -trigger delay: \nOptimized scan timing compared with a fixed delay \n*K. Qi*, J. Liu; Zhengzhou/CN \n(qk_kkkkkkk@163.com) \n \nPurpose or Learning Objective: To validate the feasibility of using bolus \ntracking with a patient-tailored post-trigger delay  (PTD) in run-off CTA and to \ncompare image quality with that using a fixed PTD. \nMethods or Background: Participants undergoing run-off CTA with bolus \ntracking were prospectively assigned at random, coh ort A comprised 30 \nparticipants with a fixed 10-second PTD and cohort B comprised 30 \nparticipants with a patient-tailored PTD. The atten uation of abdominal and \nlower limb arteries was measured in 11 different an atomical positions in one \nleg and divided into four vascular segments accordi ng to the anatomical \nlocation: aortoiliac, femoropopliteal, tibioperonea l and foot. The mean \nattenuation, SNR and CNR of each vessel segment wer e calculated. Two \nreaders rated subjective image quality. Two-way ana lysis of variance was used \nto assess the mean attenuation of four vascular seg ments. Sidak's multiple \ncomparison was used to determine differences in att enuation between the two \ncohorts and at each anatomical location within each  cohort. Mann-Whitney test \nwas used to determine SNR and CNR between two group s, while the chi-\nsquare test compared subjective image quality score s. \nResults or Findings: Cohort B using new bolus tracking algorithm predict s the \nmean patient-tailored PTD of 12 ± 1.8 s. The demographic and frequency of \nPAD revealed no statistically significant differenc es. Cohort B showed greater \nattenuation of tibioperoneal (432±76 HU vs 364±69HU, p=0.001) and foot \n(369±79 HU vs 281±77HU, p=0.001) segments. SNR (p<0.002), \nCNR(p<0.002) and subjective image quality (excellen t or good image quality, \n96.7% vs 74.3%, p=0.038) were higher in cohort B th an in the fixed cohort. \nConclusion: Bolus tracking with a patient-tailored PTD provides  reliable scan \ntiming, resulting in improved image quality and opt imized vessel opacification \nin run-off CTA . \nLimitations: Further research is needed on the relationship betw een \nfrequency of PAD and PTD. \nFunding for this study: The Science and Technology Research Project of \nHenan Provincial Health Commission (No. 21210231014 2) \nEthics committee - additional information: Ethics Committee of Zhengzhou \nUniversity \nAuthor Disclosures:  \nJie Liu: Nothing to disclose \nKe Qi: Nothing to disclose \n \n \nShear Wave Elastography in Differentiating Acute an d Subacute \nThrombosis of Dialysis Arteriovenous Fistulas \n*Ö. Altun*, A. Dablan, M. Sam Özdemir, M. Karagülle , M. Cingöz, M. F. Arslan; \nIstanbul/TR \n(omeraltun1996@gmail.com) \n \nPurpose or Learning Objective: To evaluate the utility of shear wave \nelastography (SWE) in distinguishing between acute and subacute thrombi in \nthrombosed dialysis arteriovenous fistulas (AVFs). \nMethods or Background: This retrospective study analyzed 32 dialysis \npatients with thrombosed AVFs treated between June 2022 and June 2024. All \npatients underwent Doppler ultrasound and SWE to de termine thrombus \ncharacteristics. Based on ultrasound findings and c linical history, patients were \ncategorized into acute or subacute thrombus groups.  Thrombus stiffness was \nquantified using SWE in terms of average, median, a nd maximum kilopascal \n(kPa) values. \nResults or Findings: The study included 16 patients with acute thrombi a nd \n16 with subacute thrombi. SWE measurements revealed  significantly higher \nstiffness values in subacute thrombi compared to ac ute thrombi (p < 0.001). A \nstrong positive correlation was observed between th rombus age and SWE-\nderived kPa values (average: r = 0.770, median: r =  0.727, maximum: r = \n0.835). Receiver operating characteristic (ROC) ana lysis demonstrated SWE’s \nhigh accuracy in differentiating thrombus age, with  an optimal average cut-off \nvalue of 31.7 kPa, resulting in a sensitivity of 90 .5% and specificity of 73.9%. \nConclusion: Shear wave elastography shows promise as a non-inva sive tool \nfor differentiating between acute and subacute thro mbi in thrombosed AVFs, \naiding in personalized treatment planning for dialy sis patients. \nLimitations: This study's patient numbers was not much. \nFunding for this study: This study was not supported by any funding. \nEthics committee - additional information: All procedures performed in \nstudies involving human participants were in accord ance with the ethical \nstandarts of the instituional and/or national resea rch commitee and with the \n1964 Helsinki declaration and its later amendments or comparable ethical \nstandarts. \nAuthor Disclosures:  \nMerve Sam Özdemir: Nothing to disclose \nMehmet Karagülle: Nothing to disclose \nMustafa Fatih Arslan: Nothing to disclose \nÖmer Altun: Nothing to disclose \nMehmet Cingöz: Nothing to disclose \nAli Dablan: Nothing to disclose \n \n \nMRI-Histology as a Predictive Tool for Crossing Fai lure in Below-the-\nKnee Peripheral Arterial Disease \n*J. Csőre*, A. Crichton, C. Karmonik, B. Benfor, T. L. Roy ; Houston, TX/US \n(csore.judit@gmail.com) \n \nPurpose or Learning Objective: Recent randomized trials have challenged \nthe traditional endovascular-first approach for tre ating below-the-knee arterial \ndisease. Scoring systems like TASC and GLASS overlo ok lesion composition \nand morphology, which influence peripheral vascular  intervention (PVI) \n\n \n \nAbstract-based Programme \n \n 23  \nWednesday \nsuccess. Conventional imaging provides limited plaq ue composition insight, \nwhile ultrashort echo time (UTE) MRI can distinguis h between soft (e.g., fibrous \ntissue, thrombus) and hard (e.g., calcification, de nse collagen) plaque \ncomponents. This study aimed to assess if MRI-histo logy could better predict \nlesion crossing failure compared to GLASS and TASC scoring. \nMethods or Background: Amputated limbs were collected from patients with \nchronic limb-threatening ischemia (CLTI) and scanne d ex-vivo on a 3T MRI \nusing UTE and T2w contrasts. Lesions were classifie d as 'hard' if >50% of the \nlumen was occluded by calcium or dense collagen bas ed on the MRI. The \ndistribution of hard components (eccentric, concent ric, central), lumen stenosis \ncaused by hard/soft components, and collagen densit y were recorded. Ex-vivo \nPVIs were carried out in a hybrid operating room an d TASC and GLASS \nscoring was performed. \nResults or Findings: Seventeen patients yielded 29 target lesions, 76% \n(22/29) of which were classified as ‘hard.’ Of thes e, 45% had a collagen-\ndominated composition. 'Hard' lesions showed a sign ificantly higher crossing \nfailure rate compared to 'soft' lesions (95% vs. 14 %, p<.001). MRI scoring of \n'hard' lesions was strongly associated with crossin g success (p<.001), \noutperforming TASC and GLASS scoring (p=.062 and p= .112, respectively). \nTotal vessel occlusion was not predictive of failur e (p=0.64). Most crossing \nfailures (64%) occurred in centrally distributed 'h ard' lesions, though this was \nnot significant. \nConclusion: This MRI-histology scoring system identifies plaque  composition \nas a predictor of PVI failure, outperforming TASC a nd GLASS scoring, and \nhighlighting MRI's potential role in preoperative a ssessment and device \nselection for CLTI patients. \nLimitations: Single-center study, small cohort \nFunding for this study: Jerold B. Katz Academy of Translational Science \n(project ID 15790002, recipient: Trisha Roy); Ameri can Heart Association \nTransformational Award (project ID: 17590004, recip ient: Trisha Roy); National \nInstitutes of Health Research Project grant (R01) ( award ID: R01HL174587, \nrecipient: Trisha Roy) \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board under study ID PRO000272 58. \nAuthor Disclosures:  \nAlexander Crichton: Nothing to disclose \nBright Benfor: Nothing to disclose \nJudit Csőre: Nothing to disclose \nChristof Karmonik: Nothing to disclose \nTrisha L. Roy: Nothing to disclose \n \n \nAssessment of left renal vein areas ratios on CT-ph lebography as \nsurrogate parameter for pressure gradient in pelvic  congestion syndrome \n*T. Nemirovskaya*, R. Bredikhin, R. Akhmetzyanov, E . Fomina, D. Ryabinina, \nA. Yaglova; Kazan/RU \n(tanya.nemirovsky@gmail.com) \n \nPurpose or Learning Objective: Pelvic venous diseases are considered \nwidespread problem. Among etiological factors compr ession of left renal vein \n(LRV) between aorta and superior mesenteric artery (SMA), so-called \nnutcracker syndrome, is considered primary for symp toms evolvement. \nDiagnostic workflow includes selective phlebography , but CT-phlebography is \ngaining popularity. Primary issue is applicability of CT-phlebography results \nconcerning surgical correction selection. Study obj ective was parameters \nassessment that could be accounted surrogate charac teristics of pressure \ngradient in the left renal vein. \nMethods or Background: Prospective assessment of CT-phlebography in \npatients with pelvic congestion syndrome was perfor med. Expiratory CT \nscanning was performed with 130 s delay after Iodin ated contrast medium \nadministration. Following measurements were perform ed: maximum diameter \nof pelvic veins, gonadal veins diameter, aorta/SMA angle. Three LRV areas \nwere outlined orthogonally projected to gonadal tri butary, aorta/SMA angle, \ninferior vena cava сonflux (IVC) with areas ratio calculation. LRV narr owing \nextension was also measured. \nResults or Findings: 74 patients underwent IVC and tributaries delayed C T-\nangiography from 2022 to 2024. Mean age 37 years, 1 5 male, 59 female. All \nhad pelvic veins enlargement with associated sympto ms according to \nultrasound examination and history. No severe devel opment anomalies was \nobserved except 3 cases of retroaortal LRV. Twelve cases were accompanied \nwith May-Turner variant. Lineal regression assessme nt displayed correlation \nbetween aorta/SMA angle and maximum/minimum LRV are as ratio with p < \n0.001, also with LRV narrowing extension (p = 0.003 ). May-Turner variant was \ncontributing factor with more severe pelvic veins e nlargement (p < 0.001). \nCorrelation was found with pressure gradient accord ing to direct phlebography, \nhowever only 19 patients underwent it. \nConclusion: CT-phlebography may be considered as supportive dia gnostic \nmodality to direct phlebography for selection candi dates for surgical correction \nof pelvic congestion syndrome with LRV compression.  \n \n \nLimitations: Retrospective direct phlebography data collection w ith incomplete \ncohort coverage. \nFunding for this study: No funding \nEthics committee - additional information: Institutional ethics board of \nInterregional Clinical Diagnostic Center \nAuthor Disclosures:  \nRustem Akhmetzyanov: Nothing to disclose  \nAlina Yaglova: Nothing to disclose \nDaria Ryabinina: Nothing to disclose \nElena Fomina: Nothing to disclose  \nTatiana Nemirovskaya: Nothing to disclose  \nRoman Bredikhin: Nothing to disclose \n \n \nEvaluation of low-dose upper extremity CTA with art ificial intelligence \niterative reconstruction for hemodialysis arteriove nous fistula/graft: \nImage quality and diagnostic value of stenosis dete ction \n*B. Shou*¹, J. Li², Y. Zou², W. Zhang¹, G. Zhang², X. Hu¹, F. Jiang¹, H. Hu¹; \n¹Hangzhou, Zhejiang/CN, ²Shanghai/CN \n(Y217180070@zju.edu.cn) \n \nPurpose or Learning Objective: To assess the image quality and diagnostic \nvalue of artificial intelligence iterative reconstr uction (AIIR) in low-dose upper \nextremity CT angiography for hemodialysis arteriove nous fistula or graft \n(AVF/G). \nMethods or Background: A total of 56 patients with suspected or known \nAVF/G dysfunction were prospectively enrolled and w ere randomly divided into \ntwo groups: routine-dose group (RD-group, n=28) and  low-dose group (LD-\ngroup, n=28). RD-group employed a routine CTA proto col (tube voltage: \n100kVp; contrast dosage: 1.0ml/kg) with hybrid iter ative reconstruction, while \nLD-group used the low-dose protocol (tube voltage: 80kVp; contrast dosage: \n0.6ml/kg) with AIIR. Two radiologists independently  scored the overall image \nquality using a 4-point scale (1=poor; 4=excellent) . Area under the curve \n(AUC), accuracy, sensitivity, and specificity of tw o groups for detecting \nsignificant (>50%) stenosis were calculated on a pr e-segment basis, using \ndigital subtraction angiography (DSA) as the refere nce standard. Signal-to-\nnoise ratio (SNR) and contrast-to-noise ratio (CNR)  in fistula were also \nanalyzed. \nResults or Findings: No significant differences in demographics characte ristic \nwere observed between the two groups (all p>0.05). The radiation dose and \ncontrast dosage in LD-group were reduced by 53% (22 4.56mGyxcm vs. \n479.24mGyxcm) and 42% (36mL vs. 63mL), respectively , compared to the RD-\ngroup. The mean subjective scores between the RD-gr oup and LD-group \nshowed no significant difference (3.86±0.36 vs. 3.68±0.48, p=0.12). The AUC, \naccuracy, sensitivity, and specificity were 0.91, 9 0% (47/52 segments), 92%, \nand 99% for RD-group and were 0.94, 90% (47/52 segm ents), 100%, and 98% \non a pre-segment basis for LD-group. In fistula, SN R and CNR of LD-group \nwere 130% and 140% higher than those of RD-group, r espectively (both \np<0.001). \nConclusion: Low-dose CTA with AIIR provides superior image qual ity and \nmaintains high accuracy for detecting stenosis in A VF/G, while significantly \nreducing radiation dose and contrast dosage. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: This study was approved by the \nlocal ethics Committee at the University Hospital. \nAuthor Disclosures:  \nYixuan Zou: Nothing to disclose \nXi Hu: Nothing to disclose \nHongjie Hu: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nBeili Shou: Nothing to disclose \nFeng Jiang: Nothing to disclose \nWenming Zhang: Nothing to disclose \nJing Li: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 24  \nWednesday \n10:00-11:00 Research Stage 4 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 201 \nWhat's new in biliary diseases? \n \nModerator \nJ.-H. Yoon; Seoul/KR  \n(jhjhry@gmail.com) \n \n \nMulticenter validation of the DiStrict score, a nov el classification and \nprognostic score for individuals with primary scler osing cholangitis \n(PSC) \n*A. Grigoriadis*¹, S. G. Hamma¹, G. Kemmerich², J. S. Nayagam³,  \nK. Horsthuis⁴, M-C. Londoño⁵, D. Assis⁶, S. Charanjeet⁷, A. Bergquist¹; \n¹Stockholm/SE, ²Oslo/NO, ³London/UK, ⁴Amsterdam/NL, ⁵Barcelona/ES, \n⁶Connecticut, CT/US, ⁷New Haven, CT/US \n(aristeidis.grigoriadis@ki.se) \n \nPurpose or Learning Objective: To validate the reproducibility and prognostic \nvalue of the DiStrict-score in a multicenter intern ational cohort. \nMethods or Background: The DiStrict-score is an MRCP-based classification \nof the severity of ductal changes (ranging from 1 t o 8) and is reproducible and \nassociated with liver-transplantation and liver-rel ated death. However, it has \nnot been validated. For this retrospective multicen ter study with participation of \neight international PSC-centers, hepatologists from  each center retrieved data \nof consecutive adult PSC individuals (MRCP, demogra phics, liver-tests, PSC \ndiagnosis date, hepatobiliary cancer development, l iver- transplantation, death, \nand cause of death). Two radiologists from each cen ter applied the DiStrict-\nscore independently to the patients of their center . Cases of disagreement \nwere resolved in consensus. Interreader agreement w as assessed for each \npair of radiologists with the intraclass correlatio n coefficient (ICC), with a two-\nway random-effects model, absolute-agreement, and s ingle-measurement. The \nassociation of the DiStrict-score with outcomes (tr ansplant-free survival, \ndevelopment of hepatobiliary malignancy) was assess ed with Cox-regression. \nSurvival estimates were calculated with Kaplan-Meie r curves and the curves \nwere compared with the log-rank test. \nResults or Findings: 415 patients (271 males, 248 with ulcerative coliti s) with \nmedian diagnosis age of 39 years, were included. Du ring a median follow-up of \n84 months 101 patients developed outcomes (liver-tr ansplantation; n=78, liver-\nrelated death; n=10, hepatobiliary cancer; n=13). T he interreader agreement \nranged between 0.61 and 0.91 for the different cent ers. Patients with high \nDiStrict-scores (5–8) had a higher risk of developi ng outcomes compared to \npatients with low scores (1–4) (log-rank test; p=0. 0008) with a hazard ratio of \n1,98 (95%CI; 1.32-2.96, p=0.001). \nConclusion: The DiStrict-sore is reproducible with good interre ader agreement \nand is associated with transplant-free survival and  development of \nhepatobiliary malignancy. \nLimitations: The limitations of the study are its retrospective design and the \nnon-standardized MRCP acquisition technique. \nFunding for this study: Funding was provided by Medical Diagnostics \nKarolinska. \nEthics committee - additional information: Each center obtained ethical \napproval by local ethical committees . \nAuthor Disclosures:  \nSingh Charanjeet: Nothing to disclose \nJeremy Shanika Nayagam: Nothing to disclose \nAristeidis Grigoriadis: Speaker: Have received inst itutional honoraria by \nJANSSEN-CILAG AB \nAnnika Bergquist: Nothing to disclose  \nGunter Kemmerich: Nothing to disclose \nKarin Horsthuis: Nothing to disclose \nStefan Gmail Hamma: Nothing to disclose \nDavid Assis: Nothing to disclose \nMaría-Carlota Londoño: Nothing to disclose \n \n \nPrevalence, prognostic value, and interreader agree ment of high-grade \nstrictures in individuals with primary sclerosing c holangitis (PSC) \n*A. Grigoriadis*, S. G. Hamma, A. Bergquist; Stockh olm/SE \n(aristeidis.grigoriadis@ki.se) \n \nPurpose or Learning Objective: To assess the prevalence and reproducibility \nof the evaluation of the presence of high-grade str ictures (HGS) in MRCP, in \nPSC individuals. Moreover, to assess the predictive  value of HGS for \ndevelopment of hepatobiliary malignancy, liver-tran splantation, and liver-\nrelated death. \nMethods or Background: AASLD and EASL have introduced in their \nguidelines for PSC the term HGS defined as a strict ure seen in MRCP with \n>75% reduction of the lumen of common and/or biliar y ducts. However, the \nprevalence, reproducibility of their detection, and  their potential value for \npredicting outcomes have not been assessed. Two rad iologists independently \nassessed the presence of HGS in MRCPs of 203 indivi duals with PSC \nrecruited at Karolinska University Hospital in the SUPRIM study between 2012 \nand 2015. MRCP, demographic, clinical-laboratory an d outcome data \n(hepatobiliary malignancy, liver-transplantation, a nd liver-related death) were \nretrieved for all patients. Interreader agreement o f the evaluation of HGS was \ncalculated with the intraclass correlation coeffici ent (ICC) using a two-way \nrandom-effects model, single-measurement, and absol ute-agreement. The \nassociation of HGS with outcomes was assessed with Cox-regression. Survival \nestimates were calculated with Kaplan-Meier curves and the curves were \ncompared with the log-rank test. \nResults or Findings: After exclusion, 171 patients (103 males, 95 with \nulcerative colitis) with a median diagnosis age of 40 years were included. \nDuring a median follow-up of 124 months 49 patients  developed outcomes \n(liver-transplantation=36, liver-related death=5, h epatobiliary malignancy=8). \n80 patients (47%) had HGS. The agreement was modera te with ICC=0.72 \n(95%CI; 0.64-0.78). Patients with HGS had a higher risk to develop outcomes \n(p=0.01) with a hazard-ratio of 2.08 (95%CI; 1.17-3 .71). \nConclusion: HGS are common, can be identified with acceptable \nreproducibility, and are associated with outcomes. \nLimitations: The limitations of the study are its retrospective design and that \nno intrareader agreement analysis was performed. \nFunding for this study: Funding was provided by Medical Diagnostics \nKarolinska. \nEthics committee - additional information: The study was approved by the \nSwedish ethical review authority (2011/824-31/2, 20 18/1111-32, 2018/1494-\n31/3). \nAuthor Disclosures:  \nAristeidis Grigoriadis: Speaker: Received Instituti onal honoraria from \nJANSSEN-CILAG AB \nAnnika Bergquist: Nothing to disclose \nStefan Gmail Hamma: Nothing to disclose \n \n \nDeep Learning of Preoperative Gadoxetic Acid-Enhanc ed MRI for \nPrediction of Perineural Invasion in Intrahepatic C holangiocarcinoma \n*X. Zhou*¹, J. Hu², S-T. Feng¹; ¹Guangzhou/CN, ²Bei jing/CN \n \nPurpose or Learning Objective: To preoperatively predict the Perineural \ninvasion (PNI) in intrahepatic cholangiocarcinoma ( ICC) on gadoxetic acid \n(EOB)-enhanced MRI, the deep learning with clinical  model based fusion \nmodel was developed and evaluated. \nMethods or Background: A total of 165 patients with pathologically diagnos ed \nICC who underwent preoperative EOB-enhanced MRI wer e retrospectively \nrecruited from two independent centers (center1, tr aining set, n = 115; \nvalidation set, n = 14; internal test set, n = 15; center 2, external test set, n = \n21). The medmanba was used to extract image feature s on the pre-contrast, \narterial phase, portal venous phase, and hepatobili ary phase of MRI. These \nfeatures combined with clinical factors (such as Ne utrophils, lymphocytes, and \nserum tumor markers), and classified by a linear la yer. For comparison, a DL \nmodel was constructed by removing clinical factors,  and a clinical model was \nestablished by the random forest selection on the c linical features. Model \nperformance was evaluated with the area under the r eceiver operating \ncharacteristic curve (AUC). Gradient-weighted class  activation mapping (Grad-\nCAM) heatmaps were used to show the focus area in p redicting PNI. \nResults or Findings: The PNI positive rate was 42.4% (61/144) in center 1 \nand 28.6% (6/21) in center 2. On the internal test and external test set, the \ncombined model showed the highest AUC of 0.944 and 0.844. The DL model \nachieved the moderate AUC of 0.926 and 0.833. The p erformance of clinical \nmodel is relatively low, with AUCs of 0.852 and 0.7 11. Grad-CAM showed the \nDL model focused on the solid component of the tumo r, especially the margin \narea. \nConclusion: MRI based DL model can accurately predict PNI-posit ive ICC, \nand the tumor margin area may have important indica tions for the model. \nLimitations: Retrospective study; limited sample size. \nFunding for this study: National Natural Science Foundation of China \n(82271958) \nEthics committee - additional information: The Institutional Review Board of \nThe First Affiliated Hospital, Sun Yat-sen Universi ty(approval number: \n[2023]014) \nAuthor Disclosures:  \nShi-Ting Feng: Nothing to disclose \nJing Hu: Nothing to disclose \nXiaoqi Zhou: Nothing to disclose \n \n \n\n \n \nAbstract-based Programme \n \n 25  \nWednesday \nPreoperative prediction of IDH1-mutation and perine ural invasion in \nintrahepatic cholangiocarcinoma based on Gd-EOB-DTP A-enhanced MRI \nand MRI-derived habitats \nX. Zhou, M. Chen, *S-T. Feng*; Guangzhou/CN \n(fengsht@mail.sysu.edu.cn) \n \nPurpose or Learning Objective: To preoperatively predict isocitrate \ndehydrogenase 1 (IDH1) mutation and perineural inva sion (PNI) of intrahepatic \ncholangiocarcinoma (ICC) based on the Gd-EOB-DTPA-e nhanced MRI and \nMRI-derived habitat imaging to improve the reliabil ity and interpretability. \nMethods or Background: A total of 129 ICC patients with Gd-EOB-DTPA-\nenhanced MRI before resection between 2018 and 2024  were collected, and \nrandomly assigned to training set and the test set in a ratio of 7:3. IDH1 \nmutation and PNI status were assessed on pathologic  tissue slides. Clinical \ninformation and MRI features were qualitatively and  quantitatively collected. \nMatchable tumors in the pre- and post-enhancement T 1 mapping images were \nmanually outlined for habitat analysis and divided into five habitats based on \nkmeans clustering (Habitat 1-5). A combined nomogra m model was \nconstructed based on clinical features, MRI feature s and habitat fraction. The \ndiagnostic accuracy was evaluated using the area un der the receiver operating \ncharacteristic curves (AUCs). \nResults or Findings: The IDH1 nomogram model consists of age, T2 central  \nbrightness, liver ADC value, tumor T1 value reducti on rate and percent of \nHabitat 4, with AUCs of 0.926 and 0.924 in the trai ning and validation sets. The \nPNI nomogram model consists of CEA, tumor location,  intrahepatic bile duct \ndilation and percent of Habitat 1, with AUCs of 0.8 54 and 0.896 in the training \nand validation sets. By mapping the habitats to mul ti-sequence MRI, Habitat1 \nis located predominantly at the edge of the tumor, with signals suggestive of a \nparenchymal component representing the aggressive e dge of the tumour. \nHabitat 4 is located intratumorally, with signals s uggestive of an intratumoral \nfibrotic area with little tumor component. \nConclusion: MRI and habitat imaging can noninvasively and preop eratively \ndetermining the IDH1 mutation and PNI of ICC with g ood accuracy and \ninterpretability. \nLimitations: Retrospective study, limited sample size. \nFunding for this study: National Natural Science Foundation of China \n(82271958) \nEthics committee - additional information: The Institutional Review Board of \nThe First Affiliated Hospital, Sun Yat-sen Universi ty(approval number: \n[2023]014) \nAuthor Disclosures:  \nShi-Ting Feng: Nothing to disclose \nXiaoqi Zhou: Nothing to disclose \nMeicheng Chen: Nothing to disclose \n \n \nDistinguishing Bile Sludge from Physiological Bile Concentration on \nAbdominal MRI: Key MRI Features and Diagnostic Accu racy \n*K. Kadirli*, A. Usta, S. Sahin, A. Cantürk, S. Özk an, H. Mutlu; Istanbul/TR \n(kenankadirli7@gmail.com) \n \nPurpose or Learning Objective: Routine abdominal MRI often shows signal \nchanges in the gallbladder on T1- and T2-weighted i mages, aside from \ngallstones. Differentiating between bile sludge and  physiological bile \nconcentration is key for accurate diagnosis, but li terature shows overlapping \nsignal characteristics with limited guidance. This study aims to identify specific \nMRI findings to distinguish between bile sludge and  physiological bile \nconcentration for improved diagnosis and clinical m anagement. \nMethods or Background: This retrospective study included patients with non -\nstone signal changes on upper abdominal MRI from Ja nuary 2022 to April \n2024, who had follow-up ultrasounds within one mont h. The study involved 42 \npatients in the sludge group and 44 in the non-slud ge group according to \nultrasound findings. Key parameters assessed were l ayering, fluid-fluid levels \non T2, T1 signal characteristics, and signal loss o n out-of-phase (OOP) \nsequences. Quantitative measures included the T2 ab normal signal/spleen \nratio, ADC values, and the normal-bile/abnormal-sig nal ratios on T2 and ADC \nmaps. Multivariate regression was performed on para meters with p < 0.05, and \nthe model's diagnostic performance was evaluated. \nResults or Findings: T1-weighted hypointensity and fluid-fluid levels we re \nstrongly associated with bile sludge (p < 0.001), w hile layering and OOP signal \nloss associated with normal bile(p<0,001). The best  multivariate model, using \nT1W, fluid levels, and OOP signal loss, showed 83.7 2% specificity, 72.22% \nsensitivity, and 78.48% accuracy, with an AUC of 0. 889. \nConclusion: An abnormal signal in the gallbladder lumen on MRI that does \nnot exhibit OOP signal loss but shows fluid-fluid l evels and T1-weighted \nhypointensity may be indicative of bile sludge. \nLimitations: This study is limited by its retrospective design, which may \nintroduce selection bias. Additionally, ultrasound was used as the gold \nstandard, biochemical evaluation of the bile was no t conducted. \nFunding for this study: No funding \nEthics committee - additional information: Ethics committee approval was \nobtained from the relevant institution. \nAuthor Disclosures:  \nAnıl Usta: Nothing to disclose \nKenan Kadirli: Nothing to disclose \nSuat Özkan: Nothing to disclose \nAli Cantürk: Nothing to disclose \nHakan Mutlu: Nothing to disclose \nSerdar Sahin: Nothing to disclose \n \n \n11:30-12:30 Research Stage 1 \nResearch Presentation Session: Neuro \nRPS 311 \nIlluminating the brain: neuroimaging \ninsights into epilepsy and \nneuroinflammation \n \nModerator \nS. Gerevini; Cremona/IT  \n \n \nExpanding Language Assessment in Epilepsy Patients through the \nIndividual Functional Connectome \n*R. Stepponat*, M. Berger, L. Schäfer, M. S. Yildir im, J. Leinkauf,  \nF. Fischmeister, S. Bonelli, G. Kasprian; Vienna/AT  \n(radheshyam.stepponat@meduniwien.ac.at) \n \nPurpose or Learning Objective: Pre-surgical language evaluation has \nremained largely unchanged for years, underscoring the need for improved \nmethods. Given that epilepsy and language both oper ate as network functions, \nanalyzing the individual functional connectome may offer greater clinical \nprecision, enabling more accurate predictions and b etter-informed surgical \ndecisions for enhanced patient care. \nMethods or Background: Language is predominantly left-lateralized, and \nfMRI is commonly used for preoperative evaluation o f lateralization by \ncalculating the lateralization index (LI). However,  a survey by Benjamin et al. \nhighlights significant uncertainties in fMRI result s among clinicians. In this \nstudy, 46 patients with temporal lobe epilepsy and 25 healthy controls \nunderwent preoperative fMRI. Data preprocessing was  performed with \nfMRIPrep, and correlation maps were created using t he CONN toolbox. Seed-\nbased connectivity (SBC) analysis of core language areas, based on \nFredarenko et al. 2024, was conducted, and a LI bas ed on graph \nmeasurements (degree) was calculated. This was comp ared to traditional LI \nanalysis and neuropsychological data, with all anal yses conducted in native \nspace. \nResults or Findings: Results indicated that the connectivity-based \nlateralization index (LI) provided a more consisten t measure of language \nlateralization compared to standard LI methods, ali gning better with \nneuropsychological assessments. Patients with lesio nal epilepsy showed \ngreater variability in lateralization compared to M R-negative patients, while \nhealthy controls exhibited stronger left-lateraliza tion as expected. The use of \nseed-based connectivity (SBC) analysis enhanced the  detection of individual \ndifferences in language network organization, under scoring its potential clinical \nvalue in preoperative assessment. \nConclusion: The initial findings suggest that connectivity-base d LI may \nimprove the accuracy of preoperative language asses sments. This approach \noffers a promising enhancement over traditional LI methods, providing more \nreliable insights for surgical planning and potenti ally leading to better \npredictions of post-surgical cognitive outcomes. \nLimitations: -No neuropsychological data for the controls. \n-Retrospective data. \nFunding for this study: This study has been conducted as part of a PhD-\nthesis at the medical university of Vienna. Nothing  to disclose. \nEthics committee - additional information: EK-Number: 1141/2023 \nAuthor Disclosures:  \nMehmet Salih Yildirim: Nothing to disclose \nLaurin Schäfer: Nothing to disclose \nJoel Leinkauf: Nothing to disclose \nFlorian Fischmeister: Nothing to disclose \nMarc Berger: Nothing to disclose \nSilvia Bonelli: Nothing to disclose \nRadheshyam Stepponat: Nothing to disclose  \nGregor Kasprian: Nothing to disclose \n \n \n\n \n \nAbstract-based Programme \n \n 26  \nWednesday \nTemporal lobe epilepsy with isolated amygdala enlar gement: anatomo-\nelectro-clinical features and long-term outcome \nM. Ferro¹, J. Ramos², *F. M. Doniselli*³, G. Didato ³; ¹Lisbon/PT, ²Gaia/PT, \n³Milan/IT \n(fabio.doniselli@gmail.com) \n \nPurpose or Learning Objective: This study focuses on the radiological \ncharacteristics of patients with temporal lobe epil epsy and isolated amygdala \nenlargement (TLE-AE).We aim to assess the imaging f indings, particularly in \nrelation to amygdala size and signal alterations, a nd their correlation with \nclinical and pathological data to guide better diag nostic and therapeutic \ndecisions. \nMethods or Background: We conducted a retrospective analysis of 143 brain \nMRI scans from adult patients at a tertiary neurolo gy center. Forty-one patients \nwith TLE-AE were selected. Imaging was reviewed by two neuroradiologists for \namygdala size and T2-hyperintensity. Fluorodeoxyglu cose-PET (FDG-PET) \ndata were also included for a subgroup of patients. Amygdala signal alterations \nwere quantified and correlated with clinical, neuro physiological, and \npathological findings. Statistical analyses assesse d relationships between \nimaging features and clinical outcomes. \nResults or Findings: Out of 41 patients, 32% had bilateral amygdala \nenlargement, while the remainder had unilateral fin dings. T2-hyperintensity \nwas noted in 65.9% of cases, significantly correlat ing with amygdala \nenlargement. FDG-PET showed temporomesial hypometab olism in 64% of \npatients, further supporting the epileptogenic invo lvement of the \namygdala.Surgical pathology confirmed various under lying etiologies, including \ngliosis, low-grade tumors, and inflammatory infiltr ates. Post-surgical outcomes \nwere favorable, with 70.6% of operated patients bei ng seizure-free at last \nfollow-up. \nConclusion: Radiologically, amygdala enlargement with or withou t T2-\nhyperintensity is a significant marker of TLE, ofte n associated with underlying \nstructural abnormalities or inflammation. FDG-PET i s a valuable adjunct for \nidentifying hypometabolic regions corresponding to AE, supporting its role in \nthe epileptic network.MRI findings, combined with s urgical evaluation, can \nimprove clinical outcomes in TLE-AE patients, parti cularly those with drug-\nresistant epilepsy. \nLimitations: The study's retrospective nature and variability in  MRI protocols \nover time limit the consistency of radiological fin dings. Additionally, not all \npatients underwent advanced imaging techniques, suc h as PET or follow-up \nMRIs, reducing the ability to track longitudinal ch anges. \nFunding for this study: None. \nEthics committee - additional information: Retrospective study. \nAuthor Disclosures:  \nFabio Martino Doniselli: Nothing to disclose \nMargarida Ferro: Nothing to disclose \nJoao Ramos: Nothing to disclose \nGiuseppe Didato: Nothing to disclose \n \n \nThe relationship of glutamate and glutamine and met abolic profiling in \nfocal epilepsy using 7T CRT-FID-MRSI \n*S. Chambers*, H. Shayeste, P. Lazen, L. Haider, E.  Pataraia, G. Kasprian,  \nW. Bogner, S. Trattnig, G. Hangel; Vienna/AT \n(stefanie.chambers@meduniwien.ac.at) \n \nPurpose or Learning Objective: Identifying epileptogenic foci is essential in \ntherapy-planning and predictive for post-operative seizure freedom in epilepsy. \nIn this work we present a novel MRSI technique (CRT -FID) at 7T, allowing for \nultra-high resolution whole-brain maps in focal epi lepsy. We offer a qualitative \nanalysis of its feasibility in identifying and char acterizing metabolic alterations \nover multiple pathologies. \nMethods or Background: Following informed written consent, forty-two \npatients with focal epilepsy (16-52 years, 21 femal es/21 males) underwent a \n3D-MRSI protocol in 15min with a 3.4 mm isotropic r esolution at 7T using a \n32Rx/1Tx-coil. Data processing involved spectral qu antification and ratio \nmapping of Glu, Gln, Ins, tCho, tCr and NAA normali zed to NAA and tCr. \nFurthermore, the concentration estimates of Glu and  Gln were qualitatively \nassessed in seizure onset zones. \nResults or Findings: Though we could find no consistent metabolic patter n \nthroughout pathologies, ratio maps showed high dete ctability of lesions, \nidentifying alterations in seizure onset zones in 7 8,6% when normalized to \nNAA. This detection rate was reduced to 71,2% when normalized to creatine. \nOf the assessed ratios, Ins/tNAA and Cho/tNAA showe d highest stability \nacross patients, whereas Glu/tNAA and Gln/tNAA show ed varying directionality \nof changes. Assessing these changes further in rela tion to clinical parameters \nsuch as the seizure frequency, we found a trend for  relative increases of \nGln/Glu in patients with low seizure frequencies an d the inverse for high \nseizure frequencies. \nConclusion: 7T CRT-FID-MRSI shows promising results in identify ing \nmetabolic alterations in patients suffering from fo cal epilepsy, even in the \nabsence of structural lesions. Furthermore, this or ientational study may point to \nan altered glutamate/glutamine cycle in epilepsy, p otentially the result of \naltered energy metabolism pathways in chronic epile psy. \nLimitations: This study is limited by the small cohort size and qualitative \nnature of the analysis. \nFunding for this study: This research was funded by the FWF grant \n10.55776/KLI1121, of the Mayor of the Federal Capit al Vienna (Project \nNumber 21186). \nEthics committee - additional information: Ethic committee number: EK \n1039/2020 \nAuthor Disclosures:  \nWolfgang Bogner: Nothing to disclose \nPhilipp Lazen: Nothing to disclose \nStefanie Chambers: Nothing to disclose \nLukas Haider: Nothing to disclose \nEkatarina Pataraia: Nothing to disclose \nSiegfried Trattnig: Nothing to disclose \nGilbert Hangel: Nothing to disclose \nGregor Kasprian: Nothing to disclose \nHaniye Shayeste: Nothing to disclose \n \n \nMultiparametric MR-based assessment supports the in flammatory nature \nof symptomatic CSF HIV Escape \nS. Capelli¹, A. Caroli¹, G. Pezzetti², F. Ferretti³ , R. Vercesi⁴, P. Cinque⁴,  \n*S. Gerevini*²; ¹Ranica/IT, ²Bergamo/IT, ³London/UK , ⁴Milan/IT \n \nPurpose or Learning Objective: Symptomatic cerebrospinal fluid (CSF) HIV \nescape is an infrequent but severe condition occurr ing in persons living with \nHIV (PLWH) undergoing combination antiretroviral th erapy (cART). It is \ncharacterized by HIV-RNA in the CSF despite being u ndetectable in plasma. \nSimilarly to HIV encephalitis in cART-untreated PLW H (HIV-ENC), HIV CSF \nescape (HIV-ESC) is accompanied by neurological imp airments and brain MRI \nalterations. This study aimed to investigate the ne uroimaging features of HIV-\nESC in comparison to HIV-ENC and neuro-asymptomatic  controls. \nMethods or Background: Brain structural and microstructural alterations we re \nquantified in: HIV-ESC (n=12), HIV-ENC (n=10), neur o-asymptomatic PLWH \n(n=11) and neuro-asymptomatic HIV-negative controls  (n=12). The quantitative \nanalysis included measurements of normalized FLAIR signal intensity, \nApparent Diffusion Coefficient (ADC) from diffusion -weighted MRI, and brain \ntissue volumes from T1-weighted MRI. \nResults or Findings: Both HIV-ESC and HIV-ENC demonstrated significantly  \nhigher FLAIR signal intensity in white matter (WM),  elevated ADC values in \nboth white and gray matter (GM) and reduced GM volu mes as compared to \nneuro-asymptomatic controls, while the HIV-ESC grou p had higher WM \nvolumes compared to HIV-ENC. In the HIV-ESC group, GM ADC values were \nnegatively correlated with nadir CD4+ T-cell counts , while GM volume showed \na positive correlation. In contrast, in HIV-ENC, WM  ADC, FLAIR signal \nintensity, and WM volume all positively correlated with nadir and current CD4+ \nT-cell counts. \nConclusion: WM hyperintensities and increased ADC values in HIV -ESC and \nHIV-ENC reflect active WM damage, while reduced GM volumes are indicative \nof long-term brain atrophy. However, the higher WM volume in HIV-ESC \nsuggests persistent inflammation. These findings, a long with their correlation to \nlaboratory data, support the hypothesis that inflam mation is the primary \nmechanism of brain damage in HIV-ESC. \nLimitations: Retrospective study with limited patient sample and  lack of a \nstandardized MRI protocol across all subjects. \nFunding for this study: National Institutes of Health (NIH); University of \nCalifornia, San Francisco (UCSF) \nEthics committee - additional information: The study was approved by \n“Comitato Etico IRCCS Ospedale San Raffaele” on 15/ 12/2016 (ref. 235/2015) \nAuthor Disclosures:  \nSerena Capelli: Nothing to disclose \nGiulio Pezzetti: Nothing to disclose \nAnna Caroli: Nothing to disclose \nFrancesca Ferretti: Nothing to disclose \nRiccardo Vercesi: Nothing to disclose \nPaola Cinque: Nothing to disclose \nSimonetta Gerevini: Nothing to disclose \n \n \nUnravelling PIRA brain atrophy pattern and progress ion \n*T. Fakhreddine*¹, A. Tamanti¹, C. Salvatore², D. C alderaro¹, S. Ziccardi¹,  \nM. Calabrese¹, A. Fattorini¹, I. Castiglioni², F. B . B. Pizzini¹; ¹Verona/IT, \n²Milan/IT \n(tom.fakhreddine@gmail.com) \n \nPurpose or Learning Objective: Progression independent of relapse activity \n(PIRA) is the primary factor contributing to irreve rsible disability accumulation \nin relapsing multiple sclerosis (MS). This study ai med to investigate differences \nin brain volumes through Magnetic Resonance Imaging  (MRI). \n\n \n \nAbstract-based Programme \n \n 27  \nWednesday \nMethods or Background: We conducted a retrospective study with MS \npatients with and without cognitive PIRA who perfor med T1-weighted 3D MRI \nstudy (at baseline and at 24-months follow-up) at t he Verona University \nHospital. The TRACE4AD™ medical-device software (De epTrace \nTechnologies, Italy) performed automatic structural  brain segmentation from \nMR scans into 53 regions and calculated correspondi ng total volumes and \nasymmetry indices, defined as the percentage differ ence between brain region \nvolumes on each side. Atrophy progression was measu red as the percentage \ndifference in volumes at follow-up versus baseline.  A statistical comparison of \nvolumes at baseline and atrophy progression was per formed using a two-\nsample t-test between MS patients with and without PIRA. \nResults or Findings: We collected 95 MRIs from 40 patients with PIRA (26  \nwith follow-up) and 55 patients without PIRA (43 wi th follow-up). At baseline, \nthe difference in asymmetry indices of total white matter, gray-matter occipital-\nlobe volume and parieto-occipital cortex volume wer e statistically significant \n(p<0.05): PIRA patients showed more right-side atro phy in the white matter \nvolume and less in the right occipital lobe and par ieto-occipital cortex than non-\nPIRA patients. Atrophy progression was statisticall y different (p<0.05) for the \ngray-matter volume of the right temporal lobe, bein g faster in patients with \nPIRA than in those without (-3.1% vs -1.3%). \nConclusion: These results indicate that atrophy progression may  be faster in \nMS patients with PIRA in regions related to early d ementia and non-verbal \nlanguage functions. \nLimitations: Cohort size and statistical power show potential fo r improvement. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective. \nAuthor Disclosures:  \nDavide Calderaro: Nothing to disclose \nIsabella Castiglioni: Nothing to disclose \nAgnese Tamanti: Nothing to disclose \nChristian Salvatore: Nothing to disclose \nTommaso Fakhreddine: Nothing to disclose \nMassimiliano Calabrese: Nothing to disclose \nFrancesca Benedetta Benedetta Pizzini: Nothing to d isclose \nAnna Fattorini: Nothing to disclose \nStefano Ziccardi: Nothing to disclose \n \n \nMean Upper Cervical Cord Area (MUCCA) in MOGAD comp ared to MS, \nNMOSD and healthy controls \n*E. Lotan*¹, V. Anania², T. Billiet², I. Kister¹, I. Lotan³; ¹New York, NY/US, \n²Leuven/BE, ³Petach Tikva/IL \n(vincenzo.anania@icometrix.com) \n \nPurpose or Learning Objective: Relatively little is known about how mean \nupper cervical cord area (MUCCA) changes in MOGAD c ompared to MS, \nNMOSD, and healthy controls (HC). We aim to assess MUCCA values in \nMOGAD as compared to MS, NMOSD, and HC. \nMethods or Background: We retrospectively reviewed the NYU Multiple \nSclerosis Care Center database to identify all adul t MOGAD patients with \navailable brain MRI performed in stable remission a nd compared them with \nNMOSD and MS patients and HC. Cross-sectional MUCCA  from T1 brain \nMRIs was quantified using icobrain ms+ (version 5.1 5.0) and normalized for \nhead size. A linear modeling analysis was used to e valuate the impact of \ncovariates on cross-sectional MUCCA. The covariates  were age, T1 slice-\nthickness, sex, and group. Post hoc testing was con ducted using estimated \nmarginal means (EMMEANS) to evaluate group differen ces while controlling \nfor covariates. \nResults or Findings: 20 MOGAD patients, 37 AQP4+ NMOSD patients, 40 \nMS patients, and 31 HC were included in the analysi s. Age, sex, and group \nshowed significant effects on MUCCA measurements. T he EMMEANS of \nMUCCA values were lower for the NMOSD group (86.1+/ -1.5), followed by \nMOGAD (89.3+/-1.7), MS (90.3+/-1.2), and HC (91.6+/ -1.5). Pairwise \ncomparison between groups showed no statistically s ignificant differences \nbetween the MOGAD and other groups. In contrast, a statistically significant \ndifference between the NMOSD and HC groups and a tr end towards \nsignificance between the NMOSD and MS groups were o bserved. \nConclusion: Our proof of concept study shows the feasibility of  calculating \ncervical volume loss from routine brain MRI using a utomated software. While \ngreater spinal cord tissue loss is evident in NMOSD , the degree of spinal cord \ntissue loss in MOGAD is lower and not significantly  different compared to MS \nand HC. Additional analyses on a larger cohort are underway. \nLimitations: N/A \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under study number i20-01556 \nAuthor Disclosures:  \nEyal Lotan: Nothing to disclose \nVincenzo Anania: Employee: icometrix \nThibo Billiet: Employee: icometrix \nIlya Kister: Nothing to disclose \nItay Lotan: Nothing to disclose \nMedial lemniscus as a diagnostic marker: differenti ating multiple \nsclerosis from small vessel disease \n*W. H. E. Hamed*, D. Werring, D. S Lynch, R. Jäger,  T. A. Yousry; London/UK \n(weaam.hamed@nhs.net) \n \nPurpose or Learning Objective: Evaluate the diagnostic utility of medial \nlemniscus (ML) in differentiating between multiple sclerosis (MS) and small \nvessel disease(SVD). \nMethods or Background: Distinguishing between MS and SVD remains a \nsignificant challenge in the elderly due to their o verlapping clinical \npresentations, emphasising the necessity for reliab le, non-invasive \nneuroimaging markers such as ML. A retrospective st udy analysed 270 MRI \nscans(100 MS, 170 SVD). SVD subtypes included arter iosclerotic(50), cerebral \namyloid angiopathy(CAA)(50), mixed(50), and genetic (20). The signal intensity \nof ML in the pontine tegmentum was assessed visuall y on T2-w and FLAIR \nimages. Statistical analysis included univariable t ests to identify differences \nbetween MS and SVD, followed by multivariable logis tic regression to \ndetermine independent predictors of ML involvement.  \nResults or Findings: SVD patients were significantly older than MS patie nts \n(mean age:68±13vs43±10 years, p<0.001) and had lower female \npredominance (44.7%vs64%). MS patients had no ML in volvement(87%) or \nunilateral involvement(13%) with no bilateral cases , while 38% of SVD patients \nhad bilateral involvement(p<0.001). Among SVD categ ories, ML involvement \nwas most frequent in genetic(80%), followed by mixe d(64%) and \narteriosclerotic(46%). The least affected was CAA(2 0%). In MS cohort, ML \nhyperintensity was associated with higher lesion lo ad(p<0.001) but not the \ndisease duration(p=0.4). Among SVD subtypes, the mu ltivariable analysis \nrevealed that lesion load(p<0.001) and hypertension (p=0.002) were the \nstrongest predictors of ML involvement, followed by  age(p=0.007), \ndiabetes(p=0.023), and hypercholesterolemia(p=0.048 ). Factors such as \nalcohol, gender, and smoking were not significant p redictors. \nConclusion: Our study establishes ML as a reliable radiological  marker for \ndistinguishing MS from SVD, particularly when there  is bilateral involvement, \nevidenced by a statistically significant presence i n SVD and a notable absence \nin MS. Moreover, the variation in ML involvement ac ross SVD subtypes, \nspecifically its minimal presence in pure CAA, sugg ests its potential role in \ndifferentiating specific SVD categories. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: A retrospective study \nAuthor Disclosures:  \nTarek A. Yousry: Nothing to disclose \nWeaam Hamed Elsayed Hamed: Nothing to disclose \nRolf Jäger: Nothing to disclose \nDavid Werring: Nothing to disclose \nDavid S Lynch: Nothing to disclose \n \n \nBrain Disconnection and Atrophy Assessment Multiple  Sclerosis \nConverters \n*S. Hannoun*, S. Ghazal, L. Halawi, C. Al-Dahouk, N . El-Ayoubi, S. Khoury; \nBeirut/LB \n(sh156@aub.edu.lb) \n \nPurpose or Learning Objective: Differentiating patients who convert from \nrelapsing-remitting Multiple sclerosis (RRMS) to se condary progressive MS \n(SPMS) remains a critical challenge, as early ident ification of converters can \nsignificantly impact treatment strategies. This stu dy explores specific brain \nregions associated with disconnection probabilities  and volume reductions, \naiming to identify potential MRI biomarkers predict ive of conversion. This study \naimed to investigate whether distinct patterns of W M disconnection and \nregional brain atrophy are associated with RRMS con version to SPMS. \nMethods or Background: We retrospectively analyzed 47 RRMS patients (17 \nConverters and 30 non-converters) who underwent bas eline and follow-up MRI \nscans approximately 1.5 years apart. Mixed-effects models evaluated the \ninteraction between conversion status (converters v s. non-converters) and time \nacross various brain regions, focusing on disconnec tion probabilities and \nvolumetric changes measured using the Vol2Brain too l. \nResults or Findings: Converters exhibited significant disconnection in k ey \nwhite matter tracts, including the uncinate fascicu lus, corticobulbar tract, \nsuperior longitudinal fasciculus, and cingulum para hippocampal parietal. These \ndisruptions are linked to cognitive, emotional, and  motor functions. Additionally, \ngrey matter atrophy was more pronounced in converte rs, particularly in the \nprecentral gyrus, temporal lobe, and thalamus. Lesi on burden and volume, \nespecially in juxtacortical areas, were greater in converters, with increased \nthird ventricle volume indicating more severe brain  atrophy. \nConclusion: Specific patterns of white matter disconnection and  regional brain \natrophy are associated with conversion from RRMS to  SPMS. These MRI \nbiomarkers provide valuable insights into disease p rogression and offer \npotential therapeutic targets. Further validation i n larger cohorts is needed to \nintegrate these findings into clinical practice. \n\n \n \nAbstract-based Programme \n \n 28  \nWednesday \nLimitations: A limitation of our study is its relatively small s ample size and \nretrospective nature, which may limit the generaliz ability of the findings. \nFunding for this study: No funding. \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board (IRB), and all participa nts provided informed \nconsent. \nAuthor Disclosures:  \nSamia Khoury: Nothing to disclose \nLean Halawi: Nothing to disclose \nSalem Hannoun: Nothing to disclose \nNabil El-Ayoubi: Nothing to disclose \nSola Ghazal: Nothing to disclose \nCezar Al-Dahouk: Nothing to disclose \n \n \n11:30-12:30 Research Stage 2 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 305 \nCareer, workforce issues and radiologist \nvisibility \n \nModerator \nF. Mankertz; Greifswald/DE  \n \n \nRadiological discrepancy review: a novel, customisa ble, cloud-based \ntechnology to implement the REALM paradigm \n*P. Brennan*¹, Y. Hughes-Roberts², J. Richenberg², I. Francis², M. Suleiman¹; \n¹Sydney/AU, ²West Sussex/UK \n(patrick.brennan@detectedx.com) \n \nPurpose or Learning Objective: To develop a geographically-limitless \nREALM infrastructure that facilitates an optimised and reflective educational \nactivity. \nMethods or Background: Reviewing radiological discrepancies is a well-\nknown activity that promotes diagnostic excellence,  encourages reflection and \nminimises future errors. Formalisation of discrepan cy reviews can be seen with \nthe Royal College of Radiology REALM (Radiology Eve nts and Learning \nMeetings) program where radiologists in the UK can engage both as authors \nand recipients. However, to promote a clinically-re alistic REALM activity, \ncomprehensive radiologic interactions with full res olution anonymised images \nshould occur in a geographically limitless way. The  infrastructures to support \nthis, are elusive. An existing technology DxCARES w as modified for this \nactivity. This incorporated: multi-modality viewing  capabilities and 3D and multi-\nplanar reconstructions (MPR): advanced AI-powered a nonymization and \nmasking engines to automatically remove sensitive d ata while maintaining \ndiagnostic integrity; a user-friendly interface for  users with different levels of \nexpertise; a web-based architecture supporting high  performance DICOM \nstreaming and real time interaction with large data  sets. \nResults or Findings: We have built a new technology which allows clinici ans \nto upload REALM cases from PACs systems and distrib ute across a \ngeographically-limitless health enterprise to limit less recipients. Customisable \nauthorship of each case is available so that each c reator can demand from \nrecipients the type of case-specific interaction or  reflection that is required, \nregardless of image or pathology-type. The multifun ctional cloud-based viewer \nallows the examination and manipulation of all case s as would occur with a \nprimary diagnostic workstation. All REALM outputs a nd interactions are \navailable for review and CPD accumulation. The new technology is currently \nbeing implemented across clinical centres. \nConclusion: This new technical innovation should promote widesp read \nclinically-realistic REALM engagement by simplifyin g time/cost implications. \nThe potential of radiological discrepancy review sh ould be maximised. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nJonathan Richenberg: Nothing to disclose \nPatrick Brennan: Founder: DetectedX Pty Ltd \nIan Francis: Nothing to disclose \nYnyr Hughes-Roberts: Nothing to disclose \nMoayyad Suleiman: Founder: DetectedX Pty Ltd \n \n \nSex differences in inappropriate imaging requests: Insights from the \nMedical Imaging Decision And Support (MIDAS) trial \n*S. Dijk*¹, C. Wollny², T. Kroencke², M. G. M. Huni nk¹; ¹Rotterdam/NL, \n²Augsburg/DE \n(stijntjedijk@gmail.com) \n \nPurpose or Learning Objective: We analyzed sex-related disparities in \ninappropriate imaging requests using data from the Medical Imaging Decision \nAnd Support (MIDAS) trial. \nMethods or Background: This study analyzed baseline data from the MIDAS \ntrial, a multi-center cluster randomized trial cond ucted at three German \nacademic hospitals. The study population encompasse d all imaging requests \nsubmitted to the 26 participating departments via t he computerized order entry \nsystem during a 15-month period. Imaging appropriat eness was assessed \nusing the ESR iGuide, a clinical decision support s ystem (CDSS). Requests \nwere categorized as inappropriate if imaging was de emed unlikely to be \nindicated or if the potential risks outweighed the benefits for the patient. Chi-\nsquare tests were employed to compare the proportio n of inappropriate \nimaging requests between men and women, with a sign ificance level of 0.01. \nSecondary analyses explored differences in inapprop riate requests by age \ngroup and exam type, with a Bonferroni correction a pplied to account for \nmultiple testing. \nResults or Findings: Women had more inappropriate imaging requests \n(7.32%) than men (6.08%; χ² = 37.176, p < 0.001, OR 1.22 [95%CI 1.12-1.33]). \nThis disparity was particularly evident in the 25-6 5 age group and for MRI \nexaminations. Further research is needed to explore  the underlying causes of \nthis discrepancy, including potential differences i n physician awareness of \nguidelines, adherence to guidelines, or limitations  in the CDSSs ability to \naccount for female-specific factors. \nConclusion: In our study clinicians were 22% more likely to req uest \ninappropriate imaging for women than for men across  nearly all modalities and \nage-groups. While the absolute percentage-point dif ferences were small \n(1.24%), the disparity warrants further investigati on \nLimitations: Our analysis judges each request independently, wit hout \nconsidering the cumulative impact of requests per i ndividual, underuse, or time \nbetween symptom onset and imaging. Our data only di stinguished male/female \nsexes. \nFunding for this study: The MIDAS study was funded by the German \nInnovation Fund (reference: Förderkennzeichen 01VSF 18008). \nEthics committee - additional information: Approval from the Medical Ethics \nReview Committee was obtained under protocol number s 20-069 (Augsburg), \nB 238/21 (Kiel), 20-318 (Lübeck) and 2020-15125 (Ma inz). The trial is \nregistered in the ClinicalTrials.gov register under  registration number \nNCT05490290. \nAuthor Disclosures:  \nThomas Kroencke: Nothing to disclose \nClaudia Wollny: Nothing to disclose \nMyriam G. M. Hunink: Nothing to disclose \nStijntje Dijk: Nothing to disclose \n \n \nMulticenter and multimodality evaluation of radiolo gical workload and \ndevelopment of a benchmarking metric \n*P. Dankerl*, J. Lang, A. Glaser, H. P. Beyer, M. F orsting; Dortmund/DE \n(Peter.Dankerl@evidia.de) \n \nPurpose or Learning Objective: The increasing complexity and volume of \nradiological examinations have led to growing conce rns about radiologists' \nworkload, diagnostic efficiency and accuracy. The a im of this study was to \nevaluate and compare radiological workload in a mul ticenter and multimodality \nanalysis while creating a benchmarking metric. \nMethods or Background: Over 100 days radiologists’ reading times for all \nexaminations and modalities across 34 centers have been collected utilizing \nRIS-export, while grouping these into 67 different body regions. In order to sort, \nretrieve and evaluate the various output formats fr om the RIS-data, a uniform \nnomenclature was introduced and all examinations re ceived these additional \nunique identifying labels. For benchmarking, report ing times were translated \ninto relative value units – as defined by us as the  mean reporting time of all X-\nray exams and termed RADPoints. \nResults or Findings: We examined 290.748 examinations and found \nsignificant variations in average reporting times a cross modalities and body \nregion, e.g. abdominal MRI 14.83min and CT 13.59min . Reporting times varied \nconsiderably, with the highest average times observ ed in complex \nexaminations such as cardiac MRI (26.04min) and CT (16.59min). Conversely, \nregions like the fingers showed much lower averages , e.g. 10.95min for MR \nand 9.83min for CT. We found our relative value uni t time of one RADPoint to \nequivalent to 2.23min which further served as commo n devisor in order to \nallocate specific RADPoints to all body region spec ific examinations. \nConclusion: The findings underscore the need for targeted workl oad \nmanagement strategies in radiology departments, par ticularly when high-\ncomplexity cases are in the mix. Benchmarking repor ting times across \nmodalities and body regions by the means of present ed RADPoints provides a \n\n \n \nAbstract-based Programme \n \n 29  \nWednesday \ncritical reference for optimizing radiologist workl oad, potentially leading to \nenhanced diagnostic accuracy and efficiency. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Retrospective evaluation and \nblinding of patinet as well as doctor identyfiers m akes this not applicable \nAuthor Disclosures:  \nMichael Forsting: Employee: Evidia \nPeter Dankerl: Employee: Evidia \nHaemi Phaedra Beyer: Employee: Evidia \nJochen Lang: Employee: Evidia \nAndrzej Glaser: Employee: Evidia \n \n \nAssessing the perceived impact of ESOR training pro grams on \nradiologists' professional development \n*J. Gregory*¹, M. L. Kofoed-Ottesen², B. Lindlbauer ², C. Loewe², V. Vilgrain¹; \n¹Clichy/FR, ²Vienna/AT \n(jules.gregory@aphp.fr) \n \nPurpose or Learning Objective: This study evaluates the perceived impact of \nEuropean School of Radiology (ESOR) training progra ms on radiologists' \nprofessional development. \nMethods or Background: A cross-sectional survey targeted alumni who \nparticipated in ESOR fellowships from 2011 to 2023.  The survey included \nquestions on demographics, professional background,  ESOR program details, \nand career impact. Data were collected via a web-ba sed questionnaire and \nanalyzed using descriptive statistics and thematic analysis. \nResults or Findings: A total of 190 participants responded, with a media n age \nof 31 years (range 29-33), and 54% were female. Mos t worked in public \nhealthcare (62%) and were involved in academic acti vities (24%). Fellowship \ntypes included Visiting Scholarship Program (44%), Bracco Fellowship (32%), \nand Exchange Program for Fellowships (25%). The maj ority (59%) reported the \nfellowship helped them reach their current position , and 35% noted it upgraded \ntheir CV. Significant application of learned skills  was reported by 69%. Ongoing \ncooperation with former tutors was maintained by 54 %. Financial support was \ncrucial, with 41% stating they could not have compl eted the training without it, \n33% considering it very important, and 13% deeming it important. Participants \nrated the impact on clinical skills with a median s core of 9 out of 10. Other \nareas of impact included research skills (median sc ore 7), subspecialization \n(median score 9), exposure to diverse practices (me dian score 9), networking \nopportunities (median score 10), and personal and p rofessional growth \n(median score 10). \nConclusion: ESOR training programs significantly enhance radiol ogists' \nprofessional development through comprehensive supp ort, high-quality \ntraining, and substantial financial aid, ensuring p articipants are well-equipped \nfor career advancement. \nLimitations: This study has limitations, including reliance on s elf-reported \ndata, potential recall-bias, and a 20% response rat e. The survey may not fully \nrepresent all ESOR alumni, and program heterogeneit y could influence the \ngeneralizability of results. \nFunding for this study: None \nEthics committee - additional information: Given the nature of the survey \ninvolving professional feedback without sensitive p ersonal data, ethical \napproval was not required. However, all participant s were informed about the \npurpose of the survey and the anonymous handling of  their data. \nAuthor Disclosures:  \nJules Gregory: Nothing to disclose \nBrigitte Lindlbauer: Board Member: European School of Radiology Office \nValérie Vilgrain: Board Member: European School of Radiology Office \nChristian Loewe: Board Member: European School of R adiology Office \nMathias Lange Kofoed-Ottesen: Board Member: Europea n School of \nRadiology Office \n \n \nExpanded AI learning: AI as a tool for human learni ng \nS. Faghani¹, C. Tiegs-Heiden², M. Moassefi², G. Pow ell², M. Ringler², \nB. J. Erickson², *N. Rhodes*²; ¹Minneapolis, MN/US,  ²Rochester, MN/US \n(nickgrhodes@hotmail.com) \n \nPurpose or Learning Objective: To use artificial intelligence (AI) as a \nteaching tool to identify new imaging findings and improve the radiologist’s \nability to recognize subtle imaging findings withou t additional AI assistance. \nMethods or Background: We studied the learning of a new task by humans \nusing a deep learning (DL) model that can identify sex differences from frontal \nknee radiographs with high accuracy. We then ascert ained imaging features \nvia occlusion interpretation maps (“heat maps”) to help human readers improve \ntheir ability to perform this task. Three human rea ders were tasked to classify \n50 frontal knee radiographs into male and female se x. They were then \ninformed of the patient’s sex and were given the as sociated AI-derived “heat \nmaps” for subsequent study. After two weeks, the gr oup was retested with a \nnew set of 50 radiographs. \nResults or Findings: The DL model categorized sex with 0.96 accuracy. Th e \naverage accuracy of the 3 human readers was initial ly 0.59 (range: 0.56-0.66). \nAfter study of AI-derived “heat maps” and associate d radiographs, the average \naccuracy of the human readers increased to 0.80 (ra nge: 0.76-0.84), a \nstatistically significant improvement (p=0.0270). \nConclusion: We believe this improvement serves as a proof of co ncept for the \nuse of AI as a tool for discovery science to advanc e human learning, an idea \nthat we have not seen advanced in the radiology lit erature. \nLimitations: 1. This is an education item that does not fall wel l into an AI or \nMSK section. 2. Understanding how AI algorithms wor k is limited. 3. Our study \nis small. Large enough to prove an improvement afte r training, but too small to \ndissect more. \nFunding for this study: None. \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nBradley James Erickson: Nothing to disclose \nMana Moassefi: Nothing to disclose \nGarret Powell: Nothing to disclose \nShahriar Faghani: Nothing to disclose \nChristin Tiegs-Heiden: Nothing to disclose \nMichael Ringler: Nothing to disclose \nNicholas Rhodes: Nothing to disclose \n \n \nFrom observation to interpretation: elevating repor ting skills in radiology \n*S. Ramirez*; Bogota/CO \n(sandritamrt@gmail.com) \n \nPurpose or Learning Objective: The preparation of a radiological report \ninvolves integrating knowledge, skills, and trainin g. While many articles offer \nrecommendations for quality reports, there is a lac k of literature on teaching \nstrategies for efficiently constructing these repor ts. Despite this, some \nresidents and radiologists have developed this comp etency over time, \nrevealing a gap in understanding how residents acqu ire the critical thinking \nskills needed. This work aims to determine how radi ology residents learn to \ncreate their reports. \nMethods or Background: A qualitative multi-case study was conducted \ninvolving three cases represented by residents of a  radiology residency \nprogram, the program's faculty, and a group of spec ialists in high-quality report \nproduction. The study included document reviews of the syllabus, semi-\nstructured interviews, and a focus group to capture  experiences, perspectives, \nand learning processes, as well as the factors and strategies that influenced \nthe development of the knowledge and skills needed for creating radiological \nreports. \nResults or Findings: Information was collected on concepts, perceptions,  \nbeliefs, emotions, interactions, and experiences to  construct coherent \nmeanings about the topic. Common patterns and chall enges in learning to \ncreate radiological reports were identified. Partic ipants described the hidden \ncurriculum as as the main component of the educatio nal process through \nwhich residents learn to write reports. Preparing p reliminary reports was \nregarded as the most valuable pedagogical activity for developing critical \nthinking skills. All residents agreed on the need t o design educational activities \nfocused on learning to prepare radiological reports . \nConclusion: This study represents, to our knowledge, the first research on \nhow the critical thinking skills necessary for crea ting a radiological report are \nacquired. By understanding how a person effectively  acquires a skill, teaching \nstrategies can be designed and adapted to maximize meaningful learning. \nLimitations: None \nFunding for this study: No \nEthics committee - additional information: Ethics committee Fundación \nUniversitaria Sanitas \nAuthor Disclosures:  \nSandra Ramirez: Nothing to disclose \n \n \nThe use of Instagram in medical education: gender d ifferences and \nstudent Satisfaction and gender differences compare d to traditional \nlearning platforms (Blackboard) \n*M. Alvarez García*, M. E. Pueyo, P. Boldó, C. Urta sun Iriarte,  \nA. Ezponda Casajus, P. Chico, J. Pueyo Villoslada; Pamplona/ES \n(malvarezgar@alumni.unav.es) \n \nPurpose or Learning Objective: This study aimed to explore the student \nsatisfaction with Instagram as an educational tool in radiology education for \nfourth-year medical students, compared to a traditi onal learning platform. \nMethods or Background: The study focused on gender differences and \noverall preferences in platform usage among fourth- year medical students \n(64% female). Both Instagram and the traditional pl atform were used to provide \nmultiple-choice test questions with images based on  radiology class content. \nStudents chose their preferred platform for complet ing quizzes, and their \nexperiences were assessed via anonymous satisfactio n surveys. Statistical \nanalysis explored gender-based differences in platf orm usage, preferences, \n\n \n \nAbstract-based Programme \n \n 30  \nWednesday \nand perceived learning benefits. The Mann-Whitney U  test was applied to \nanalyze the results. \nResults or Findings: A total of 65.8% of participants chose Instagram as  their \npreferred platform for answering clinical cases. No  significant gender \ndifferences were observed in overall platform choic e. The majority (71%) of \nstudents expressed moderate to high satisfaction wi th Instagram as a \ncomplementary learning tool. A significant gender d ifference (p=0.019) was \nfound in the reasons for preferring Instagram: wome n favored its user-friendly \ninterface, while men prioritized its speed. No sign ificant differences were found \nbetween genders regarding Instagram's usefulness fo r personal study or \nwhether it caused distractions (p=0.31). Additional ly, 97.1% of students \npreferred Instagram over other social networks. \nConclusion: Instagram was generally preferred over traditional learning \nplatforms and other social media platforms for radi ology education, with a \nmoderate to high level of student satisfaction. \nLimitations: Certain students do not use Instagram. There might be missing \ndata. Instagram tool “Stories” only allows users (s tudents) to see the content \nduring a period of 24h. \nFunding for this study: None \nEthics committee - additional information: This study has no ethical \nimplications \nAuthor Disclosures:  \nPablo Chico: Author: Study design, Reviewer \nCesar Urtasun Iriarte: Author: Data analysis, revie wer \nMaría Elena Pueyo: Author: Data analysis \nPatricia Boldó: Author: Data analysis \nJesus Pueyo Villoslada: Author: Study design, data analysis, reviewer \nMiguel Alvarez García: Author: Study design, data c ollection, data analysis, \nreviewer. \nAna Ezponda Casajus: Author: Reviewer \n \n \nStock photos lead to inaccurate representations of radiologists by the \nmedia \n*L. Hartog*¹, J. M. L. Bosmans², J. Bouziotis², P. M. Parizel³, A. Snoeckx²; \n¹Antwerp/BE, ²Edegem/BE, ³Perth/AU \n(laurahartog@live.nl) \n \nPurpose or Learning Objective: Value-based radiology requires thorough \nunderstanding by patients of the role of radiologis ts in healthcare. News media \nfrequently enlighten content on radiology with seem ingly outdated images from \nstock photo databases. We conducted a critical eval uation of the ability of \nstock photos in three major databases to reliably r epresent contemporary \nradiology. \nMethods or Background: We collected two hundred consecutive photos from \nthree major commercial databases, using the keyword  ‘radiologist’. Each \nimage was evaluated for overall quality, image desc ription, descriptive \nelements, displayed body regions, diagnostic or int erventional context, \ndepicted modality, inclusion of medical professiona ls, and accurate display on \nthe viewing device. \nResults or Findings: Just 6% of the stock photos received a high overall  \nquality score. Radiographs were the most frequently  depicted modality (52%), \nfollowed by MRI (21%) and CT (19%), with a focus pr imarily on diagnostic \nimaging (99%). On 4% of the stock photos, images we re displayed on a PACS \nworkstation, while 66% showed a radiologist holding  up films ‘in the air’, and \n10% on an X-ray viewbox. In 48% of the photos, the lead person was shown \nwearing a stethoscope. In all, only 9% of the stock  photos presented a realistic \nview of a radiologist’s job. \nConclusion: The vast majority of stock photos fail to accuratel y portray the \ncontemporary role of radiologists, contributing to gross misinterpretation of our \nprofession by the general public. \nLimitations: The evaluation was done by a radiology resident but  reviewed by \na consultant radiologist. To our knowledge, this ki nd of evaluation of an \nessential source of information for the public has never been undertaken \nbefore. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: No patient data was used for this \nstudy. \nAuthor Disclosures:  \nAnnemiek Snoeckx: Nothing to disclose \nPaul M. Parizel: Nothing to disclose \nJason Bouziotis: Nothing to disclose \nJan Maria Lodewijk Bosmans: Nothing to disclose \nLaura Hartog: Nothing to disclose \n \n \n \n \n \n11:30-12:30 Research Stage 3 \nResearch Presentation Session: \nInterventional Radiology \nRPS 309 \nImage-guided tissue biopsies and thyroid \ninterventions \n \nModerator \nG. T. Yusuf; London/UK  \n(Gibran.yusuf@nhs.net) \nAuthor Disclosures:  \nGibran Timothy Yusuf: Speaker: Siemens, Samsung, GE , Terumo, Bracco \n \n \nDiagnostic accuracy of core needle biopsy in patien ts with \nlymphoproliferative disorders: an optimized protoco l in 478 patients \n*P. Marra*, L. Dulcetta, R. Muglia, F. S. Carbone, A. Weber, S. Ferrari,  \nA. Rambaldi, P. A. Erba, S. Sironi; Bergamo/IT \n(pmarra@asst-pg23.it) \n \nPurpose or Learning Objective: Surgical excision biopsy of lymph nodes \nstands as the gold standard for histological charac terization of \nlymphoproliferative disorders (LD). However, contem porary clinical practice \nincreasingly leans towards core needle biopsy (CNB) . This study seeks to \nexplore the factors influencing the diagnostic yiel d of CNB in LD. \nMethods or Background: This unicentric retrospective study presents data \nfrom patients referred for suspicion of new or rela psing LD. All patients \nunderwent image-guided CNB of the target lesion bas ed on CT/PET findings. \nThe primary endpoint was the diagnostic outcome, co mparing the ability to \nachieve a definitive diagnosis according to interna tional guidelines with CNB \nversus the necessity for subsequent excisional biop sy. \nResults or Findings: We enrolled 478 consecutive patients undergoing CNB , \ncategorized into two cohorts. Cohort A comprised pa tients who underwent \nCNB using 18-20G full-core Menghini needles, with a  median macroscopic \nfragment dimension of 1 cm. Cohort B included patie nts who underwent CNB \nwith 16-18G semiautomatic guillotine needles, with a median macroscopic \nfragment dimension of 1.5 cm. In cohort A, the rate s of diagnostic and non-\ndiagnostic (or non-sufficiently detailed) CNBs were  95 (73%) versus 35 (27%), \nrespectively. In cohort B, these rates were 299 (86 %) versus 49 (14%). \nConclusion: The type and size of the needle used for CNB, as we ll as the \nhistologic variant of LD, emerged as factors influe ncing diagnostic yield and \naccuracy. Given the swiftness of CNB compared to su rgical excision, \noptimizing this technique could streamline the diag nostic and therapeutic \nworkflow for patients with suspected LD. \nLimitations: Retrospective study; lack of control group undergoi ng surgery \nFunding for this study: None \nEthics committee - additional information: Comitato Etico di Bergamo - \nLymphoid Cancer Registry (NCT03131531) \nAuthor Disclosures:  \nSandro Sironi: Nothing to disclose \nAlessandro Rambaldi: Nothing to disclose \nPaola Anna Erba: Nothing to disclose \nPaolo Marra: Nothing to disclose \nFrancesco Saverio Carbone: Nothing to disclose \nRiccardo Muglia: Nothing to disclose \nAlessandra Weber: Nothing to disclose \nSilvia Ferrari: Nothing to disclose \nLudovico Dulcetta: Nothing to disclose \n \n \nThe interplay of time and angle with the incidence of Pneumothorax in a \nCT-guided Lung Biopsy \n*N. Maalouf*¹, M. Abou Mrad¹, R. Benayed¹, R. A. Pu gliesi², J. C. Apitzsch¹; \n¹Pforzheim/DE, ²Stuttgart/DE \n(nourmaalouff@gmail.com) \n \nPurpose or Learning Objective: This study evaluates the relationship \nbetween the needle-pleura angle and the duration of  needle traversal (NTD) \nthrough lung tissue during CT-guided lung biopsies,  and their impact on the \nincidence of pneumothorax. \nMethods or Background: 96 patients (54 m, 42 f, median age: 71 years) \nunderwent CT-guided lung biopsies between January 2 020 and March 2024. \nProcedures were performed using a semi-automatic 18 G needle and a 17G \ntrocar. The minimum delta (δmin) was calculated as the absolute difference \nbetween a 90° angle and the measured angles to the pleura and correlated \n\n \n \nAbstract-based Programme \n \n 31  \nWednesday \nwith pneumothorax occurrence. NTD was recorded from  needle puncture to \nretraction. Patients with immediate intraprocedural  pneumothorax were \nexcluded. A multivariate analysis compared four pat ient groups, categorized by \nδmin (greater or less than ten degrees) and NTD (les s than or more than six \nminutes). \nResults or Findings: 96 biopsies were performed, with six patients exclu ded. \nPneumothorax occurred in 22 out of 90 procedures. S ix of these 22 patients \nhad δmin greater than 10 ° and NTD over six minutes; fiv e had δmin less than \nten degrees and NTD under six minutes. In contrast,  68 patients did not \nexperience pneumothorax. Of these, 29 had δmin less than ten degrees with \nNTD under six minutes, while 20 had δmin less than ten degrees with NTD \nover six minutes. Pneumothorax occurred in 14.7% of  patients with δmin less \nthan ten degrees and NTD under six minutes, compare d to 33.3% with δmin \ngreater than ten degrees and NTD over six minutes. \nConclusion: Optimizing the needle-pleura angle and minimizing n eedle \ntraversal duration during CT-guided lung biopsies c an reduce pneumothorax \nrisk. \nLimitations: A relatively small number of patients, as it was a single-center \nobservational study. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Approved by the local ethics \ncommittee (F-2021-038) \nAuthor Disclosures:  \nRosa Alba Pugliesi: Nothing to disclose  \nJonas Christoph Apitzsch: Nothing to disclose \nMazen Abou Mrad: Nothing to disclose \nRoua Benayed: Nothing to disclose \nNour Maalouf: Nothing to disclose \n \n \nSensitivity and specificity of FNAC with ROSE of lu ng lesions: a single-\ncenter experience on 643 patients \n*C. Verde*, L. Tarotto, S. Stilo, R. D'Angelo, V. S toia, V. Sanna, N. Martucci,  \nL. Arenare, F. Fiore; Naples/IT \n(caterina.verde@libero.it) \n \nPurpose or Learning Objective: The purpose of this study was to evaluate \nsensitivity and specificity of FNAC with ROSE in th e diagnosis of lung lesions. \nMethods or Background: The study was conducted at the INT of Naples \n“Foundation Pascale” between 2013 and 2017. CT, CBC T FNAC was \nperformed on 643 patients, of which 195 subsequentl y underwent surgical \nresection. Exclusion criteria were: platelet count (< 50,000) and INR (> 1.5). \nUnenhanced TC scans or CBCT were performed pre-biop sy. A 18 G coaxial \nneedle is used and a thinner needle (23-22G) is ins erted into the lesion (2-3 \ntimes, if necessary). FNAC is associated with the e xtemporaneous examination \n(ROSE), which evaluates the adequacy of the sample.  Pneumothorax is the \nmost frequent complication but it is asintomatic in  most cases and only 7 % of \ncases requests pleural drainage. No evidence of oth er major complications. \nResults or Findings: Comparing surgical cytological and histological sam ples, \npositive cytological samples are neoplastic in 99.3 % of cases (152/153) and \nnon-neoplastic in only 0.6% of cases (1/153). The s ensitivity of FNAC is \n86.8%, the positive predictive value 99.3% and the specificity 75%. \nConclusion: FNAC is a reference diagnostic tool in the characte rization of \nlung lesions for the purpose of target therapy and immunotherapy. It is an \neffective procedure with high sensitivity and speci ficity and low complication \nrate. Extemporaneous testing increases sensitivity,  reducing the number of \ninadequate samples and false negatives. \nLimitations: Single center study. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: No information provided by the \nsubmitter. \nAuthor Disclosures:  \nVincenzo Stoia: Nothing to disclose \nNicola Martucci: Nothing to disclose \nFrancesco Fiore: Nothing to disclose \nCaterina Verde: Nothing to disclose \nRoberto D'Angelo: Nothing to disclose \nVeronica Sanna: Nothing to disclose \nLuca Tarotto: Nothing to disclose \nLaura Arenare: Nothing to disclose \nSalvatore Stilo: Nothing to disclose \n \n \nTargeted Fine-Needle Aspiration of Thyroid Nodules Guided by Shear \nWave Elastography: A Novel Diagnostic Approach \n*M. Khaleghi*¹, A. Aziz Ahari²; ¹Tehran/IR, ²Boston , MA/US \n(Mokh.med2025@gmail.com) \n \nPurpose or Learning Objective: The aim of this study is to investigate target \nthyroid nodules based on shear wave elastography. \nMethods or Background: Suspicious nodules requiring fine-needle aspiration  \n(FNA) can sometimes be heterogeneous. Even though t hey may appear \nhomogeneous on ultrasound at times, they are hetero geneous on shear wave \nelastography examination. Sampling from highly susp icious areas in shear \nwave elastography evaluation can lead to decreased inadequate samples and \nincreased detection of malignant thyroid diseases. In the present study, \npatients referred for thyroid nodule sampling under go elastography \nassessment and are included in the study if they ex hibit heterogeneity. They \nthen undergo thyroid FNA under ultrasound guidance using a 22-gauge spinal \nneedle with stylet. \nResults or Findings: As of today, five patients have been examined. In t hese \npatients, inadequate or inappropriate samples were not observed. Suspicious \nareas in elastography in these patients had high in dices above 80 kPa, while \nnon-suspicious areas were below 40 kPa. All these p atients were diagnosed \nwith papillary thyroid cancer. \nConclusion: Thyroid nodules have always been a significant chal lenge. Shear \nwave elastography presents a new criterion for bett er diagnosis of suspicious \nnodules. Target thyroid FNA as a new concept should  receive attention. \nLimitations: One of the limitations of this study is the necessi ty of a highly \nskilled radiologist who can accurately sample high- risk areas. \nFunding for this study: No Funding is received. \nEthics committee - additional information: This research has been reviewed \nand approved by the National Ethics Committee under  the number \nIR.IUMS.FMD.REC.1399.177 \nAuthor Disclosures:  \nMohammadreza Khaleghi: Nothing to disclose \nAlireza Aziz Ahari: Nothing to disclose \n \n \nComparison of radiofrequency ablation and microwave  ablation in the \ntreatment of benign thyroid nodules \n*P. Glielmo*, G. Mauri, D. Albano, S. Gitto, S. Fus co, L. M. Sconfienza; \nMilan/IT \n \nPurpose or Learning Objective: Radiofrequency ablation (RFA) and \nmicrowave ablation (MWA) are minimally invasive tec hniques used to treat \nbenign thyroid tumours. This study aims to compare the efficacy and safety of \nthese two methods. \nMethods or Background: We retrospectively evaluated all patients with \nbenign thyroid nodules treated with either RFA or M WA at our Istitution \nbetween January 2021 and December 2021. The primary  outcomes assessed \nwere the volume reduction rate (VRR) of the ablated  areas at 1, 6, 12 and 24 \nmonths, procedure duration, and complication rates.  \nResults or Findings: A total of 56 patients were enrolled, 35 treated wi th RFA \nand 21 with MWA. At 1 month, the VRR was 57% in the  MWA group and 48% \nin the RFA group (p=0.045). At 6 months, both group s showed a VRR of 72%. \nAt 12 months, the VRR was 75% in the MWA group and 76% in the RFA group \nand at two years of 76% in MWA group and 75% in RFA  group. Both \ntechniques achieved significant volume reduction wi th no major complications \nreported. \nConclusion: Both RFA and MWA are effective and safe for treatin g benign \nthyroid nodules. MWA demonstrated a higher initial VRR at 1 month, while \nboth techniques achieved similar efficacy at 6, 12 and 24 months. These \nfindings support the use of either method as viable  non-surgical alternatives for \npatients seeking treatment options for benign thyro id nodules. \nLimitations: Limitations of this study include its retrospective  design, relatively \nsmall sample size, and absence of a control group f or direct comparison. \nFunding for this study: None \nEthics committee - additional information: “This study was approved by the \nEthical Committee of IRCCS Ospedale Galeazzi - Sant 'Ambrogio, and all \nparticipants provided informed consent in accordanc e with the Declaration of \nHelsinki.” \nAuthor Disclosures:  \nPierluigi Glielmo: Nothing to disclose \nStefano Fusco: Nothing to disclose \nSalvatore Gitto: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nGiovanni Mauri: Nothing to disclose  \nDomenico Albano: Nothing to disclose \n \n \nImage-guided thermal ablation as an alternative to surgery for papillary \nthyroid microcarcinoma, a 7-year experience \nG. Mauri, *P. Glielmo*, D. Albano, S. Gitto, S. Fus co, L. M. Sconfienza; \nMilan/IT \n \nPurpose or Learning Objective: Thermal ablation has emerged as a \nminimally invasive treatment option for thyroid mic rocarcinomas, offering an \nalternative to surgery. This study presents a 7-yea r experience (2018-2024) in \ntreating thyroid microcarcinomas with thermal ablat ion. \nMethods or Background: We retrospectively evaluated all patients who \nunderwent US-guided thermal ablation for thyroid mi crocarcinomas between \nJanuary 2018 and September 2024. We assessed the te chnical efficacy of the \nprocedure, complications, and local or distant recu rrences. \n\n \n \nAbstract-based Programme \n \n 32  \nWednesday \nResults or Findings: Of the 59 patients referred for evaluation, 5 were \ndeemed unsuitable for thermal ablation, and 1 opted  for surgery. The \nremaining 53 patients (40 females, 13 males; mean a ge 51.4 ± 7.7 years) \nunderwent thermal ablation successfully. The proced ure was well-tolerated \nwith no major adverse events. The follow-up period ranged from 71 to 2 \nmonths (mean 36 months). No local or distant recurr ences occurred. \nConclusion: Thermal ablation is a safe and effective treatment for thyroid \nmicrocarcinomas, providing a viable alternative to surgery. This 7-year \nexperience supports its role as a standard treatmen t option for thyroid \nmicrocarcinomas, demonstrating excellent outcomes i n terms of safety and \nefficacy in tumour control. \nLimitations: Limitations of this study include its retrospective  design, relatively \nsmall sample size, and absence of a control group f or direct comparison. \nFunding for this study: None \nEthics committee - additional information: This study was approved by the \nEthical Committee of IRCCS Ospedale Galeazzi - Sant 'Ambrogio, and all \nparticipants provided informed consent in accordanc e with the Declaration of \nHelsinki. \nAuthor Disclosures:  \nPierluigi Glielmo: Nothing to disclose \nStefano Fusco: Nothing to disclose \nSalvatore Gitto: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nGiovanni Mauri: Nothing to disclose \nDomenico Albano: Nothing to disclose \n \n \nTransarterial Embolization Outperforms Radiofrequen cy Ablation for \nThyroid Goiters Exceeding 100 mL: A Study on Effica cy and Safety \nW-C. Lin, *Y. J. Lee*, C-K. Wang, A-N. Lin, Y-S. Ch en, C. Y. Lee, P-L. Chiang, \nC. Y. Lu; Kaohsiung City/TW \n(wesly128@cgmh.org.tw) \n \nPurpose or Learning Objective: This study aims to compare the efficacy and \nsafety of radiofrequency ablation (RFA) and transar terial embolization (TAE) in \nmanaging large benign thyroid nodules (BTNs), parti cularly those exceeding \n100 mL, where established guidelines are limited. \nMethods or Background: This retrospective multicenter study, conducted \nfrom January 2018 to May 2022, included 70 patients  with a total of 76 large \nBTNs. Of these, 53 underwent RFA and 17 underwent T AE. Nodules were \ncategorized by initial volume (<50 mL, 50–100 mL, > 100 mL) and diameter (<6 \ncm, 6–9 cm, >9 cm). Treatment efficacy was evaluate d using the volume \nreduction rate (VRR) at 6 months. Complications, as  well as improvements in \nsymptom and cosmetic scores, were documented and an alyzed. \nResults or Findings: At 6 months, TAE demonstrated a significantly highe r \nmean VRR than RFA (p = 0.007), especially for nodules larger than 100 mL \n(TAE: 63.34% vs. RFA: 49.71%; p = 0.035). The complication rate in the TAE \ngroup (5.88%) was lower than that in the RFA group,  where transient \nhoarseness and hematoma were common complications. Both treatments \nresulted in significant improvements in symptom and  cosmetic scores \n(p < 0.001), with TAE providing greater improvements in larger nodules. \nConclusion: TAE is more effective and has fewer complications t han RFA for \nthe treatment of large BTNs exceeding 100 mL. These  findings suggest that \nTAE may serve as a minimally invasive alternative t o surgery for patients with \nlarge thyroid nodules. Further prospective studies are necessary to develop \nsize-specific guidelines for selecting between RFA and TAE. \nLimitations: Our study's limitations include the absence of gros s pathological \nconfirmation, retrospective design, short-term foll ow-up, and variations in TAE \nprotocols. \nFunding for this study: No funding was received for this article. \nEthics committee - additional information: Chang Gung Medical Foundation \nInstitutional Review Board IRB No.: 202401138B0 \nAuthor Disclosures:  \nYun Ju Lee: Nothing to disclose \nCheng-Kang Wang: Nothing to disclose  \nWei-Che Lin: Nothing to disclose \nPi-Ling Chiang: Nothing to disclose \nChia Yin Lu: Nothing to disclose \nYueh-Sheng Chen: Nothing to disclose \nAn-Ni Lin: Nothing to disclose \nChih Ying Lee: Nothing to disclose \n \n \n \n \n \n \n11:30-12:30 Research Stage 4 \nResearch Presentation Session: \nMusculoskeletal \nRPS 310 \nSelected applications of MSK ultrasound \n \nModerator \nF. Zaottini; Genoa/IT  \n(federico.zaottini.fz@gmail.com) \n \n \nIncidence of Palmar Fibromatosis Nodules Following Carpal Tunnel \nRelease: A Prospective High-Resolution Ultrasound S tudy \n*S. A. Jengojan*¹, A. Piacentini¹, F. P. Papa¹, V. König¹, D. Albano², Ž. Snoj³, \nG. Ivanac⁴, G. Bodner¹; ¹Vienna/AT, ²Milan/IT, ³Ljubljana/SI,  ⁴Zagreb/HR \n(suren.jengojan@meduniwien.ac.at) \n \nPurpose or Learning Objective: Palmar fibromatosis (PF), also known as \nDupuytren’s disease (DD), is characterized by fibro us nodules and thickening \nof the palmar fascia, potentially leading to functi onal impairment of the hand. \nWhile its etiology remains unclear, we have observe d over and over again in \nour daily routine that surgical procedures, such as  carpal tunnel syndrome \n(CTS) release may trigger the development of fibrom atosis in the palmar \nFascia. In this study we investigated whether patie nts who undergo CTS \nsurgery are more likely to develop palmar fibromato sis nodules, compared to \nhealthy individuals and those with non-operated CTS , using high-resolution \nultrasound (HRUS) imaging. \nMethods or Background: In this prospective study, we examined 100 patients  \nwho had previously undergone CTS surgery, screening  for palmar fibromatosis \nnodules with HRUS. We further evaluated two control  groups: 50 healthy \nvolunteers and 30 patients diagnosed with CTS who h ad not yet had surgery. \nThe prevalence of PF nodules across these groups wa s documented and \ncompared. \nResults or Findings: Out of the 100 patients who had CTS surgery, 35% \n(n=35) were found to have PF nodules. In contrast, only 1 individual in the \ngroup of healthy volunteers had nodules 2% (n=1). N one of the patients with \nnon-operated CTS showed evidence of nodules. These findings suggest a \nhighly significant increase in PF among post-CTS su rgery patients compared \nto both healthy controls and those with non-operate d CTS. \nConclusion: Our findings suggest that CTS release surgery may b e linked to a \nhigher incidence of DD. This raises important quest ions about the role of \nsurgical intervention in the development of fibroma tosis. \nLimitations: A limitation of our study is the lack of preoperati ve imaging, which \ncould not exclude the possibility that the palmar f ibromatosis nodules already \nexisted before CTS surgery. \nFunding for this study: No funding was needed \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board. (vote Number 2028/2024)  approved the study. \nAuthor Disclosures:  \nAlessio Piacentini: Nothing to disclose \nGerd Bodner: Nothing to disclose \nFrancesco Pio Papa: Nothing to disclose \nSuren Armeni Jengojan: Nothing to disclose \nGordana Ivanac: Nothing to disclose \nŽiga Snoj: Nothing to disclose \nViktoria König: Nothing to disclose \nDomenico Albano: Nothing to disclose \n \n \nThe effect of soft tissue compression on shear wave  velocity of \nperipheral nerves \nJ. Peterca, Ž. Snoj, *G. Omejec*; Ljubljana/SI \n(gregor.omejec@gmail.com) \n \nPurpose or Learning Objective: The objective was to determine the effect of \nsoft tissue compression applied by the US probe on peripheral nerve shear \nwave velocity (SWV) measurements \nMethods or Background: Lower and upper limbs of 20 fresh cadavers were \ndissected at hip and shoulder region, strapped to i nox tray and emerged into \nthe 37°C water bath. The US probe was mounted on ro botic arm Yaskawa \nMH6 (Yaskawa Electric, Fukuoka, Japan). Median nerv e at wrist and forearm, \nand sciatic nerve just proximal to tibial and fibul ar nerve branching were \nscanned longitudinally using US Resona R9 (Mindray Bio-medical Electronics, \nShenzhen, China) and L14-3WU transducer. The compre ssion of the soft \ntissue applied by US probe was gradually increased from no contact to 10mm \n(wrist), 20mm (forearm), and 30mm (posterior thigh)  in successive steps of 1, \n\n \n \nAbstract-based Programme \n \n 33  \nWednesday \n2, 3, 5, 10, 15, 20, 25 and 30mm. After every succe ssive step, SWV was \nmeasured three times and median was used for statis tical analysis. \nResults or Findings: Average distances between the US probe and \nperipheral nerve at baseline were 5.4mm for median nerve at the wrist, \n18.9mm for median nerve at the forearm, and 24.8mm for sciatic nerve. For \nmedian nerve at wrist, significant changes in SWVs were found for \ncompression distances of ≥2mm, and for median nerve at forearm for \ncompression distances of ≥15mm (p <0.05). No changes in SWV were found \nfor sciatic nerve. \nConclusion: The influence of soft tissue compression on SWV mea surements \ndepends on the thickness of the soft tissue between  the probe and peripheral \nnerve. No or only minor pressure is advised when me asuring SWV of \nsuperficial peripheral nerves and more pressure can  be applied when \nmeasuring SWV of deeper peripheral nerves. \nLimitations: Possible small changes in ROI position after roboti c arm \nmovement \nFunding for this study: The authors report grants from The Republic of \nSlovenia Research Agency (Grant No. P3-0338 and J3- 4507) \nEthics committee - additional information: The study was approved by the \nNational Medical Ethics Committee of Slovenia (Nr. 0120-25/2023/3). \nAuthor Disclosures:  \nGregor Omejec: Nothing to disclose \nJošt Peterca: Nothing to disclose \nŽiga Snoj: Nothing to disclose \n \n \nDominance differences in hamstring stiffness among athletes: Insights \nfrom ultrasound shear wave elastography \n*B. S. Alvarez De Sierra*, P. Nieto; Madrid/ES \n \nPurpose or Learning Objective: Hamstring injuries are prevalent among \nfootball players, necessitating a deeper understand ing of muscle stiffness and \nits implications. This study investigates the diffe rences in hamstring stiffness \nbetween dominant and non-dominant legs in athletes using Shear Wave \nElastography (SWE). \nMethods or Background: A retrospective study was conducted involving 30 \nfootball players (15 males, 15 females). Ultrasound  SWE was used to measure \nshear wave velocities (SWV) in the semitendinosus, semimembranosus, and \nbiceps femoris muscles. Data were analyzed based on  leg dominance and sex, \nwith statistical significance set at p < 0.05. \nResults or Findings: Significant differences were found in semitendinosu s \nstiffness between sexes, particularly in the non-do minant leg (p = 0.02). \nMales exhibited higher SWV in the semitendinosus mu scle compared to \nfemales, with mean SWV of 3.8 m/s (dominant) and 4. 1 m/s (non-dominant) for \nmales versus 2.9 m/s (dominant) and 2.8 m/s (non-do minant) for females. No \nsignificant SWV differences were observed in the se mimembranosus and \nbiceps femoris muscles between sexes or based on le g dominance. \nConclusion: SWE effectively identifies dominance-related differ ences in \nhamstring stiffness, offering valuable insights for  optimizing athletic \nperformance and reducing injury risks. Future resea rch should explore the \nimplications of these findings on long-term injury prevention and rehabilitation \nstrategies. \nLimitations: The sample size was small (30 players), the exact d egree of \nstiffness remains uncertain without a histological correlation serving as a \nstandard of reference for the hamstrings. Other fac tors such as muscle \ngeometry, fiber orientation, and intramuscular pres sure may influence SWV. \nFunding for this study: No funding \nEthics committee - additional information: The study was approved by the \nClinical Research Ethics Committee. CEIC 2024.232 \nAuthor Disclosures:  \nPatricia Nieto: Nothing to disclose  \nBeatriz Sierra Alvarez De Sierra: Nothing to disclo se \n \n \nUltrasound-Guided Interphalangeal Injection (US-IPI ) of Mucoid Cysts: \nTechnical Notes and Clinical Efficacy \nE. Faiella, *E. Vergantino*, D. Santucci, A. Bruno,  G. Pacella, R. F. Grasso; \nRome/IT \n(elva.vergantino@unicampus.it) \n \nPurpose or Learning Objective: Digital mucous cysts (DMCs) are common \nsoft tissue tumors affecting interphalangeal joints . Various treatment options \nexist, with surgical excision being the standard. U ltrasound-guided cortisone \ninjection into the distal interphalangeal (US-IPI) joint has been proposed as a \ntherapeutic alternative. This study aims to assess the technical success and \nclinical efficacy of US-IPI in terms of swelling re solution and pain control. \nMethods or Background: This study assessed corticosteroid infiltration for  \nmucoid cysts in interphalangeal joints. Fifty-two p atients (16 males, 36 \nfemales; median age 53, range 45-73) were treated b etween January 2020 \nand March 2023. Inclusion criteria included joint s welling, growth, and chronic \npain >3 months; infections were excluded. The ultra sound-guided procedure \nused a 26-gauge needle and Triamcinolone acetonide (Kenacort). Pain and \nswelling were evaluated via the Numeric Rating Scal e (NRS) at 2 weeks, 1 \nmonth, 3 months, and 6 months. Statistical analysis  was performed using \nSPSS (v.22) \nResults or Findings: The study assessed pain relief from ultrasound-guid ed \ncorticosteroid infiltration for mucoid cysts. NRS s cores decreased significantly \nfrom a median of 6.8 pre-procedure to 3.2 at two we eks (p < 0.01) and to 0 by \none month (p < 0.01). Eighty-three percent of patie nts responded positively; \n17% experienced persistent pain (average NRS 6.2). A secondary infiltration \nreduced their scores to 3.5 at two weeks and 0 by o ne month (p < 0.01). For \njoint swelling, 68% had a 50% reduction within one month, with complete \nresolution by three months. No recurrences were obs erved at the six-month \nfollow-up. \nConclusion: Ultrasound-guided injection for digital mucous cyst s offers \neffective pain relief and aesthetic improvement. Fu ture studies should assess \nlong-term outcomes and compare efficacy with other treatments. \nLimitations: The lack of extended follow-up beyond one year post -treatment \nlimits the assessment of long-term outcomes. \nFunding for this study: No funding. \nEthics committee - additional information: The study was conducted \naccording to the guidelines of the Declaration of H elsinki. Ethical review and \napproval were waived for this study due to its retr ospective nature. \nAuthor Disclosures:  \nRosario Francesco Grasso: Nothing to disclose \nAmalia Bruno: Nothing to disclose \nEliodoro Faiella: Nothing to disclose \nElva Vergantino: Nothing to disclose \nGiuseppina Pacella: Nothing to disclose \nDomiziana Santucci: Nothing to disclose \n \n \nClinical and Radiological Outcomes of Ultrasound Gu ided Closed- Circuit \nIrrigation of Calcific Tendinitis of the Shoulder: a prospective study \n*A. De Grandis*¹, C. D'Alessandro¹, G. Sussan¹, A. Crimì¹, D. Coraci¹,  \nS. Masiero¹, R. Ragazzi², E. Quaia¹, F. Crimì¹; ¹Pa dova/IT, ²Venice/IT \n(andrea.degrandis96@gmail.com) \n \nPurpose or Learning Objective: Ultrasound-guided percutaneous treatments \nare a recognized and effective option for calcific tendinopathy of the shoulder. \nIn this study, we enhanced the standard double-need le technique with a \nclosed-circuit irrigation system and evaluated the clinical/radiological outcomes \nof the procedure. \nMethods or Background: We prospectively enrolled 24 patients (14 females; \nmedian age 54years, IQR:50-62) with painful calcifi c tendinopathy of the \nshoulder between October 2023 and March 2024. All p atients had a \ncalcification >5 mm treated with ultrasound-guided closed-circuit irrigation, and \nthe procedure duration was recorded. Ultrasound, ra diography evaluation, and \nOSS and SPADI clinical questionnaires were administ ered before and 3 \nmonths after the procedure. \nResults or Findings: After the procedure, there was a significant reduct ion in \nthe size of the calcifications (12 mm, IQR:10-20 mm  vs. 5.5 mm, IQR:2-10 mm; \np=0.0001). The median duration of the procedure was  41 minutes (IQR:39-45 \nminutes). After the procedure, none of the patients  experienced infections \nwhile two developed bursitis. There was a significa nt improvement in the OSS \nscore 3 months after the procedure (16.5, IQR:10-23  vs. 32, IQR:36-45.5; \np<0.0001) and a significant reduction in SPADI scor es: pain (88, IQR:74-95 \nbefore vs. 13, IQR:4-24; p<0.0001), disability (72,  IQR:60-90 before vs. 8, \nIQR:4-20; p<0.0001), and total score (78, IQR:66-91  before vs. 11, IQR:4-20; \np<0.0001). The improvement of SPADI total score was  higher, although not \nsignificantly (p=0.2891), compared to a 2015 review . \nConclusion: The closed-circuit double-needle barbotage for calc ific \ntendinopathy of the shoulder is an effective treatm ent that improves both \nshoulder pain and function with a very low risk of short-term complications. \nLimitations: No limitations were identified. \nFunding for this study: No founding were received for this study. \nEthics committee - additional information: CET-ACEV code: 471n/AO/24 \nAuthor Disclosures:  \nGiovanni Sussan: Nothing to disclose \nCarlo D'Alessandro: Nothing to disclose \nDaniele Coraci: Nothing to disclose \nRoberto Ragazzi: Nothing to disclose \nFilippo Crimì: Nothing to disclose \nAlberto Crimì: Nothing to disclose \nEmilio Quaia: Nothing to disclose \nAndrea De Grandis: Nothing to disclose \nStefano Masiero: Nothing to disclose \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 34  \nWednesday \nReal-Time Elastosonography of the Achilles Tendon a nd Plantar Fascia: \nPredictive Insights for Diabetic Foot Ulcers \nV. Burulday¹, S. Ceylan Durmaz², A. Gungunes³, *A. Tezcan*¹; ¹Malatya/TR, \n²Ankara/TR, ³Kirikkale/TR \n(alperentezcan@hotmail.com) \n \nPurpose or Learning Objective: This study aims to evaluate the plantar \nfascia and Achilles tendon in patients with diabeti c foot ulcers using two \nelastosonography methods: strain elastography (SE) and shear wave \nelastography (SWE). The goal is to identify biomech anical changes in these \nstructures, which may predict the risk of diabetic foot ulcer development. This \nstudy is the first in the literature to evaluate bo th the Achilles tendon and \nplantar fascia together using both SE and SWE metho ds. \nMethods or Background: Twenty-five patients with type 2 diabetes and \ndiabetic foot ulcers, along with 30 healthy individ uals, were evaluated. B-mode \nultrasound, SE, and SWE methods were used to assess  the thickness and \nstiffness of the Achilles tendon and plantar fascia . Diabetic foot ulcers were \nstaged, and measurements were performed by a single  experienced \nradiologist. The study compared patients with diabe tic ulcers to those without, \nanalyzing changes in tissue stiffness and structure . \nResults or Findings: The study found significant increases in Achilles t endon \nthickness and stiffness in diabetic foot patients c ompared to the control group \n(p<0.0001). Plantar fascia stiffness was also signi ficantly higher in diabetic \npatients (p<0.0001). Subgroup analysis revealed tha t patients with foot ulcers \nhad even greater Achilles tendon and plantar fascia  stiffness than those \nwithout ulcers. No significant difference in planta r fascia thickness was \nobserved between groups (p=0.539), but stiffness ch anges were evident. \nConclusion: The elastosonographic evaluation of the Achilles te ndon and \nplantar fascia provides valuable insights into the biomechanical changes in \ndiabetic foot patients. Both SE and SWE are complem entary methods that may \nhelp predict the development of diabetic foot ulcer s. Early detection of stiffness \nand tissue changes through these methods can be cru cial in preventing ulcer \nformation. \nLimitations: Operator dependence, lack of MRI comparison, small sample \nsize, absence of reliability assessment \nFunding for this study: None \nEthics committee - additional information: Decision number 15/03 dated \n01.10.2018 \nAuthor Disclosures:  \nAlperen Tezcan: Nothing to disclose \nAskin Gungunes: Nothing to disclose \nVeysel Burulday: Nothing to disclose \nSenay Ceylan Durmaz: Nothing to disclose \n \n \n11:30-12:30 Room G1 \nResearch Presentation Session: \nRadiographers \nRPS 314 \nInnovative imaging practices and patient-\ncentred care: radiographers at the \nforefront of diagnostic excellence \n \nModerators \nC. Messina; Milan/IT  \nH. Precht; Middelfart/DK \n(hepr@ucl.dk) \nAuthor Disclosures:  \nCarmelo Messina: Grant Recipient: Bracco Imaging It alia, Echolight \n \n \nAnalysis of a focus group survey on person-centred care (PCC) practices \nin Nuclear Medicine: insights from the European Con gress of Radiology \n2024 \n*P. S. Costa*¹, D. Fonseca Ribeiro², M. Champendal³ , S. Murphy⁴, C. Baun⁵,  \nC. Andersson⁶, A. Karangelis⁷, K. Borg Grima⁸, A. Geão⁹; ¹Porto/PT, \n²London/UK, ³Lausanne/CH, ⁴Dublin/IE, ⁵Odense/DK, ⁶Uppsala/SE, ⁷Patra/GR, \n⁸Naxxar/MT, ⁹Lisbon/PT \n(psc@ess.ipp.pt) \n \nPurpose or Learning Objective: This study aimed to gather insights into the \nunderstanding, implementation, and challenges of Pe rson-Centered Care \n(PCC) practices among Nuclear Medicine professional s in Europe. \nAdditionally, the focus group validated the questio ns used in the survey, \nserving as the ground work for a forthcoming Europe an-wide study involving \nNuclear Medicine Radiographers. \nMethods or Background: A focus group survey was conducted during the \nEuropean Congress of Radiology 2024. The participan ts included \nRadiographers and Nuclear Medicine Radiographers/Te chnologists from \nvarious European countries. The survey covered demo graphics, professional \nbackground, understanding of PCC, its application i n clinical settings, and \nfactors aiding or hindering its implementation. The  focus group also provided \nsome informal feedback to validate the survey quest ionnaire for future use. \nResults or Findings: Thirty-two participants participated in this focus group \nand contributed to the preliminary survey, with n=5 ; 45% Radiographers and \nn=6; 55% NM Radiographers/Technologists . Most part icipants had between 1-\n10 years of experience. Participants were from Alba nia (9%), Italy (27%), \nDenmark (36%), Malta (9%), and Portugal (3%). Repor ted Key factors aiding \nPCC implementation included good communication (45% ), empathy (40%), \nand sensitivity to patient characteristics (36%). T he reported barriers to the \nimplementation of PCC included burnout (82%), compl exity of procedures \n(36%), and time constraints (27%). Additionally, th e focus group provided \nfeedback on the questions used while aiding to impr ove the survey set-up. \nConclusion: The findings revealed a strong recognition of the i mportance of \nPCC among Nuclear Medicine professionals, but also significant challenges \nsuch as burnout and time constraints which could hi nder its implementation. \nRecommendations included, amongst others, the use o f the validated \nquestionnaire to gain a broader understanding of PC C practices in Nuclear \nMedicine across a wider spectrum of European countr ies. \nLimitations: The small sample size and the limited geographical spread of the \nparticipants. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Not applicablw \nAuthor Disclosures:  \nDaniela Fonseca Ribeiro: Nothing to disclose \nKaren Borg Grima: Nothing to disclose \nApostolos Karangelis: Nothing to disclose \nAna Geão: Nothing to disclose  \nCamilla Andersson: Nothing to disclose \nChristina Baun: Nothing to disclose \nPedro Silva Costa: Nothing to disclose \nShauna Murphy: Nothing to disclose \nMélanie Champendal: Nothing to disclose \n \n \nComparison of Virtual Monoenergetic Images of a cli nical Photon-\nCounting-Detector CT with a state-of-the-art Energy -integrated-Detector \nCT in Dual-Energy Mode \n*M. Scheweder*¹, D. Ecker¹, E. Wurzinger², A. Gratz er², P. Kullnig²; ¹Linz/AT, \n²Graz/AT \n(mario.scheweder@fhgooe.ac.at) \n \nPurpose or Learning Objective: Photon-counting detector computed \ntomography (PCD-CT) is a promising novel technique for clinical CT, with new \nopportunities for image optimization while reducing  radiation dose compared to \nconventional energy-integrated detector CT (EID-CT) . Therefore, a PCD-CT \nand an EID-CT were compared to assess the image qua lity on spectral data \nand image reconstructions. The goal was to explore technical potentials for \nclinical practice. \nMethods or Background: A whole-body phantom was scanned on an EID-CT \nin Dual-Energy (DE) Mode and a PCD-CT with similar CTDIvol. Virtual \nmonoenergetic images (VMI) were processed at 16 keV  levels (40-190 keV) \nand different reconstructions. Signal-to-noise Rati o (SNR) and contrast-to-\nnoise ratio (CNR) ROIs were evaluated from liver eq uivalent tissue for each \nlevel and reconstruction. Mann-Whitney U test was u sed to compare image \nquality. Besides, a dose-reduced PCD-CT scan was co mpared to the EID-CT \nscan. \nResults or Findings: PCD-CT-VMI data show significantly higher (all p<.0 5) \nSNR and CNR (all p<.001) than EID-CT across all keV  levels and \nreconstruction methods. However, SNR and CNR highly  depended on the \nreconstruction method and keV level. PCD-CT SNR/CNR  exceeded ≥80 keV \n(SNR: +7% to +502% / CNR: +3% to +801%). A 40% dose -reduced PCD-CT \nscan provided higher dose-normalized SNR against th e EID-CT scan at 100-\n190 keV (+35% to +272%). However, at 40-90 keV, PCD -CT had lower SNR (-\n53% to -1%). \nConclusion: PCD-CT-VMI demonstrate higher SNR/CNR capabilities \ncompared to EID-CT in DE-Mode. This advantage can b e used to optimize \nscan setups regarding radiation exposure. Few studi es have compared the \nVMI data of PCD-CT and EID-CT for this purpose. How ever, in addition to \nthese promising objective results, subjective image  assessment by radiologists \nis necessary to clarify diagnostic accuracy. \nLimitations: This project is a phantom study and requires furthe r clinical \nresearch to apply findings in practice. \nFunding for this study: This project was internally funded by the Universit y of \nApplied Sciences for Health Professions Upper Austr ia. Project-Number: P-\n2002-003 \n\n \n \nAbstract-based Programme \n \n 35  \nWednesday \nEthics committee - additional information: The project was evaluated by the \ncore team of the Institutional Review Board of the University of Applied \nSciences for Health Professions Upper Austria, with  the following conclusion: \n\"There are no objections to the execution of this s tudy in its current form\". IRB-\nNumber.: A-2022-018 \nAuthor Disclosures:  \nPeter Kullnig: Nothing to disclose \nAlexander Gratzer: Nothing to disclose \nEric Wurzinger: Nothing to disclose  \nMario Scheweder: Nothing to disclose  \nDominik Ecker: Author: DE is an employee at Siemens  Healthineers. During \nthe course of this work, he was a master’s student at the University of Applied \nSciences for Health Professions Upper Austria and w orked on the project. \n \n \nRadiographer-operated urgent diagnostic imaging in hybrid mobile \nstroke unit: a pilot technical study in challenging  conditions \n*D. Pakizer*¹, A. Chalánková², R. Líčeník³; ¹Ostrava/CZ, ²Olomouc/CZ, \n³Peterborough/UK \n(davidpakizer@gmail.com) \n \nPurpose or Learning Objective: Diagnostic assessment shift and treatment \nat emergency site proved beneficial for acute strok e or other neurological \npatients by using hybrid-mobile stroke unit (h-MSU)  ambulance with mobile \ncomputed tomography (CT), portable ultrasound (US),  and telemedicine \nonboard. We aimed to determine feasibility, safety,  and efficacy of \nradiographer-operated h-MSU CT/US in advanced preho spital work-up for \npatients with acute neuroemergencies in challenging  geographical/weather \nconditions. \nMethods or Background: In our pilot prospective open-label cohort study, h -\nMSU was available constantly for 19 consecutive day s (November/December \n2023) in Czechia mountain/rural region. Patients we re examined by CT \nintracranially in standby ambulance. Extracranial c arotid US underwent \nselected stable patients with time of transport >30 min to stroke center \n(ambulance on move). The h-MSU CT/US efficacy was c ompared with \nstandard in-hospital modalities; feasibility and sa fety were assessed. \nResults or Findings: Of 46 patients, 37 brain CT, 1 intracranial CT \nangiography, and 8 extracranial carotid US examinat ions were conducted. CT \nhelped find contraindications for thrombolysis in 3  patients; 6 patients received \nthe treatment. Of 106 CT scans, only 3% of full exa minations had to be \nrepeated. Mobile-CT mean radiation dose was only sl ightly higher compared to \nin-hospital CT, findings and image quality were sim ilar. Low temperatures and \nuneven mountainous terrain were responsible for 4% of repeat CT scans. \nGood-quality US images were achieved and no hemodyn amically significant \nstenosis was found but approximately 15min extracra nial carotid examination \nwas needed. Moreover, 2 US-guided cannulations were  performed. \nConclusion: Mobile-CT and portable US onboard of h-MSU were saf e, \nfeasible, and effective modalities for neuroemergen cy patients and proved to \nbe beneficial for time reduction and faster treatme nt decision-making in \nchallenging geographical/weather conditions. \nLimitations: Older age of mobile-CT/US, CT contrast agent only f or study \nsecond-half, low patient number examined by US, and  lack of Doppler US. \nFunding for this study: None. \nEthics committee - additional information: Tomas Bata Hospital Zlin Ethics \nCommittee (approval 2023-66) \nAuthor Disclosures:  \nAnežka Chalánková: Nothing to disclose \nRadim Líčeník: Nothing to disclose \nDavid Pakizer: Nothing to disclose \n \n \nOptimisation in CT using tin filtration: A systemat ic Review \n*A. Bellizzi*, J. L. Portelli, P. Bezzina, G. Galea , F. Zarb; Msida/MT \n(bellizziandrea22@gmail.com) \n \nPurpose or Learning Objective: To identify which non-contrast CT \nexaminations benefit from image quality and radiati on dose optimisation using \ntin filtration (TF) and which optimisation strategy  is best suited for this purpose. \nMethods or Background: The review was registered in PROSPERO, and \nused PICO to create the research question, and excl usion/inclusion criteria. \nFrom the research question, MeSH search terms were obtained and inputted \ninto five electronic databases: PUBMED, Scopus, CIN AHL complete, \nCochrane Central Register of Controlled Trials and Health & Medical \nCollection. Studies identified from the search were  loaded into Covidence and \nreviewed by a team of 3 experts using PRISMA guidel ines. The Joanna Briggs \nInstitute (JBI) critical appraisal tool was used to  evaluate the studies while data \nextraction was performed using a self-designed vali dated tool. \nResults or Findings: From the retrieved 1479 studies, 410 were found to be \nduplicates leaving 1069 studies for title/abstract screening. Subsequently, 130 \nstudies were included for full text-review, with a final 84 studies included for \nevaluation. TF was used to optimise scanning in 14 protocols. Scan \nparameters used in conjunction with TF as an optimi sation strategy were: \niterative reconstruction (IR) level, tube voltage ( kV), pitch, rotation time and \nreference mAs. Use of TF achieved a significant dos e reduction ranging from \n17-95% in all protocols. \nConclusion: TF is an efficient dose reduction technique in non- contrast CT \nexaminations, but has limitations meriting consider ation in terms of objective \nimage quality. These limitations could potentially be solved by varying the IR \nlevel, however more studies are needed for this to be confirmed. \nLimitations: Six studies could not be retrieved. A meta-analysis  could not be \nconducted due to the heterogeneity of the studies. Paediatrics and CT \nprotocols using IVCM were excluded. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: None required - systematic \nreview. \nAuthor Disclosures:  \nGabriel Galea: Nothing to disclose \nJonathan Loui Portelli: Nothing to disclose \nFrancis Zarb: Nothing to disclose \nAndrea Bellizzi: Nothing to disclose \nPaul Bezzina: Nothing to disclose \n \n \nThe use of Ultra-Low Dose Computed Tomography in th e diagnosis of \nSuspected Physical Abuse in paediatric patients - A  phantom study \nE. K. Mahon, A. A. Mohammed, A. England, R. Young, *N. Moore*,  \nG. A. Curran, M. F. Mcentee; Cork/IE \n(niamh.moore@ucc.ie) \n \nPurpose or Learning Objective: Suspected physical abuse (SPA) poses a \nglobal threat to children, particularly those under  the age of two. Current \npractices utilise conventional radiography skeletal  surveys (SSs) to help with \nthe diagnosis and management of SPA. The recent exp loration of ultra-low \ndose CT (ULDCT) could improve diagnostic accuracy o f SPA whilst minimising \nradiation exposure and thus is the focus of this st udy. \nMethods or Background: CT datasets were acquired on a paediatric phantom \nusing ULDCT (DLP=1.49 mGycm2) and standard-dose (DL P=22.92 mGycm2) \nCT protocols. Participants (radiographers and radio logists) subjectively scored \nthe image quality (IQ) of both protocols using a 5- point Likert scale. Mann-\nWhitney U tests assessed significant differences in  IQ between protocols. \nParticipants also estimated radiation dose differen ces and confidence in \ndiagnosing SPA using the CT datasets provided. \nResults or Findings: Responses from 46 participants were included. Data \nwere categorised into four anatomical areas; head/n eck, thorax, \nabdomen/pelvis and extremities. IQ scores were cons istently higher for STD \nwhen compared to ULDCT (H&N 2.7 vs 2.1; Thorax 2.8 vs. 2.2; Abdo/Pelvis \n2.8 vs. 2.0 and Extremities 2.9 vs. 2.2; p<0.05). I mportantly, the ULDCT \nprotocol scored highest in “Borderline acceptable” for all regions. Interestingly, \n45 (97.8%) participants underestimated the dose dif ference (1.49 versus 22.92 \nmGycm2) between the two protocols. The ULDCT protoc ol scored a total of \n38% (88/230) confidence for the diagnosis of SPA, w hereas the STD protocol \nscored a total of 70% (161/230) confidence. \nConclusion: Despite the IQ difference between protocols, ULDCT results are \npromising. Results may suggest that with adjustment s to the ULD protocol \n'optimisation' CT could contribute further SPA diag nosis but further research, \nincluding clinical studies, are needed. \nLimitations: Phantom-based study. \nFunding for this study: None. \nEthics committee - additional information: Medical School Social Research \nEthics Committee - University College Cork \nAuthor Disclosures:  \nMark F. Mcentee: Nothing to disclose \nNiamh Moore: Nothing to disclose \nRena Young: Nothing to disclose \nAhmed Abdulahad Mohammed: Nothing to disclose \nGráinne Alison Curran: Nothing to disclose \nAndrew England: Advisory Board: RoClub Research/Gra nt Support: GE \nHealthcare, Organon Pharma Ltd. Board Member: EFRS \nEimear Kate Mahon: Nothing to disclose \n \n \nDeep-learning based image reconstruction in body CT  imaging: What is \nthe real gain? A quantitative study \n*D. Delarbre*¹, E. Maturana¹, M. Scheffler¹, J. L. Navarro Quirante¹, D. Racine²; \n¹Geneva/CH, ²Lausanne/CH \n(david.delarbre@hug.ch) \n \nPurpose or Learning Objective: Compare the performance between deep \nlearning-based Advanced intelligent Clear-IQ Engine  (AICE), iterative Adaptive \nIterative Dose Reduction (AIDR-3D) and filtered bac k projection (FBP) \ncomputed tomography (CT) image reconstruction algor ithms in terms of image \ntexture, low-contrast lesion detectability, and dos e reduction potential. \nMethods or Background: An abdominal anthropomorphic phantom was \nscanned at five computed tomography dose index (CTD Ivol) level settings: \n\n \n \nAbstract-based Programme \n \n 36  \nWednesday \n10.3, 6.4, 3.3, 2.4, and 1.9 mGy. Images were recon structed using AICE, \nkernel “Body Sharp”, 1mm slice thickness, then AIDR -3D, kernel “FC08”, 1mm \nand 2mm slice thicknesses, and classic FBP includin g quantum denoising \nsystem (QDS+) reconstruction. Noise and contrast-de pendent spatial \nresolution were assessed through noise power spectr a (NPS) and target \ntransfer functions (TTF). Texture similarity of the se algorithms was evaluated \nusing peak frequency difference (PFD) and root mean  square deviation \n(RMSD). Lesion detectability was quantified using a  non-prewhitening (NPW) \nobserver model with eye filter. The area under the curve (AUC) and receiver \noperating characteristic (ROC) served as the figure  of merit (FOM). Dose \nreduction potential for AICE 1 mm, compared to AIDR -3D with 2mm slice \nthickness, was calculated to achieve equivalent AUC  values. \nResults or Findings: At higher dose levels, AIDR-3D better preserved FBP -\nlike noise texture. At lower doses, this difference  diminished. The PFD for \nAIDR-3D ranged from -0.05 to -0.14, while for AICE it ranged from -0.13 to -\n0.17. RMSD values followed a similar trend. AICE co nsistently achieved higher \nAUC values than AIDR-3D with 1mm slice thickness, w ith an increasing \ndifference as dose decreased. AICE especially demon strated a radiation dose \nreduction potential of up to 45% compared to AIDR-3 D with 2mm slice \nthickness. \nConclusion: AICE provides equivalent low-contrast lesion detect ability at \nsignificantly reduced radiation doses compared to A IDR-3D, without adversely \naffecting noise texture. \nLimitations: Constant kV, phantom morpho-type M \nFunding for this study: 0 \nEthics committee - additional information: None \nAuthor Disclosures:  \nDamien Racine: Nothing to disclose \nJose Luis Navarro Quirante: Nothing to disclose \nDavid Delarbre: Nothing to disclose \nEnrique Maturana: Nothing to disclose  \nMax Scheffler: Nothing to disclose \n \n \nThe effect of simulated reduced temporal resolution  and motion artefacts \non CT-derived cardiac left ventricular ejection fra ction \n*M. W. Kusk*¹, O. Gerke², S. Hess², S. J. Foley³; ¹ Esbjerg/DK, ²Odense C/DK, \n³Dublin/IE \n(martin.weber.kusk@rsyd.dk) \n \nPurpose or Learning Objective: To test whether mid-range CT-scanners, \nwith low temporal resolution (TR) can reliably meas ure left ventricular ejection \nfraction, (LVEF), compared to high-end cardiac scan ners. \nMethods or Background: 77 low-dose functional CT datasets, from a 3rd \ngeneration DSCT scanner, with 66 msec (TR) , recons tructed at 5% intervals of \nthe entire cardiac cycle were used Cardiac MRI serv ed as gold standard for \nclassifying patients with potential heart failure ( LVEF below 50%). Reduced TR \ndataset were simulated by temporal averaging betwee n adjacent phases, using \na MATLAB script. Furthermore, in 25 artefact-free d atasets, we inserted \nsimulated discontinuity artefacts of varying magnit ude and location. LVEF was \nmeasured on a clinical workstation using standard ( ST) and Blood Volume \n(BV) modes. Absolute LVEF was compared between orig inal and simulated \nimages with Bland-Altman plots and t-tests, while c orrelation between effective \nTR was examined wiht Spearman rank-correlation. \nResults or Findings: For BV-mode, LVEF was not significantly different \nbetween original and reduced TR images, (p=0.88) bu t significantly lower in \nST-mode by 2.8% (p<0.01.) However, no patients were  reclassified according \nto the 50% threshold. There was significant negativ e correlation between LVEF \nand effective TR in the ST, but not in BV mode. Mot ion artefacts increased the \nwidths of 95% limits of agreement to 5.6% in ST mod e vs 2.8% in BV mode. \nConclusion: Halving effective TR did not affect LVEF in the BV- mode . This \nmode also provided lower dispersion in the presence  of motion artafacts, \nmaking it the recommended measurement method. Mid-r ange scanners \nappear suitable for LVEF measurement, e.g. in conju nction with standard \nChest-Abdomen-Pelvis scans. \nLimitations: The main limitation is the lack of a true reference  standard with \nsuperior temporal, and similar spatial resolution, thus making the effect of the \ntwo factors hard to disentangle. \nFunding for this study: The study was funded by: The Esbjerg Fund, \nResearch Fund of Danish Radiographers Association, Karola Jørgensen Fund \nEthics committee - additional information: Regional \nAuthor Disclosures:  \nMartin Weber Kusk: Nothing to disclose \nSøren Hess: Nothing to disclose \nOke Gerke: Nothing to disclose \nShane J Foley: Nothing to disclose \n \n \n \n \n \nPatient-tailored contrast optimization in coronary CT angiography on \nPhoton Counting CT \n*T. Busselot*, P. Giordano, W. Coudyzer, H. Bosmans , S. Dymarkowski; \nLeuven/BE \n(t.busselot@gmail.com) \n \nPurpose or Learning Objective: Standardized injection protocols are often \nintegrated in coronary CTA (cCTA), yet failing to a ccount for patient variability. \nThis study aims to introduce an evidence based inje ction protocol, achieving a \ntarget enhancement (expressed in terms of HU) in th e overall patient \npopulation. \nMethods or Background: In a first retrospective study, 162 cCTA patients, \nscanned on a photon-counting CT, were included. Inj ection parameters, \ndemographic data and coronary enhancement were retr ospectively collected \nand HU enhancement was measured in the aorta ascend ens and proximal \nRCA using circular ROIs. Using the principle that i odine concentrations and HU \ncorrelate linearly, ideal iodine delivery rates (IDR) that would have provided a \nHU target of 500 in 55keV mono-energetic images wer e calculated from the \nHU of the original scans. Linear regression analysi s was performed with the \ndifferent demographic parameters and their ideal ID R, to obtain an evidence \nbased IDR. Secondly, 62 patients were prospectively  recruited in an IRB-\napproved study and injected with newly proposed IDR s, wherefrom volume \nwas calculated. To obtain lowest volumes iodine-sal ine dilution ratios were \nused. A minimal total volume of 60ml and fixed inje ction duration of 17s was \ndefined to ensure complete coronary enhancement. Co ronary enhancement \nwas measured in the same ROIs as retrospectively. \nResults or Findings: Ideal IDRs correlated best with bodyweight adjusted  for \nDeurenberg’s fat formula (Pearson r=0.69), with IDR =0.0116x+0.4556 Mean \ncoronary enhancement was 530±99HU (median: 539 and IQR:462 – 599), \nagreeing with the target HU of 500. Mean injected i odine volume was 51ml \n(median:51ml and IQR:45–57ml). All low iodine volum es could be administered \nunder the same duration and injection rates (from 3 .5–3.8ml/s). \nConclusion: A new method for injection protocol optimization, b ased on linear \nregression and integrating iodine-saline dilutions,  was introduced and achieved \nthe preset target HU enhancement. \nLimitations: Mono-centric study. \nFunding for this study: AI-POD project. \nHORIZON Action Grant Budget-Based. Grant number 101 080302. \nEthics committee - additional information: UZ/KU Leuven ethics, internal \nnumber S58042 \nAuthor Disclosures:  \nWalter Coudyzer: Nothing to disclose \nHilde Bosmans: Nothing to disclose \nTim Busselot: Nothing to disclose \nPierpaolo Giordano: Nothing to disclose \nSteven Dymarkowski: Nothing to disclose \n \n \n13:00-14:30 Research Stage 1 \nResearch Presentation Session: Breast \nRPS 402 \nAdvances in breast imaging: innovations \nshaping the future of cancer care \n \nModerator \nR. Woitek; Vienna/AT  \n(ramona.woitek@dp-uni.ac.at) \n \n \n18F-fluoroestradiol hybrid imaging in clinical mana gement of breast \ncarcinoma \n*J. Ferda*, E. Ferdova, T. Barakova, S. Vokurka; Pl zen/CZ \n(ferda@fnplzen.cz) \n \nPurpose or Learning Objective: 18F-fluoroestradiol is a novel \nradiopharmaceutical useful in breast carcinoma, the  indications in clinical \nscenarios are under development. The purpose of the  study is to assess the \nclinical impact of the imaging of the breast carcin oma with estrogen-positive \nreceptors (ER+) using 18F-fluoroestradiol (18F-FES)  PET/CT or PET/MRI \naccording to its impact to the treatment decision m aking. The study is \nconcerned in the different preference of PET/CT and  PET/MRI in the staging \nand restaging. \nMethods or Background: 40 patients with estrogen positive breast carcinoma  \nunderwent the hybrid imaging after intravenous appl ication of 18F-FES, in 25 \ncases it was used PET/CT, in 15 cases PE/MRI. The r adiopharmaceutical was \n\n \n \nAbstract-based Programme \n \n 37  \nWednesday \ninjected in the activity of 2,5 MBA/kg and the imag ing was performed the \nimaging. In 10 patient PET/MRI was used as restagin g method, PET/MRI was \nperformed in the 5 cases of the staging before surg ery with targeted full \ndiagnostic MRI imaging of the breast in prone posit ion, followed by the trunk \nimaging in supine position, in five to seven steps.  All PET/MRI were performed \nwith the gadolinium contrast material, the imaging included brain imaging in T1 \nSTARVIBE. PET/CT was performed using the continuous  PET acquisition after \nCT with the administration of the iodinated contras t material, in 5 cases was \nperformed in staging, in 20 cases in rest-aging \nResults or Findings: The most important information was detection of ER+  \nmetastases when 18F-FDG-PET was negative (12x) - in cluding brain and liver \nmetastases, the persistent ER+ of the metastases (7 x), staging of the disease \n(10x), the loss of the ER (4x) and the negative fin ding for metastases (2), no \nadded information was found in 5 examinations. \nConclusion: 18F-FES-PET provided the important clinical informa tion to \ntreatment strategy, PET/MRI improves the imaging of  brain and liver. \nLimitations: Small sample \nFunding for this study: No \nEthics committee - additional information: according to the Helsinky \ndeclaration \nAuthor Disclosures:  \nEva Ferdova: Nothing to disclose \nTana Barakova: Nothing to disclose \nSamuel Vokurka: Nothing to disclose \nJirí Ferda: Nothing to disclose \n \n \nConspicuity as new CEDM descriptor: likelihood of M alignancy and \nRelationship With Breast Tumor Receptor Status \nL. Corradini, *D. Ballerini*, A. Bonanomi, E. D'Asc oli, G. Della Pepa,  \nC. De Berardinis, E. Ancona, C. Depretto, G. P. Sca perrotta; Milan/IT \n(Daniela.Ballerini@istitutotumori.mi.it) \n \nPurpose or Learning Objective: Lesion conspicuity, defined as the \"degree of \nenhancement\" relative to the background, was the fo cus of this retrospective \nmonocentric study, which aimed to explore its corre lation with malignancy of \nlesions, histology, receptor profile and grading in  breast cancer patients. \nMethods or Background: Two breast radiologist and one radiology resident \nevaluated all CEDM performed in our oncological cen ter from January 2023 to \nApril 2024, assigning degrees of conspicuity to bre ast lesions, and evaluating a \npossible correlation with Ki-67 values (≤ 20% or > 20%), HER-2 status, \nestrogen (ER) and progesterone (PGR) receptor posit ivity, molecular subtype, \nand histological grade. Statistical analysis employ ed the Cramer’s V test. \nResults or Findings: Out of 352 patients included (median age=54, IQR=18 ), \n100 were excluded due to chemotherapy controls and 53 due to B3 lesion. In \nthe 199 remaining patients we observed a moderate t o strong association \nbetween conspicuity and ER expression (V=0.534) and  Ki-67 value (V=0.36). A \nmoderate association was found between conspicuity and PGR expression \n(V=0.31). No significant correlation was noted betw een conspicuity and \nhistological grade (V=0.184) or HER2 status (V=0.2) . \nConclusion: Conspicuity, a recently incorporated descriptor in CEDM BI-\nRADS lexicon, was validated by our findings, which are in line with the initial \nevidence in the literature. \nLimitations: Retrospective monocentric study with limited number  of patients. \nConspicuity is a subjective descriptor, potentially  introducing variability in the \ndata and affecting the findings. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nGianmarco Della Pepa: Nothing to disclose \nClaudia De Berardinis: Nothing to disclose \nElisa D'Ascoli: Nothing to disclose \nAlice Bonanomi: Nothing to disclose \nDaniela Ballerini: Nothing to disclose \nEleonora Ancona: Nothing to disclose \nGianfranco Paride Scaperrotta: Nothing to disclose \nCatherine Depretto: Nothing to disclose \nLisa Corradini: Nothing to disclose \n \n \nA novel metabolic MRI method for malignant breast t umors detection \nM. Rivlin¹, *R. Sivan Hoffmann*², V. Hadar², S. Suk hotnik², N. E. Weisenberg², \nO. Shmain-Naydenov², M. Zaiss³, S. Weinmüller³, G. Navon¹; ¹Tel Aviv/IL,  \n²Kfar Saba/IL, ³Erlangen/DE \n(rotemsivan3@gmail.com) \n \nPurpose or Learning Objective: Molecular imaging with 18F-\nfluorodeoxyglucose positron emission tomography (18 FDG-PET) is a powerful \nand well-established tool in breast cancer manageme nt, as increased glucose \nuptake is a known cancer hallmark.However, it carri es the risk of repeated  \n \nradiation exposure. We have recently discovered tha t glucosamine (GlcN), a \nnon-toxic, biocompatible glucose-based material can  be detected using a \nunique MRI contrast mechanism termed chemical excha nge saturation transfer \n(CEST). CEST has emerged as an attractive molecular  imaging approach for \nproviding valuable metabolic information. Our goal is to develop an innovative \nmolecular imaging modality based on CEST-MRI of Glc N to visualize and \nmeasure breast tumors while also distinguishing bet ween benign and \nmalignant tumors. \nMethods or Background: Breast cancer patients and control group were \nscanned using the CEST-MRI pulse sequence on a 3T s canner (VIDA, \nSiemens, Germany) equipped with breast coil. The pr otocol included CEST \nscans before and two hours after drinking a solutio n of GlcN (184 mg/kg). The \ndata were evaluated using magnetization transfer as ymmetry ratio (MTRasym) \nand area under curve (AUC) analysis. \nResults or Findings: GlcN treatment resulted in higher CEST values in tu mor \nregions of interest (ROIs), with maximum net MTRasy m signal (at 2 ppm) of \n6.3±2.6% and averaged AUC (2-5 ppm) increase ratio of 3.3±2.1% (N=7). Yet, \nno significant GlcN CEST signal enhancement was det ected in healthy \nvolunteers (N=9). GlcN CEST signal values were high ly correlated with the BI-\nRADS category. \nConclusion: These findings suggest that the GlcN CEST MRI techn ique can \ndetect breast cancer while also providing molecular -level diagnostic tools for \ndiscriminating between benign and malignant breast tumors. These promissing \nresults pave the way to a future metabolic imaging of additional diseases. \nLimitations: Due to the size of the study group we were not able  to clarify into \nsub-groups. \nFunding for this study: Funding by the ISF No. 1689/18 \nEthics committee - additional information: approval number MMC0201-21 \nAuthor Disclosures:  \nOlga Shmain-Naydenov: Nothing to disclose \nNoemi Edith Weisenberg: Nothing to disclose \nMoritz Zaiss: Nothing to disclose \nRotem Sivan Hoffmann: Nothing to disclose \nStephanie Sukhotnik: Nothing to disclose \nSimon Weinmüller: Nothing to disclose \nVivian Hadar: Nothing to disclose \nMichal Rivlin: Nothing to disclose \nGil Navon: Nothing to disclose \n \n \nArtificial intelligence in digital mammography and serial breast \ntomosynthesis for neoadjuvant breast cancer treatme nt response \nprediction \n*D. Förnvik*¹, S. Zackrisson¹, I. Skarping²; ¹Malmö /SE, ²Lund/SE \n(daniel.fornvik@med.lu.se) \n \nPurpose or Learning Objective: To predict neoadjuvant chemotherapy \n(NACT) treatment response by applying artificial in telligence (AI) to digital \nmammography (DM) and serial breast tomosynthesis (D BT) images. \nExplainable AI (XAI) for enhanced clinical credibil ity is explored. \nMethods or Background: NACT for early-stage breast cancer (BC) has \nrecently become an attractive approach to patients who are eligible for \nchemotherapy. MRI is the imaging modality of choice  for evaluating tumor \nresponse but not as readily available as DM and lat ely DBT. Nevertheless, \npredicting residual cancer, as assessed by radiolog ist, after NACT using \nimaging is challenging; thus, AI could be an altern ative. Between 2005/2014 - \n2019, 453 (DM) and 149 (DBT) patients, respectively , at Skane University \nHospital, Sweden, comprised the cohorts. Two deep l earning architectures \n(DM: ResNet24, DBT: backbone 3D ResNet) applied to images from both the \ncancer and contralateral healthy breasts acquired a t three time points: pre-\nNACT (DM and DBT), mid-NACT (DBT) and post-NACT (DB T) were used to \npredict pathological complete response (pCR). For D BT, GradCAM was used \nto produce saliency maps to obtain insights into th e model-based decisions. \nResults or Findings: The DM and the DBT AI models predicted pCR as \nrepresented by the area under the ROC curve of 0.71  (95% CI: 0.53–0.90; \np = 0.035) and 0.83 (95% CI: 0.63–1.00; p = 0.008), res pectively. The spatial \ncorrelation of saliency maps for DBT volumes from t he same patient but at \ndifferent time points was high, likely indicating t hat the model focuses on the \nsame areas during decision-making. \nConclusion: The DBT model demonstrates a high discriminative pe rformance \nfor predicting pCR/non-pCR, possibly outperforming radiologists' assessment. \nLimitations: Availability of larger datasets and inclusion of cl inicopathological \nvariables would permit more comprehensive training of the models and more \nrigorous evaluation of their prediction performance  for future patients. \nFunding for this study: Swedish Breast Cancer Group (BRO), Allmänna \nSjukhusets i Malmö Stiftelse för bekämpande av canc er, and the \nGovernmental Funding of Clinical Research within th e National Health \nServices. \nEthics committee - additional information: Regional Ethics Committee in \nLund, Sweden (committee reference numbers: 2014/13,  2014/521, and \n2016/521, and 2021-05637-02). \n \n\n \n \nAbstract-based Programme \n \n 38  \nWednesday \nAuthor Disclosures:  \nDaniel Förnvik: Nothing to disclose \nIda Skarping: Nothing to disclose \nSophia Zackrisson: Nothing to disclose \n \n \nImage Quality and Diagnostic Values of Diffusion-We ighted Breast MRI: \nA Comparison of Single-Shot EPI with Deep Learning Reconstruction and \nMulti-Shot EPI with Simultaneous Multislice \n*H. S. Ahn*, S. H. Kim, M. J. Hong, H-S. Lee; Seoul /KR \n \nPurpose or Learning Objective: To evaluate the image quality and diagnostic \nvalue of single-shot echo-planar imaging (ss-EPI) w ith deep learning \nreconstruction (DLR) versus simultaneous multi-slic e echo-planar imaging \n(SMS rs-EPI) in breast MRI. \nMethods or Background: This study included 77 cases of breast cancer from \n74 patients who underwent preoperative breast MRI. All patients underwent \nbreast MRI that included the ss-EPI sequence combin ed with post-processing \nusing DLR, as well as the SMS rs-EPI sequence. Two radiologists \nindependently assessed qualitative image quality fa ctors and determined their \npreferences, while the cancer detection rate (CDR) was calculated. \nQuantitative analysis included measurements of appa rent diffusion coefficient \n(ADC), signal-to-noise ratio (SNR), contrast-to-noi se ratio (CNR), and lesion \ncontrast, including phantom measurements. \nResults or Findings: Regarding qualitative image quality parameters, ss- EPI \nwith DLR demonstrated significantly higher scores i n fat suppression and \noverall image quality as assessed by both radiologi sts, with comparable scores \nin artifact presence and lesion conspicuity to the SMS rs-EPI sequence. The \nCDR showed no significant difference between the tw o sequences. Both \nradiologists preferred ss-EPI with DLR (Reader 1: 7 8.4%, Reader 2: 89.2%). \nFor quantitative parameters, ss-EPI with DLR exhibi ted significantly higher \nCNR (p = 0.002) and lesion contrast (p < 0.001), wh ile ADC and SNR values \nwere comparable. In phantom measurements, mean ADC was lower for ss-EPI \nwith DLR (DLR: 1.08 ± 0.58 vs. SMS: 1.12 ± 0.59, p = 0.007), but SNR was not \nsignificantly different (DLR: 607.45 ± 346.1 and SMS: 630.03 ± 624.51, p = \n0.911). The acquisition time was shorter for ss-EPI  with DLR (2:06 min) \ncompared to SMS rs-EPI (3:29 min). \nConclusion: ss-EPI with DLR provided superior image quality and  greater \nreader preference compared to SMS rs-EPI. \nLimitations: This is a retrospective study which performed at si ngle center. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nHyun-Soo Lee: Nothing to disclose \nMin Ji Hong: Nothing to disclose \nSung Hun Kim: Nothing to disclose \nHye Shin Ahn: Nothing to disclose \n \n \nDual Imaging Power: CT and Contrast-Enhanced Mammog raphy (CEM) \nfor Advanced Detection of Metastatic Breast Cancer \n*M. Balbino*¹, M. Montatore², F. Masino³, F. A. Car pagnano⁴, G. Capuano²,  \nG. Guglielmi⁵; ¹Triggiano/IT, ²Barletta/IT, ³Bari/IT, ⁴Foggia/IT, ⁵Andria/IT \n(marinabalbino93@gmail.com) \n \nPurpose or Learning Objective: This study, one of the first in Italy, aims to \nevaluate the efficacy of performing CT and CEM cons ecutively using the same \ncontrast medium in a single imaging session. The go als include reducing the \namount of contrast agent injected into oncological patients and enhancing the \ndetection of metastases in various districts throug h CT, while also providing a \nprecise diagnosis of breast extension and identifyi ng additional foci within the \nbreasts through CEM. \nMethods or Background: A cohort of female patients with confirmed primary \nbreast cancer and suspected metastatic disease were  enrolled. Each patient \nunderwent a CT scan followed immediately by a CEM u sing the same contrast \nmedium. The CT was performed to identify visceral m etastases, while the CEM \ntargeted the detection of additional breast lesions  and regional lymph node \ninvolvement. Both imaging modalities utilized iodin e-based contrast agents, \nadministered intravenously. The diagnostic outcomes  were compared with \nthose from conventional imaging techniques, includi ng standard \nmammography, ultrasound, and MRI. \nResults or Findings: The combined CT and CEM approach demonstrated a \nhigher sensitivity and specificity in detecting met astatic sites compared to \ntraditional imaging methods. In particular, CEM rev ealed additional lesions in \nthe breast and regional lymph nodes that were not i dentified by CT alone. The \nconcurrent use of the same contrast medium was foun d to be safe and well-\ntolerated, with no significant increase in adverse reactions. The integrated \nimaging protocol provided comprehensive anatomical and functional \ninformation, leading to more accurate staging and b etter-informed treatment \ndecisions. \n \nConclusion: The integration of CT and CEM using the same contra st agent \noffers a promising advancement in the diagnostic im aging of metastatic breast \ncancer. This combined approach enhances the detecti on of metastatic lesions, \nproviding a more comprehensive assessment of diseas e spread. \nLimitations: No \nFunding for this study: No \nEthics committee - additional information: No \nAuthor Disclosures:  \nMarina Balbino: Nothing to disclose \nGiuseppe Guglielmi: Nothing to disclose \nFrancesca Anna Carpagnano: Nothing to disclose \nManuela Montatore: Nothing to disclose \nFederica Masino: Nothing to disclose \nGiulia Capuano: Nothing to disclose \n \n \nNovel and robust approach to breast density predict ion: utilising the \nTree-Structured Parzen Estimator algorithm-driven t ransfer learning \napproach \n*M. Bobowicz*, M. Kosno, K. P. Brzozowski, M. Rygus ik; Gdańsk/PL \n(maciej.bobowicz@gumed.edu.pl) \n \nPurpose or Learning Objective: The breast density visual assessment in \nmammography is a subjective process prone to errors  but impacting diagnostic \ndecisions. To overcome this problem, we developed a  robust and reliable AI \nmodel that employs convolutional neural network-bas ed transfer learning, \nspecifically ResNet, DenseNet, and EfficientNet arc hitectures, to predict breast \ndensity. Our research benefits from the Tree-struct ured Parzen Estimator \n(TPE) algorithm, an advanced tool for hyperparamete r optimisation. \nMethods or Background: A dataset of 2101 digital MLO mammography \nimages performed at the Medical University of Gdans k from 2014 to 2022 was \nselected for analysis. The images were acquired usi ng various devices from \nSIEMENS, GE HEALTHCARE, and HOLOGIC to ensure a hig h degree of \nimage characteristics variability. The dataset was divided into 80% training and \n20% validation sets. ResNet, DenseNet, and Efficien tNet architectures were \ntrained using the TPE algorithm. The assembly model  comprises three five-fold \ncross-validated convolutional networks. \nResults or Findings: An ensemble model resulted in good performance \nmetrics: AUC-ROC (0.99), accuracy (0.91), F1-score (0.91), and recall values \n(0.90) for the test dataset. The TPE algorithm faci litates the development of \nhigh-performance models on a relatively small datas et, eliminating the need for \nimage segmentation to extract the skin and pectoral  muscle opacities, which is \nchallenging to implement and often burdened with si gnificant errors. \nConclusion: Our methodology enables the straightforward trainin g of a robust \nmodel that can provide highly precise breast densit y predictions, reducing and \nautomating the burden of required density reporting . Furthermore, our findings \ndemonstrate the efficacy of advanced hyperparameter  numerical optimisation \nmethods in enhancing the efficiency of transfer dee p learning models in the \ncontext of breast density prediction. \nLimitations: The study's limitations are its relatively small da taset, single-\ncentre design, and lack of external validation. \nFunding for this study: Funding was provided by the European Union’s \nHorizon 2020 research and innovation programme unde r grant agreement No \n952103 (EuCanImage project) and was co-funded by th e Digital Europe \nprogramme under grant agreement No 101100633 (EUCAI M project). \nEthics committee - additional information: This retrospective study uses \nfully anonymised data from the EuCanImage project u nder the global ethics \ncommittee agreement for MUG. \nAuthor Disclosures:  \nMaciej Bobowicz: Nothing to disclose \nMichał Kosno: Nothing to disclose \nMarlena Rygusik: Nothing to disclose \nKrystian Paweł Brzozowski: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 39  \nWednesday \n13:00-14:30 Research Stage 2 \nResearch Presentation Session: \nGenitourinary \nRPS 407 \nWhat's new in prostate imaging: advances \nand emerging techniques \n \nModerator \nP. A. Bonaffini; Bergamo/IT  \n(pa.bonaffini@gmail.com) \n \n \nAdvancing prostate cancer imaging: a comparative an alysis of MET-\nRADS-P and PCWG3 in the assessment of mCRPC \n*L. Russo*¹, S. Bottazzi¹, O. Longoria², G. Avesani ¹, S. J. Withey², L. D'Erme¹, \nE. Sala¹, D-M. Koh², N. Tunariu²; ¹Rome/IT, ²London /UK \n \nPurpose or Learning Objective: Treatment response assessment in \nmetastatic castration-resistant prostate cancer (mC RPC) is critical because the \nProstate Cancer Working Group 3 (PCWG3) criteria ha ve notable limitations. \nThe METastasis Reporting and Data System for Prosta te Cancer (MET-RADS-\nP) provides standardised guidelines using whole-bod y MRI (WBMRI). Our main \naim was to compare MET-RADS-P and PCWG3 criteria fo r disease \nprogression categorization in mCRPC, as well as the  prognostic value of MET-\nRADS-P for progression-free survival (PFS) and over all survival (OS). \nMethods or Background: A cohort of 201 mCRPC patients treated at The \nRoyal Marsden Hospital between January 2013 and Feb ruary 2024 was \nretrospectively included. All patients underwent WB MRI, CT and BS at each \ntime point. CT and BS were interpreted according to  PCWG3 and WBMRI \naccording to MET-RADS-P. Concordance between MET-RA DS-P and PCWG3 \nin disease progression categorization was assessed overall, in bone and soft-\ntissue only. PFS and OS were evaluated using Kaplan -Meier survival curves \nwith log-rank test comparisons. \nResults or Findings: Overall, 64.5% of time points (302/468) were concor dant \nbetween MET-RADS-P and PCWG3 criteria, with MET-RAD S-P detecting \nprogression earlier in 31.8% (149/468). Discrepanci es were more pronounced \nin bone metastases, where MET-RADS-P identified pro gression in 55.1% of \ncases classified as non-progressive disease by PCWG 3. Progressing patients \nby MET-RADS-P had significantly worse PFS: median 2 .7 months versus 4.2 \nmonths by PCWG3 (p<0.001). The median OS was 12.5 m onths for \nprogressing patients by MET-RADS-P at 12-week asses sment compared to \n19.8 months for those stable or responding (p<0.001 ). \nConclusion: MET-RADS-P allowed for earlier progression detectio n compared \nwith PCWG3, particularly in bone metastases, potent ially permitting earlier \ntherapeutic interventions. MET-RADS-P also demonstr ated strong predictive \nvalue for PFS and OS, suggesting its potential role  as an imaging biomarker in \nfuture clinical trials. \nLimitations: Retrospective design and lack of cost and availabil ity comparative \nanalysis. \nFunding for this study: This study represents independent research funded \nby the National Institute for Health and Care Resea rch (NIHR) Biomedical \nResearch Centre at The Royal Marsden NHS Foundation  Trust and The \nInstitute of Cancer Research, London, and by the Ro yal Marsden Cancer \nCharity, and Cancer Research UK (CRUK) National Can cer Imaging Trials \nAccelerator (NCITA) and Prostate Cancer UK. The vie ws expressed are those \nof the author(s) and not necessarily those of the N IHR or the Department of \nHealth and Social Care. This work uses data provide d by patients and \ncollected by the NHS as part of their care and supp ort. \nEthics committee - additional information: The study was approved by the \nInstitutional Ethics Committee (reference no. 21/LO /0605). \nAuthor Disclosures:  \nSamuel Joseph Withey: Nothing to disclose \nGiacomo Avesani: Nothing to disclose \nDow-Mu Koh: Nothing to disclose \nNina Tunariu: Nothing to disclose \nOssian Longoria: Nothing to disclose \nSilvia Bottazzi: Nothing to disclose \nEvis Sala: Nothing to disclose \nLuca D'Erme: Nothing to disclose \nLuca Russo: Nothing to disclose \n \n \n \n \nAnalysis of biopsy strategy in young men with suspi cious PSA in a \nprostate cancer screening setting – data from the P ROBASE trial \n*M. Boschheidgen*¹, R. Al-Monajjed¹, J. P. Radtke¹,  H-P. Schlemmer²,  \nG. Antoch¹, L. Schimmöller¹, P. Albers¹; ¹Düsseldor f/DE, ²Heidelberg/DE \n(matthias.boschheidgen@med.uni-duesseldorf.de) \n \nPurpose or Learning Objective: To analyze the performance of targeted (TB) \nand systematic (SB) MRI/US fusion-guided prostate b iopsy within the \nprospective PROBASE prostate cancer (PC) screening trial. \nMethods or Background: Men aged forty-five from the general population \nwere invited to screening. Those with confirmed pro state-specific antigen \n(PSA) levels of 3 ng/ml or higher were offered an M RI and were referred to \nMRI/US-guided biopsy. Biopsies were performed in ev ery participant \nunrespective of the MRI result. Targeted and system atic biopsies with \nsoftware-based fusion techniques were offered. The primary endpoint of this \nanalysis was the PC detection in either TB or SB. \nResults or Findings: A total of 554 men (median age 50 (range 44-54), \nmedian PSA level 4.1 ng/ml) were analyzed who under went an MRI followed \nby MRI/US-guided biopsy. Of 217 PC diagnosed, 198 ( 91%) and 140 (65%) \nwere detected by SB and TB, respectively. 64 of 217  PC (29%) were low grade \n(ISUP 1). 40 significant tumors were found exclusiv ely by SB (26%), while 9 \nsignificant tumors were only diagnosed with TB (6%) . SB detected significantly \nmore low-grade cancer compared to TB (p<0.001). Can cer detection rate was \n20% for PIRADS 1-2, 26% for PIRADS 3, 59% for PIRAD S 4, and 92% for \nPIRADS 5. \nConclusion: In young men and in the setting presented here, sys tematic \nbiopsy in addition to targeted MRI/US guided fusion  biopsy still appears to be \njustified to adequately detect PC. Most PC (71%) we re clinically significant. \nPerforming only TB in young men without SB faces th e risk of missing a \nsignificant number of csPC even if it simultaneousl y diagnoses fewer low-grade \ncarcinomas. \nLimitations: MRI did not influence clinical decision-making; exp erience in MRI \nreading and the image quality differed widely at th e time of initiation \nFunding for this study: Deutsche Krebshilfe \nEthics committee - additional information: The study was approved by local \nethics committee. \nAuthor Disclosures:  \nJan Philipp Radtke: Nothing to disclose \nPeter Albers: Nothing to disclose \nRouvier Al-Monajjed: Nothing to disclose \nMatthias Boschheidgen: Nothing to disclose \nLars Schimmöller: Nothing to disclose \nHeinz-Peter Schlemmer: Nothing to disclose \nGerald Antoch: Nothing to disclose \n \n \nIs there an MRI phenotype for the cribiform pattern  of Prostate Cancer? \nM. D. M. Palma, D. Freire Maia Vieira, T. A. Leite De Lima, J. Nather,  \nF. Chahud, R. B. Reis, *V. F. Muglia*; Ribeirao Pre to/BR \n(fmuglia@fmrp.usp.br) \n \nPurpose or Learning Objective: Cribriform pattern (CP) is a distinct \nhistological feature present in various neoplasms, characterized by cohesive \ntumour cells surrounding circular spaces, creating a \"Swiss cheese\" \nappearance. In prostate cancer (PCa), CP is one of the four architectural \nsubtypes of the Gleason 4 pattern and has been link ed to worse outcomes \ncompared to other morphologies. This study aimed to  determine if CP presents \ndistinct features on multiparametric magnetic reson ance imaging (MRI). \nMethods or Background: In this retrospective, single-centre study, we \nidentified PCa cases with CP from 2016 to 2023, wit h MRI conducted within \nfour months of histological diagnosis. Patients wit hout CP but with equivalent \nGleason grades and risk stratification were include d as controls in a ratio of up \nto 1.5:1. Two radiologists, with over 5 and the oth er with 7 years of prostate \nimaging experience, evaluated lesion size, form, lo cation, prostate volume, \nmean apparent diffusion coefficient (ADC) values, a nd post-contrast kinetic \ncurves. Clinical staging, prostate-specific antigen  (PSA), and PSA density \n(dPSA) were also reviewed. \nResults or Findings: The study included 42 patients with CP and 72 witho ut \nCP. No significant differences were found between t he groups regarding PSA \n(p=0.43), dPSA (p=0.37), lesion size (p=0.33), loca tion (p=0.65), mean ADC \n(p=0.21), or kinetic curve type (p=0.75). Significa nt differences were observed \nfor age (p<0.0001), prostate volume (p=0.001), and PI-RADS category \n(p=0.05). In univariate logistic regression, age an d PI-RADS score were \nindependent predictors of CP presence, but only age  remained significant in \nmultivariate analysis (p=0.0001). \nConclusion: Cribrifrom pattern in PCa is more common in older m en with \nlarger prostates and higher PI-RADS scores. However , no specific \nmorphological or functional MRI parameters were ass ociated with the presence \nof this pattern. \nLimitations: Single-centre, retrospective study. \nFunding for this study: None \n\n \n \nAbstract-based Programme \n \n 40  \nWednesday \nEthics committee - additional information: Our Institutional Review Board \napproved the research with a waiver for informed co nsent due to the \nretrospective nature. \nAuthor Disclosures:  \nFernando Chahud: Nothing to disclose \nDavid Freire Maia Vieira: Nothing to disclose \nJulio Nather: Nothing to disclose \nMatheus De Moraes Palma: Nothing to disclose \nValdair Francisco Muglia: Nothing to disclose \nRodolfo B. Reis: Nothing to disclose \nThalyne Aparecida Leite De Lima: Nothing to disclos e \n \n \nThe effect of prostate volume - does PSA density al ways work? \n*S. Durmaz*¹, S-C. J. Wu², K-L. Lee², A. Shakur², I . Caglič², T. Barrett²; \n¹Istanbul/TR, ²Cambridge/UK \n(drselahattindurmaz@gmail.com) \n \nPurpose or Learning Objective: To evaluate the impact of PSA density \n(PSAd) on the probability of detecting clinically s ignificant prostate cancer \n(csPCa) across different prostate volume ranges and  PI-RADS scores. \nMethods or Background: 2097 patients undergoing multiparametric MRI \n(mpMRI) for suspected PCa were included. 738/2097 ( 35.2%) had PCa, and \n566/2097 (27%) had csPCa (Gleason ≥3+4), patients were classified as \nnegative after biopsy (n=299) or were diagnosed wit h clinically insignificant \nPCa (n=172) or having a negative mpMRI and completi ng at least one-year \nfollow-up without developing PCa (n=1060). Single-v ariable logistic regression \nanalyses were conducted to assess the impact of PSA d on the probability of \ncsPCa within the different prostate volume ranges ( <40 mL, 40-60 mL, 60-80 \nmL, >80 mL) and PI-RADS groups. \nResults or Findings: The median age, PSA, PSAd, and prostate volume was \n66 years (IQR: 61-72), 5.6 ng/mL (IQR:4.05-8.05), 0 .10 ng/mL/mL (IQR:0.07-\nO.15), and 56 mL (IQR:39-80), respectively. Logisti c regression at a PSAd of \n0.15 ng/mL/mL showed the probability of csPCa decre ased with increasing \nprostate volume: <40 mL (44%), 40-60 mL (38%), 60-8 0 mL (29%), and >80 \nmL (18%). At the same PSAd level, the probability o f csPCa increased with \nincreasing PI-RADS score: PIRADS 1-2 (3%), PI-RADS 3 (26%), PI-RADS 4 \n(63%), PI-RADS 5 (70%). Regardless of the PSAd leve l, the risk of csPCa in \npatients with PI-RADS 4-5 lesions was always >20%. \nConclusion: When using PSAd to assist in the decision to perfor m MRI, or to \nbiopsy patients with PI-RADS 1-3 scores, caution sh ould be exercised in those \nwith larger volume prostates, as the lower PSAd can  provide false \nreassurance. \nLimitations: Retrospective design. All patients underwent mpMRI and prostate \nbiopsy at a single tertiary referral center with ex tensive experience in prostate \nMRI and biopsy. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This is a retrospective study. \nAuthor Disclosures:  \nIztok Caglič: Nothing to disclose \nAmreen Shakur: Nothing to disclose \nShun-Chin Jim Wu: Nothing to disclose \nSelahattin Durmaz: Nothing to disclose \nKang-Lung Lee: Nothing to disclose \nTristan Barrett: Nothing to disclose \n \n \nA transformer-based deep learning model for early p rediction of \nbiochemical recurrence after radical prostatectomy using pretreatment \nmpMRI \n*F. Li*¹, L. Zhuo¹, L. Yue¹, L. Juan¹, L. Wang¹, R.  Liu¹, F. Wang², Y. Xiang³; \n¹Mianyang City/CN, ²Luzhou/CN, ³Leshan/CN \n(317064491@qq.com) \n \nPurpose or Learning Objective: The purpose of this study is to develop and \nverify a deep learning model using preoperative mul ti-parameter MRI images \nto predict BCR risk after radical prostatectomy. \nMethods or Background: Patients after radical surgery at 4 centers between  \nAugust 2013 and September 2021 were retrospectively  included with the \nendpoint outcome of 3-year BCR (two consecutive spe cific antigen [PSA] \nlevels > 0.2 ng/mL [0.2µg/L]). A transformer-based DL model was used to \npredict BCR after radical surgery using 3D tumor im ages, a clinical model was \nconstructed by multivariate logistic regression, Ka plan-Meier plots were used \nfor estimating recurrence-free survival, and finall y, pre- and post-surgical Capra \nmodels, a clinical model, a multi-instance model, a nd a transformer model, \nMultimodal Combine model were compared to assess th e performance of \npredicting BCR. \n \n \n \n \nResults or Findings: A total of 582 patients (median age 70 years, (IQR 44-89 \nyears) with a median follow-up of 43 months (IQR, 2 9-71 months) were \nrandomized into a training group (n=249 ), an inter nal test set (n=107), an \nexternal test set 1 (n=189), and an external test s et 2 (n=37).The AUC of the \nTransformer model in the 0.92 in the internal test set, 0.84 in the external test \nset 1, and 0.82 in the external test set 2, and the  multimodal Combine model \nfurther improves the performance, respectively, wit h 0.94 (95% CI. 0.885 - \n0.992), 0.94 (95% CI, 0.900 - 0.969), and 0.83 (95%  CI, 0.693 - 0.965), and \nearly recurrence-free survival and overall survival  could be better risk-stratified \nand predicted using the Combine model. \nConclusion: A transformer-based DL model for predicting BCR aft er radical \nsurgery was developed and internally and externally  validated, and the joint \nmodel is better and expected to guide individualize d treatment. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: Ethics （2024）014-1 \nAuthor Disclosures:  \nLu Wang: Nothing to disclose \nRuishan Liu: Nothing to disclose  \nFan Li: Nothing to disclose  \nFei Wang: Nothing to disclose  \nLiao Juan: Nothing to disclose  \nYe Xiang: Nothing to disclose \nLv Yue: Nothing to disclose  \nLihua Zhuo: Nothing to disclose \n \n \nShort MRI sequence suitable for re-identification o f prostate lesion during \nin-bore biopsy? \n*C. Peter*, A. Schaudinn, C. Ehrengut, T. Franz, L- C. Horn, N. Linder,  \nJ-U. Stolzenburg, H. Busse, T. Denecke; Leipzig/DE \n(christian.peter1@web.de) \n \nPurpose or Learning Objective: To evaluate the image quality of a rapid \nintraprocedural balanced steady-state free precessi on (b-SSFP) sequence for \nre-identification of prostate lesions during transr ectal in-bore biopsies in \ncomparison with that of a T2-weighted \nMethods or Background: In this retrospective study, 127 patients with 140 PI-\nRADS ≥ 3 (version 2.1) lesions based on multiparametric 3 T MRI (mpMRI) \nunderwent transrectal in-bore biopsies. b-SSFP imag es were acquired at 1.5T \nfor interventional guidance. Two radiologists (R1: 11 years and R2: 2 years of \nmpMRI experience) independently rated the image qua lity of both b-SSFP \n(acquisition time: 11-15 seconds) and diagnostic T2 -weighted TSE (3T, \nacquisition time: 4-5 minutes) sequences using a 4- point scale (3: good, 2: \nacceptable, 1: poor, 0: impossible for lesion ident ification). Recognition rates \n(RR) were calculated as the percentage of cases wit h sufficient image quality \n(scores of 3 or 2). Subgroup analyses were performe d by zonal location \n(PZ/TZ), lesion size (</≥ 0.5 mL), and PI RADS score (3/4-5). Differences \nbetween readers and sequences were analysed using M cNemar's test (p < \n0.05). \nResults or Findings: The RR for the T2-weighted reference sequence was \n98% for both radiologists, with subgroups ranging f rom 94% to 99%. For b-\nSSFP, the RR was 87% for R1 and 81% for R2, with su bgroups ranging from \n75% (PI-RADS 3) to 92-93% (PI-RADS 4/5, large lesio ns). No significant \ndifferences were found between readers. RR differen ces between sequences \nwere statistically significant, except for TZ and l arge lesions rated by R1. \nConclusion: b-SSFP showed only moderately lower RR than the ver y high RR \nof T2-weighted reference images, especially for the  experienced reader R1. \nGiven its much shorter acquisition time, b-SSFP of discernible lesions therefore \nhas the potential to reduce biopsy times, particula rly for large or (highly) \nsuspicious lesions (PI-RADS 4-5). \nLimitations: Retrospective; single-center. \nFunding for this study: None \nEthics committee - additional information: Ethics committee was consulted, \nwritten informed consent was obtained from particip ating patients \nAuthor Disclosures:  \nNicolas Linder: Nothing to disclose \nTimm Denecke: Nothing to disclose \nHarald Busse: Nothing to disclose \nLars-Christian Horn: Nothing to disclose \nConstantin Ehrengut: Nothing to disclose \nAlexander Schaudinn: Nothing to disclose  \nJens-Uwe Stolzenburg: Nothing to disclose \nToni Franz: Nothing to disclose  \nChristian Peter: Nothing to disclose \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 41  \nWednesday \nDiagnostic assessment of early DWI changes after Si ngle-Dose Ablative \nRadiation Therapy for localized prostate cancer \n*P. N. Franco*, C. R. G. L. O. M. Talei Franzesi, C . Maino, R. Corso,  \nD. Ippolito; Monza/IT \n(francopaoloniccolo@gmail.com) \n \nPurpose or Learning Objective: To investigate the diagnostic value of \ndiffusion-weighted (DWI) MRI early changes, 1 hour after treatment, in patients \nwith organ-confined unfavourable prostate cancer (P Ca) treated with Single-\nDose Ablative Radiation Therapy (SDART), in compari son with biochemical \nmarkers. \nMethods or Background: Twenty-four patients with intermediate unfavourable  \nor high-risk localized PCa treated with SDART (21 G y on the entire prostate \nwith boost up to 24 Gy on the focal lesion) associa ted with hormone therapy \nwere prospectively enrolled. Each patient was exami ned with a 3T scanner \nfour times: (1) 1-2 weeks before RT (t0) for treatm ent planning; (2) 1 hour after \ntreatment (t1); (3) 3 months after treatment (t2); (4) 2 years after treatment (t3). \nRegions of interest (ROIs) were plotted on apparent  diffusion coefficient (ADC) \nmaps and T2-HR sequences on lesions, benign periphe ral zone, and the entire \nprostate gland. Patients’ laboratory data (PSA and testosterone) was collected. \nResults or Findings: ADC values significantly increased in neoplastic le sions \nat t1, t2 and t3 (+22%, +43% and +53%, respectively ). Conversely, no \nsignificant changes were observed in the benign per ipheral zone and the entire \nprostate gland. On T2 sequences, signal intensity p rogressively decreased in \nthe benign peripheral zone (t1: -1%; t2: -33%; t3: -42%) and in the entire \nprostate gland (t1: -6%; t2: -24%; t3: -31%), while  no significant changes were \nobserved in lesions. All patients except one had a complete biochemical \nresponse. \nConclusion: The study findings showed high diagnostic value of DWI and a \ngood correlation between early (t1) changes in ADC values after SDART and \nlater (t2 and t3) tumour response (both biochemical  and imaging) in patients \nwith unfavourable PCa. Early DWI changes can repres ent a useful parameter \nto evaluate treatment response and predict patients ’ outcomes. \nLimitations: Small sample size; associated hormone therapy. \nFunding for this study: None \nEthics committee - additional information: The local ethics committee \nformally approved this study. \nAuthor Disclosures:  \nCesare Maino: Nothing to disclose \nCammillo Roberto Giovanni Leopoldo Oreste Massimili ano Talei Franzesi: \nNothing to disclose \nRocco Corso: Nothing to disclose \nPaolo Niccolò Franco: Nothing to disclose \nDavide Ippolito: Nothing to disclose \n \n \nMaximal radial distance as a new parameter for pred icting extraprostatic \nextension of prostate cancer on multiparametric mag netic resonance \nimaging: a histo-radiological study \n*F. Porões*, A. Nobile, L. Widmer, J. A. Vidal, J. Di Vincenzo, H. Najberg,  \nJ. M. M. Froehlich, C. Reischauer, H. Thoeny; Fribo urg/CH \n(fabioporoes@gmail.com) \n \nPurpose or Learning Objective: We introduce a new parameter for predicting \nextraprostatic extension (EPE) on multiparametric m agnetic resonance \nimaging (mpMRI): the maximal radial distance (maxRA DD). It corresponds to \nthe largest diameter of a prostate cancer focus (PC F) perpendicular to a \ncontact with the prostate pseudocapsule. We compare  accuracy and \nreproducibility of maxRADD with the previously prop osed maximal capsular \ncontact length (maxCCL) for predicting EPE. \nMethods or Background: We retrospectively and consecutively included 81 \npatients undergoing prostate mpMRI between October 2018 and December \n2020, followed by radical prostatectomy. One uropat hologist with 8 years of \nexperience collected for each PCF: location, maxCCL , maxRADD, and \npresence/absence of EPE. Four radiologists with 0, 2, 3, and 6 years of \nexperience in prostate mpMRI determined maxRADD and  maxCCL on mpMRI \nfor each PCF twice in separate readings. Accuracy i n predicting EPE was \nassessed using the area under the curve (AUC), with  the pathologic findings as \nthe gold standard. Inter-/intra-reader agreement we re assessed using \nintraclass correlation coefficients (ICCs) and Cron bach’s alpha. \nResults or Findings: On histolpathology, there was no significant differ ence in \nthe accuracy of predicting EPE between maxRADD and maxCCL \n(AUCmaxRADD = 0.92, AUCmaxCCL = 0.91, p = 0.28). Pe arson correlation \nshowed a strong correlation of both parameters dete rmined on mpMRI with \ntheir histopathological counterparts (>0.7), with t he exception of maxCCL \nassessed by the reader w/o experience in prostate m pMRI (0.54). On mpMRI, \ninter-reader agreement was significantly higher for  maxRADD (ICCmaxRADD \n= 0.96, ICCmaxCCL = 0.94, p = 0.046) and intra-read er agreement was higher \nbut did not reach significance (average alphamaxRAD D = 0.95, average \nalphamaxCCL = 0.92, p = 0.31). \nConclusion: MaxRADD permits assessing EPE with good accuracy an d \nshows higher reproducibility compared with maxCCL. \nLimitations: No significant limitation. \nFunding for this study: This study has received funding by the Swiss Nation al \nScience Foundation (Grant/Award Number: 32003B_1762 29/1) and the HFR \nResearch GRANT (2352). \nEthics committee - additional information: The study was approved by our \ninstitutional ethics committee (CER-VD). The ethics  committee notification can \nbe found under the project-ID 2020-01859. \nAuthor Disclosures:  \nHarriet Thoeny: Advisory Board: Guerbet \nJana Di Vincenzo: Nothing to disclose \nHugo Najberg: Nothing to disclose \nJohannes Malte Maria Froehlich: Consultant: Guerbet  \nAntoine Nobile: Nothing to disclose \nFabio Porões: Nothing to disclose \nLucien Widmer: Nothing to disclose \nJulian Alexis Vidal: Nothing to disclose \nCarolin Reischauer: Nothing to disclose \n \n \nImpact of Centrally Reviewed PI-QUAL v2 Scores on t he Diagnostic \nPerformance of Prostate MRI \n*G. Brembilla*, D. Cannoletta, M. Cosenza, F. Pelle grino, M. E. Porzi,  \nL. Quarta, A. Stabile, A. Briganti, F. De Cobelli; Milan/IT \n \nPurpose or Learning Objective: To assess the impact of image quality, \ndefined by PI-QUAL v2 scores, on the diagnostic yie ld of prostate MRI in \ncentrally reviewed scans. \nMethods or Background: We retrospectively identified consecutive patients \nwho underwent MRI-targeted and systematic biopsies at our Institution \n(January 2023 - June 2024), with MRI performed exte rnally. All the external \nMRI scans were centrally reviewed by an experienced  uro-radiologist, who \nassigned PI-QUAL v2 and PI-RADS v2.1 scores. We ass essed the proportion \nof PI-RADS 3 lesions and the detection rate of clin ically significant prostate \ncancer (csPCa), stratified by PI-QUAL v2 scores, be fore and after central \nrevision. Histopathological results from the biopsi es were used as the \nreference standard. \nResults or Findings: A total of 151 consecutive patients were included i n the \nanalysis. 37/151 (24%) of the MRI scans were scored  PI-QUAL 1, 72/151 \n(48%) PI-QUAL 2, and 42/151 (38%) PI-QUAL 3. Based on original reports, the \noverall proportion of PI-RADS 3 scans was 34/151 (2 3%). In PI-QUAL 1-2 vs 3 \nscans, the proportion of PI-RADS 3 in was 27% vs 12 %, respectively; the \ncsPCa detection rate was 48% vs 62%, respectively. The reclassification rate \nof PI-RADS scores at central review was 64/151 (42% ), and was higher for PI-\nQUAL 1-2 scans (47%) than for PI-QUAL 3 scans (31%) . After central revision, \nthe overall proportion of PI-RADS 3 in PI-QUAL 1-2 vs 3 was 19% vs 5%, \nrespectively; the detection rate of csPCa was 58% v s 78%, respectively. \nConclusion: Lower prostate MRI image quality, as defined by the  PI-QUAL v2 \nscoring system, is associated with a higher proport ion of equivocal scans (PI-\nRADS 3) and a reduced csPCa detection rate in centr ally revised MRI scans. \nLimitations: Small sample size, only one radiologist for review \nFunding for this study: None \nEthics committee - additional information: IRB approved \nAuthor Disclosures:  \nMaria Elena Porzi: Nothing to disclose \nDonato Cannoletta: Nothing to disclose \nMichele Cosenza: Nothing to disclose \nFrancesco Pellegrino: Nothing to disclose \nAlberto Briganti: Nothing to disclose \nArmando Stabile: Nothing to disclose \nLeonardo Quarta: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \nGiorgio Brembilla: Nothing to disclose \n \n \nProstate Volume Assessment on MRI: Comparison of fu ll manual \nsegmentation to PIRADS-based approximation in 2 pla nes and its \ninfluence on PSA-density \n*J. Uhlig*, L. Biggemann, C. Louizi, A. Uhlig; Gött ingen/DE \n(johannes.uhlig@med.uni-goettingen.de) \n \nPurpose or Learning Objective: To evaluate differences in prostate volume \nquantification on MRI comparing full manual segment ation and PIRADS-based \napproximation in 2 planes. \nMethods or Background: Patients imaged with 3T mpMRI (Siemens VIDA) \nfor suspected prostate cancer between 2021-2023 wer e included. PSA \nmeasurements were obtained at the time of mpMRI or extracted from patients \nrecords up to 3 months prior. Manual segmentation o f the prostate was \nperformed on all axial T2w slices serving as refere nce standard. Prostate \nvolume was approximated using 3 measurements on T2w  sagittal and axial \nplanes according to the PIRADSv2.1 recommendations.  Prostate volumes from \nmanual segmentation and approximation were compared  and the influence on \nPSA-density quantified using different cut-off valu es. \n\n \n \nAbstract-based Programme \n \n 42  \nWednesday \nResults or Findings: n=331 patients were included (mean age 67 ± 7 years ) \nwith a mean PSA value of 8.1 ± 5.7 ng/ml. Mean prostate volume using manual \nsegmentation was 63.3cc (± 33cc), with n=142 patients having a volume of \n<=50cc, 51-100cc: n=148, 101-150cc: n=32, and >151c c: n=8. The mean \nabsolute difference of prostate volume using approx imation vs. segmentation \nwas 9.1cc (±9.3cc, p=0.01). In general, smaller pro state volumes were \noverestimated, and larger volumes underestimated by  approximation. Using a \nPSA-density cut-off <0.1 ng/ml/cc, the approximatio n method yielded an \naccuracy = 88%. Using a PSA-density cut-off <0.15 n g/ml/cc, the \napproximation method yielded an accuracy = 90%. \nConclusion: Using PIRADSv2.1-based approximation of prostate vo lume on \nmpMRI yields a statistically significant difference  when compared to full manual \nsegmentation. These differences have a relevant eff ect on PSA-density \ncalculation with potential impact on downstream pat ient management. \nLimitations: Patients were recruited in only one tertiary center  and imaged on \none MRI scanner, which could limit the generalizabi lity of presented results. \nFunding for this study: This study received no funding. \nEthics committee - additional information: Ethics committee of the \nUniversity Medical Center Goettingen \nAuthor Disclosures:  \nAnnemarie Uhlig: Nothing to disclose \nJohannes Uhlig: Nothing to disclose \nLorenz Biggemann: Nothing to disclose \nChiheb Louizi: Nothing to disclose \n \n \nMRI without contrast media injection for prostate c ancer screening: \nresults from Prostate Cancer Secondary Screening in  Sapienza (PROSA) \n*E. Messina*, A. Borrelli, L. Laschena, S. Lucciola , M. Pecoraro,  \nV. Panebianco; Rome/IT \n(emanuele.messina@uniroma1.it) \n \nPurpose or Learning Objective: PROSA is a randomized MRI-based \nscreening protocol, investigating the role of MRI w ithout contrast media \ninjection (bi-parametric MRI, bpMRI) as secondary p revention test for prostate \ncancer (PCa) early diagnosis, comparing MRI with PS A-test. PROSA aims to \ninvestigate the efficiency of this screening protoc ol, both in terms of diagnostic \naccuracy, and cost-effectiveness. \nMethods or Background: 590 men aged 49 to 69 years were enrolled and \nblindly randomized into two different arms: (A) Men  underwent bpMRI \nregardless of their PSA values; (B) Men with increa sed PSA were directed to \nbpMRI, while those with normal PSA were not. Men sc reened positive on MRI \nwere directed to MR-directed targeted biopsy. To ev aluate the efficiency of the \nprotocol we calculated the experimental event rate (EER), control event rate \n(CER), absolute risk reduction (ARR), number needed  to treat (NNT). Health \nTechnology Assessment analysis was implemented to e valuate the cost-\neffectiveness. The cost/effectiveness ratio is calc ulated as follows: Delta Costs/ \nDelta effectiveness = (CA–CB)/(EA-EB). \nResults or Findings: 289 men were randomized on Arm A and among them \n15 clinically significant PCa (csPCa) were detected ; 291 men were randomized \non Arm B, with 6 csPCa detected (p=0.04). On arm A,  8 men diagnosed with \ncsPCa (53.3%) presented normal PSA levels. Consider ing the efficiency of the \nscreening protocol, EER was 5.23%, CER 2.06%, ARR 3 .17%, and NNT 31.6. \nTherefore 32 interventions (in this study MRIs) are  needed to find one event (in \nthis study one csPCa). The final cost/effectiveness  ratio resulted to be € \n3.562,61 for the diagnosis of one csPCa. \nConclusion: Prostate MRI without contrast media injection showe d promising \nresults compared to the use of PSA analysis alone a s a screening tool, both in \nterms of efficiency and cost-effectiveness. \nLimitations: Single center \nFunding for this study: No \nEthics committee - additional information: CE Approved \nAuthor Disclosures:  \nSara Lucciola: Nothing to disclose \nValeria Panebianco: Nothing to disclose \nAntonella Borrelli: Nothing to disclose \nEmanuele Messina: Nothing to disclose \nMartina Pecoraro: Nothing to disclose \nLudovica Laschena: Nothing to disclose \n \n \n \n \n \n \n \n \n \n13:00-14:30 Research Stage 3 \nResearch Presentation Session: Neuro \nRPS 411 \nAging brain and neurodegeneration \nimaging \n \nModerator \nF. Barkhof; Amsterdam/NL  \n(f.barkhof@vumc.nl) \nAuthor Disclosures:  \nFrederik Barkhof: Advisory Board: Combinostics, Sco ttish Brain Sciences, \nAlzheimer Europe, Merck; Author: Clinical Neuroradi ology - The ESNR \ntextbook; Consultant: Roche, Celltrion, Rewind Ther apeutics, Bracco; Founder: \nQueen Square Analytics; Grant Recipient: Roche, UK MS Society, Biogen, \nMerck, ADDI; Other: DSMB member Prothena, EISAI \n \n \nConnecting the Dots: Linking White Matter Hyperinte nsity Patterns to \nLongitudinal Cognitive Changes in Aging \n*M. M. Courtney*, R. A. Kenny, J. F. Meaney, C. De Looze; Dublin/IE \n(michaelc0001@hotmail.com) \n \nPurpose or Learning Objective: White matter hyperintensities (WMHs) are \nknown to correlate with cognitive decline, stroke, and dementia. Previous \nresearch has explored the independent effects of wh ite matter macrostructure, \nmicrostructure, and spatial distribution on cogniti ve function, yet a \ncomprehensive analysis combining elements is limite d. We provide a \ncomprehensive WMH analysis to assess their associat ion with cognitive \ndecline over a six-year period. \nMethods or Background: Data was obtained from The Irish Longitudinal \nStudy on Ageing (TILDA), including MRI scans and co gnitive performance \nscores from 497 community-dwelling older adults. WM Hs were segmented \nusing Lesion Prediction Algorithm, analysed using E xplore DTI for diffusion \nmetrics. Statistical Analysis performed in R-Studio . Linear mixed effect models \nused to assess relationship between lesion phenotyp es and cognitive decline, \nadjusting for demographic and health-related variab les. \nResults or Findings: 11,933 WMHs analysed. Average 24 lesions per subjec t. \nAverage lesion volume 263mm3. Average lesion FA 0.2 9, MD 1.10. K-means \nclustering identified 3 primary WMH phenotypes. Dee p WMHs associated with \nolder age and 2 or more cardiovascular risk factors  (p<0.001 respectively). \nHigher volume lesions were associated with cardiova scular risk factors \n(p<0.001), smoking (p<0.01) and older age (p<0.001) . High-volume, low FA \nlesions in both deep (p=0.5) and periventricular (p =0.04) white matter exhibited \naccelerated cognitive decline over six years. Incre ased number of \nperiventricular lesions was associated with cogniti ve decline (p<0.01). \nConclusion: WMHs manifest diverse phenotypes associated with co gnitive \ndecline. High-volume, low FA lesions in both perive ntricular and deep white \nmatter are predictive of cognitive deterioration. I dentifying WMH phenotypes \nmay inform early intervention strategies and improv e patient outcomes by \ntargeting individuals at higher risk of cognitive d ecline. \nLimitations: Cross-sectional MRI data implies findings are assoc iative and \ncorrelate with longitudinal findings, but correlati on does not equal causation. \nFurther investigation with serial MRI would provide  more reliable data. \nFunding for this study: The Irish Longitudinal Study on Ageing is funded by  \nthe Irish Department of Health, the Atlantic Philan thropies and Irish Life. \nEthics committee - additional information: This study was approved by the \nTrinity College Faculty of Health Sciences Research  Ethics Committee, Dublin, \nIreland. Protocols conformed with the Declaration o f Helsinki. Signed informed \nconsent was obtained from all respondents prior to participation. Additional \nethics approval was received for the magnetic reson ance imaging (MRI) sub-\nstudy from the St James’s Hospital/Adelaide and Mea th Hospital, Inc. National \nChildren’s Hospital, Tallaght (SJH/AMNCH) Research Ethic Committee, Dublin, \nIreland. Those attending for MRI also completed an additional MRI-specific \nconsent form. (De Looze et al) \nAuthor Disclosures:  \nMichael Michael Courtney: Nothing to disclose \nCéline De Looze: Nothing to disclose \nRose Anne Kenny: Nothing to disclose \nJames F Meaney: Nothing to disclose \n \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 43  \nWednesday \nNormal Aging-Related Brain Morphological Connectivi ty Network Linked \nto Multiple Neurological Diseases \n*L. Yuna*; Beijing/CN \n(yunali1094@163.com) \n \nPurpose or Learning Objective: Disentangling the complex interaction \nbetween aging and various aging-related neurologica l diseases at the brain \nnetwork level. \nMethods or Background: We presented a unified analysis framework to \nidentify aging-related morphological connectivity n etworks (MCNs) and \ndetermined their clinical relevance in various neur ological diseases (including \nmild cognitive impairment, Alzheimer's disease, Par kinson’s disease, small \nvessel disease multiple sclerosis and multiple scle rosis). First, individual MCNs \nin the HC group were constructed and further decomp osed into distinct \nsubnetworks using linked independent component anal ysis. Aging-related \nsubnetworks were defined as those significantly ass ociated with age. The \naging-related subnetworks were spatially correlated  with disease-related MCN \ndisruptions. Further, the regression coefficients o f the aging-related \nsubnetworks were calculated for each patient’s MCN using linear regression. \nThe regression coefficients were then correlated wi th various clinical variables \nwithin each disease group to assess the clinical si gnificance of the aging-\nrelated subnetworks. Finally, a series of annotated  biological maps were \nutilized to advance the biological interpretation o f the identified aging-related \nsubnetworks. \nResults or Findings: We first identified three aging-related subnetworks , \nincluding the perceptual-limbic subnetwork, attenti on-somatomotor \nsubnetwork, and somatomotor-predominant subnetwork,  that exhibited distinct \naging trajectories. Normal aging interacted with va rious neurological diseases, \nexhibiting both transdiagnostic and diagnosis-speci fic patterns at the brain \nnetwork level. The aging-related subnetworks were c losely related to cognitive \nand physical performance in patients. Biological co rrelation analysis revealed \nthat glucose metabolism and several neurotransmitte rs, such as cannabinoids \nand dopamine, played critical roles in aging-relate d subnetworks. \nConclusion: This study elucidated the network mechanisms underl ying the \ncomplex interactions between aging and neurological  diseases, offering \ninsights that could improve clinical management and  therapy development by \nevaluating aging effects. \nLimitations: This study is limited by an uneven sample distribut ion, variability \nin disease durations, and the absence of longitudin al research \nFunding for this study: We demonstrated how normal aging interacted with \nvarious neurological diseases at the brain network level, both \ntransdiagnostically and diagnosis-specifically. The  identified aging-related \nsubnetworks might serve as imaging markers to disti nguish normal aging \neffects from disease-specific mechanisms, thereby i mproving disease \nmonitoring and management. \nEthics committee - additional information: Beijing Tiantan Hospital, Capital \nMedical University, Beijing, China; No. KY-2019-050 -02 \nAuthor Disclosures:  \nLi Yuna: Nothing to disclose \n \n \nA comparative evaluation of four commercially avail able artificial \nintelligence software solutions for brain volumetry  and lesion \nsegmentation in dementia \n*G. Di Cerbo*, G. Saltarelli, A. Innocenzi, M. Cell a, C. De Felici, F. Bruno,  \nA. Splendiani, E. Di Cesare; L'Aquila/IT \n(giovannidicerbo96@gmail.com) \n \nPurpose or Learning Objective: The purpose of this study is to compare the \noperating features and analysis outputs of four dif ferent commercially available \nsoftware for brain volumetric analysis. \nMethods or Background: We analyzed consecutive brain MRI scan of 32 \npatients (25 males, aged between 50 and 90 years) e valuated in a singles \nInstitution for cognitive decline. All MRI examinat ions were performed on 3T \nscanner (GE MR750w.), including a volumetric T1 GRE  sequence (slice 1 mm, \nTR 8.5, frequency FoV 25.6, phase FoV 0.8). MRI dat a were analyzed through \nfour different dedicated softwares (S1, S2, S3, S4)  after quality check by an \nexperienced neuroradiologist. Volumetric output dat a of brain segmentation \nand volume for frontal, temporal, parietal, occipit al lobes, hippocampus, and \nlateral ventricles, were collected and compared. \nResults or Findings: The results revealed no significant consensus among  \nthe four artificial intelligence software applicati ons in measuring various brain \nareas. S1-S2 showed non statistically significant o utput values in all brain \nregions. S1-S3 showed statistically significant dif ferences in frontal and parietal \nlobe, lateral ventricles and hippocampus. S1-S4 sho wed statistically significant \ndifferences in frontal parietal and occipital lobe,  and lateral ventricles. S2-S3 \nshowed statistically significant differences in tem poral and occipital lobes. S3-\nS4 showed statistically significant differences in parietal and occipital lobes, \nhippocampus and lateral ventricles. \nConclusion: Although AI software are becoming increasingly popu lar in \nclinical practice, the findings indicate a low degr ee of concordance among the \nfour applications evaluated in this study. Therefor e, clinicians integrating these \ntools into routine practice should be aware of the limited result \ninterchangeability across different software platfo rms and consider their use as \ncomplementary aids rather than substitutes for clin ical expertise. \nLimitations: Small sample size \nFunding for this study: None \nEthics committee - additional information: Local IRB \nAuthor Disclosures:  \nAlessandra Splendiani: Nothing to disclose \nGaspare Saltarelli: Nothing to disclose \nAntonio Innocenzi: Nothing to disclose \nErnesto Di Cesare: Nothing to disclose \nGiovanni Di Cerbo: Nothing to disclose \nMarco Cella: Nothing to disclose \nClaudia De Felici: Nothing to disclose \nFederico Bruno: Nothing to disclose \n \n \nGlymphatic dysfunction mediates the impact of tau p athology on \nneurodegeneration in cognitively unimpaired individ uals and prodromal \nAlzheimer’s Disease \nX. Xu, B. Zhang, *Z. Zhu*; Nanjing/CN \n \nPurpose or Learning Objective: To elucidate the pathological mechanism of \nglymphatic system dysfunction are associated with r egional tau deposition and \ntau-mediated neurodegeneration across the preclinic al and prodromal stage of \nthe AD continuum. \nMethods or Background: Cognitively normal (CN) controls (n=94), individual s \nwith mild cognitive impairment (MCI; n = 83), and t hose with significant \nmemory concern (SMC; n =84) were included from the Alzheimer's Disease \nNeuroimaging Initiative. Tau pathology was measured  by positron emission \ntomography, the glymphatic activity assessed by dif fusion tensor image \nanalysis along the perivascular space (DTI-ALPS), a nd neurodegeneration \nreflected by hippocampal volume. Mediation analysis  was used to study the \npossible pathways. \nResults or Findings: ALPS was significantly associated with tau and tau-\nmediated neurodegeneration, especially in parahippo campal gyrus. The \nrelationship between glymphatic function and neurod egeneration was \nmediated by tau pathological deposition (indirect e ffect: 0.012, 95%CI [0.001—\n0.029]) rather than ALPS index mediated the relatio nship between tau and \nneurodegeneration (indirect effect: 0.212, 95%CI [- 0.013,0.0002]). The \nrelationship between glymphatic dysfunction and cog nitive decline were fully \nmediated by tau deposition and neurodegeneration in  preclinical AD. \nConclusion: Tau deposition in specific region may mediate the r elationship of \nglymphatic dysfunction and neurodegeneration, which  contribute to cognitive \ndecline in the preclinic AD stage, facilitating the  development of therapeutics \ntargeting tau protein and glymphatic dysfunction in  AD. \nLimitations: Firstly, the cross-sectional design employed in thi s study limits us \nto explore the causal relationships or investigate longitudinal changes over \ntime, future longitudinal studies could provide fur ther. Secondly, the ALPS \nindex is mainly used to measure the function of the  subcortical glymphatic \nsystem. Future research should continue to use othe r methods, such as \nBOLD-CSF coupling measurements, to validate the fun ction of the subcortical \nglymphatic system in AD. \nFunding for this study: This work was supported by the National Science and  \nTechnology Innovation 2030 -- Major program of \"Bra in Science and Brain-Like \nResearch\" (2022ZD0211800); the National Natural Sci ence Foundation of \nChina (82271965, 81971596, 82001793); the Fundament al Research Funds \nfor the Central Universities, Nanjing University (2 020-021414380462); the Key \nScientific Research Project of Jiangsu Health Commi ttee (K2019025); Special \nFunded Project of Nanjing Drum Tower Hospital (No. RC2022-023), \nDevelopment Plan (Social Development) Project of Ji angsu Province (No. \nBE2022679). China Postdoctoral Science Foundation ( 2023M741648). The \nNational Natural Science Foundation of China (82302 172); The funders had no \nrole in the study design, data collection and analy sis, decision to publish, or \npreparation of the manuscript. \nEthics committee - additional information: ADNI Ethics committee \nAuthor Disclosures:  \nBing Zhang: Nothing to disclose \nZhengyang Zhu: Nothing to disclose \nXinru Xu: Nothing to disclose \n \n \nThe Impact of Temporal Muscle Thickness as an Indic ator of Sarcopenia \non Clinical Status in Parkinson's Disease \nB. Atalay, K. Erincik, M. B. Doğan, *M. Gezgin*, H. Yıldız, F. B. Ozdilek; \nIstanbul/TR \n(me.egezgin@hotmail.com) \n \nPurpose or Learning Objective: To assess the impact of temporal muscle \nthickness (TMT), as an indicator of sarcopenia on c ognitive status and \nmedication dosage in patients with Parkinson's dise ase (PD). \n\n \n \nAbstract-based Programme \n \n 44  \nWednesday \nMethods or Background: A total of 54 patients with PD and 46 healthy \ncontrols were retrospectively analyzed. Brain MR im ages from both groups \nwere reviewed by two radiologists who independently  measured the right and \nleft temporal muscle thickness using T1-weighted ax ial images. Clinical \nassessments included the Unified Parkinson’s Diseas e Rating Scale (UPDRS), \nMini-Mental State Examination (MMSE), Hoehn and Yah r Scale, L-dopa \nequivalent daily dose (LEDD), and disease duration,  collected by a neurologist. \nInterobserver agreement was evaluated using the int raclass correlation \ncoefficient. The relationship between TMT and clini cal data was analyzed using \nSpearman’s correlation. \nResults or Findings: In the PD group, 33.3% of patients were female, \ncompared to 50% in the control group. Interobserver  agreement for TMT \nmeasurements was excellent. No significant differen ce in TMT was observed \nbetween the PD and control groups (p=0.16, p=0.34).  A weak but statistically \nsignificant correlation was found between TMT, LEDD , and disease duration, \nwhile no correlation was found between TMT and UPDR S or Hoehn and Yahr \nscores. A weak but significant correlation was obse rved between TMT and \nMMSE scores. \nConclusion: Sarcopenia, characterized by muscle mass loss, shar es \ncontributing factors with Parkinson’s disease. TMT,  a reliable marker of muscle \nmass, can be measured on brain MRIs to assess sarco penia risk in PD \npatients. Our findings suggest that TMT correlates with cognitive function and \nmedication dosage in PD patients, making it a valua ble tool for early \nsarcopenia detection and management in clinical set tings. \nLimitations: Lack of patient follow-up and inability to assess t he impact of \nmuscle mass increase on medication dosage. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number 2023/0912. \nAuthor Disclosures:  \nMerve Gezgin: Nothing to disclose \nHüseyin Yıldız: Nothing to disclose \nKendal Erincik: Nothing to disclose \nFatma Betül Ozdilek: Nothing to disclose \nBaşak Atalay: Nothing to disclose \nMahmut Bilal Doğan: Nothing to disclose \n \n \nCan a Coronal Swallow Tail Cleft Sign Increase the Confidence of \nVisualization of the Nigrosome-1 Layer of the Subst antia Nigra in Normal \nSubjects and Those with Parkinsonism? \n*S. Rajan*, J. S. Chatha, H. Mahajan; New Delhi/IN \n(drsriramrajan@gmail.com) \n \nPurpose or Learning Objective: To assess whether reconstruction of phase \nimages from susceptibility-weighted imaging (SWI) i n a coronal plane \nenhances the visualization confidence of the nigros ome-1 layer in the \nsubstantia nigra. The coronal Swallow Tail cleft si gn may serve as a valuable \nindicator for this structure. \nMethods or Background: A retrospective review was conducted on 433 \nconsecutive MR brain scans acquired using a routine  protocol for various \nindications. The axial phase images of SWI were ana lyzed in Phase 1 by two \nradiologists with 9 and 22 years of experience. The y evaluated the visibility of \nthe nigrosome-1 layer below the level of the red nu cleus using a 5-point Likert \nscale (1 = very difficult to 5 = easily seen). In P hase 2, coronal reconstructed \nimages of the phase SWI were assessed perpendicular ly to the substantia \nnigra in posterior sections for the presence of the  cleft sign, also scored with \nthe 5-point Likert scale. \nResults or Findings: In Phase 1, the Likert scores of 1 and 2 were simil ar (6% \nand 9%, respectively). There was a marginal decreas e in score 3 (from 4.97% \nto 4.04%). Scores of 4 decreased significantly from  16.97% in Phase 1 to \n7.96% in Phase 2, while the score of 5 increased su bstantially from 62.01% in \nPhase 1 to 72.97% in Phase 2. Statistical analysis demonstrated that the \naddition of the coronal Swallow Tail cleft sign sig nificantly enhanced \nconfidence in visualizing the nigrosome-1 layer (p < 0.01). \nConclusion: The coronal Swallow Tail cleft sign significantly i mproves the \nconfidence of visualization of the nigrosome-1 laye r in both normal subjects \nand those with Parkinsonism. This finding underscor es the utility of coronal \nreconstructions in enhancing diagnostic accuracy in  neuroimaging. \nLimitations: Correlation with nuclear scans or clinical history was not done \nFunding for this study: None \nEthics committee - additional information: Restrospective \nAuthor Disclosures:  \nJagneet Singh Chatha: Nothing to disclose \nSriram Rajan: Nothing to disclose \nHarsh Mahajan: Nothing to disclose \n \n \n \n \n \nMachine learning approach effectively discriminates  between Parkinson’s \ndisease and progressive supranuclear palsy: multi-l evel indices of rs-\nfMRI \n*W. Cheng*; Nanchang/CN \n(ndyfy04042@ncu.edu.cn) \n \nPurpose or Learning Objective: Parkinson’s disease (PD) and progressive \nsupranuclear palsy (PSP) present similar clinical s ymptoms, but their treatment \noptions and clinical prognosis differ significantly . Therefore, we aimed to \ndiscriminate between PD and PSP based on multi-leve l indices of rs-fMRI via \nthe machine learning approach. \nMethods or Background: A total of 58 PD and 52 PSP patients were \nprospectively enrolled in this study. Participants were randomly allocated to a \ntraining set and a validation set in a 7:3 ratio. V arious resting-state functional \nmagnetic resonance imaging (rs-fMRI) indices were e xtracted, followed by a \ncomprehensive feature screening for each index. We constructed fifteen \ndistinct combinations of indices and selected four machine learning algorithms \nfor model development. Subsequently, different vali dation templates were \nemployed to assess the classification results and i nvestigate the relationship \nbetween the most significant features and clinical assessment scales. \nResults or Findings: The classification performance of logistic regressi on \n(LR) and support vector machine (SVM) models, based  on multiple index \ncombinations, was significantly superior to that of  other machine learning \nmodels and combinations when utilizing automatic an atomical labeling (AAL) \ntemplates. This has been verified across different templates. \nConclusion: The utilization of multiple rs-fMRI indices signifi cantly enhances \nthe performance of machine learning models and can effectively achieve the \nautomatic identification of PD and PSP at the indiv idual level. \nLimitations: Only the rs-fMRI index was used in this study, and DTI-related \nmicrostructure data was not included. \nFunding for this study: This study was supported by the National Natural \nScience Foundation of China (82160331), Jiangxi Pro vince Double Thousand \nTalent Plan (jxsq2023201039). This project is imple mented by the Jiangxi \nClinical Research Center for Medical Imaging (20223 BCG74001), and Jiangxi \nProvince Key Laboratory for Precision Pathology and  Intelligent Diagnosis \n(2024SSY06281). \nEthics committee - additional information: This study was approved by the \nMedical Ethics Committee of the First Affiliated Ho spital of Nanchang \nUniversity (approval number: IIT2022124). \nAuthor Disclosures:  \nWeiling Cheng: Nothing to disclose \n \n \nExploring the Relationship Between Body Composition  and Brain \nMorphology in Aging: A Focus on Thigh Muscle Mass a nd Subcutaneous \nFat as Predictors of Cortical Thickness in Healthy Older Adults \n*M. Sarkinaite*¹, U. Lukoseviciute¹, N. Masiulis¹, S. Lukoševičius¹, O. Levin²,  \nR. Gleiznienė¹; ¹Kaunas/LT, ²Leuven/BE \n(milda.sarkinaite@gmail.com) \n \nPurpose or Learning Objective: Subcutaneous fat accumulation has been \nlinked to adverse brain health outcomes. This study  examines the relationship \nbetween thigh muscle mass, subcutaneous fat distrib ution, and brain structure \nin elderly adults. It explores how body composition  affects cortical thickness in \nbrain regions linked to cognitive function. \nMethods or Background: Fifty-four healthy elderly individuals underwent \nimaging of the right thigh and brain using a 3T Sie mens Avanto MRI system. \nMuscle and subcutaneous fat cross-sectional areas ( CSA) were measured at \n50% and 20% of thigh length, with the muscle-to-fat  ratio calculated at the 50% \nmark. Cortical thickness was assessed through brain  volumetric analysis using \nFreesurfer 7.4.1 software. \nResults or Findings: Significant positive correlations (r ≥ 0.2, p ≤ 0.05) were \nfound between the muscle-to-fat ratio and cortical thickness in the left \ncerebellum, cuneus, and transverse temporal cortex,  in the right entorhinal \ncortex, inferior temporal cortex, postcentral gyrus , superior temporal cortex, \nand the banks of the superior temporal sulcus (BANK SSTS). Additionally, \nsignificant negative correlations (r ≤ -0.2, p ≤ 0.05) were observed between \nsubcutaneous fat CSA at 50% of thigh length and cor tical thickness in the left \ncuneus and entorhinal cortex, in the right BANKSSTS , postcentral gyrus, and \nsuperior temporal cortex. Additionally, subcutaneou s fat CSA at 20% of thigh \nlength was inversely correlated with cortical thick ness in the left cuneus and \nright BANKSSTS (r ≤ -0.2, p ≤ 0.05). \nConclusion: Our findings demonstrate that increased thigh muscl e mass \ncorrelates with greater cortical thickness in cogni tive regions, while elevated \nsubcutaneous fat is linked to reduced thickness. Th ese results highlight the \nrole of body composition in maintaining brain healt h in the elderly and \nunderscore the importance of muscle mass in mitigat ing age-related cortical \ndecline. \nLimitations: With only 54 participants, the study's findings may  lack \ngeneralizability. \nFunding for this study: Supported by the Research Council of Lithuania \n(grant number P-MIP-22-217). \n\n \n \nAbstract-based Programme \n \n 45  \nWednesday \nEthics committee - additional information: Approved by the Kaunas region \nMedical Ethics Committee for Biomedical Research (N o. BE-2-22). \nAuthor Disclosures:  \nMilda Sarkinaite: Nothing to disclose \nNerijus Masiulis: Nothing to disclose \nSaulius Lukoševičius: Nothing to disclose \nRymantė Gleiznienė: Nothing to disclose \nUrte Lukoseviciute: Nothing to disclose \nOron Levin: Nothing to disclose \n \n \nAge-related hearing loss may be associated with sma ll vessel disease. \nPeak skeletonized mean diffusivity and TBSS study \nB. Genç, *I. C. Koc*, K. Aslan; Samsun/TR \n \nPurpose or Learning Objective: Peak skeletonized mean diffusivity (PSMD) \nis being proposed as a novel biomarker for small ve ssel disease. Tract-Based \nSpatial Statistics (TBSS) is a well-established DTI  analysis method that \nenables the fully automated detection of microstruc tural changes in white \nmatter. The aim of this study is to investigate whi te matter changes in patients \nwith age-related hearing loss using both PSMD and T BSS. \nMethods or Background: All individuals from the OpenNeuro hearing loss \nconnectome dataset were included in the study (http s://openneuro.org/ \ndatasets/ds005026). The dataset consisted of 52 hea ring loss (HL) patients \nand 30 healthy controls. From the DTI images in the  dataset, FA, MD, RD, and \nAD were obtained using FSL with preprocessing steps  including TOPUP and \neddy. The standard TBSS procedure was applied to in vestigate group \ndifferences in FA, MD, RD, and AD. The standard PSM D method with \nhistogram analysis was used to calculate the differ ence between the 95th and \n5th percentile MD values (https://www.psmd-marker.c om/). A comparison \nbetween the groups was made. \nResults or Findings: The TBSS analysis did not show any differences \nbetween the groups for any of the DTI parameters. T he PSMD value in the HL \ngroup was 225,21±23,50 x 10^-6 mm²/s, while in the control group it was \n214,53±24,01 x 10^-6 mm²/s, showing a statistically significant increase in the \nhearing loss group (p=0,039). \nConclusion: Our findings suggest that small vessel disease may underlie the \npathophysiology in patients with age-related hearin g loss. To our knowledge \nthis is the first study to show the relationship be tween age-related hearing loss \nand small vessel disease. \nLimitations: Since numerical data of hearing tests were not avai lable, the \ncorrelation between PSMD data and hearing test coul d not be analyzed. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: OpenNeuro permits the use of \npatient data under the CC0 license. Ethical approva l has already been \nobtained by the \"University of Salerno.\" Therefore,  no separate ethical approval \nhas been obtained from our institution. \nAuthor Disclosures:  \nBariş Genç: Nothing to disclose \nKerim Aslan: Nothing to disclose \nIrem Ceren Koc: Nothing to disclose \n \n \nGlymphatic Dysfunction Correlate with Spatial Navig ation Deficits in \nSubjective Cognitive Decline: Insights from 5.0T MR I and Plasma \nBiomarkers Analysis \nF. Chen¹, B. Zhang¹, Q. Chen¹, X. Fan², L. Zou², Y.  Li², G. Cheng², *G. Danni*¹; \n¹Nanjing/CN, ²Shenzhen/CN \n(gedniii@163.com) \n \nPurpose or Learning Objective: To assess the feasibility of analysis along \nthe perivascular space (ALPS) using DTI obtained fr om 5.0 Tesla MR, assess \nglymphatic system function in SCD and its correlati on with spatial navigation \nabilities and plasma biomarkers. \nMethods or Background: Glymphatic dysfunction is implicated in cognitive \nimpairment associated with AD. Spatial navigation i mpairments are among the \nearliest manifestations in individuals with SCD. Ho wever, the relationship \nbetween glymphatic function and spatial navigation remains poorly understood. \nBetween May 2023 and January 2024, 62 SCD patients and 62 matched \ncontrols underwent high-resolution DTI on 5.0T MR s canner, spatial navigation \nbehavioral tests, cognitive assessments, and Simoa plasma biomarker \nanalyses. The ALPS index reflecting glymphatic acti vity was calculated by a \nratio of the diffusivities along the x-axis in the projection and association neural \nfibers to the diffusivities perpendicular to them a nd compared according to the \ngroups with use of multivariate analysis of varianc e. Inter-reproducibility of \nALPS index among 5.0TMR and 3.0TMR scanners was eva luated using \nconsistency interclass correlation coefficient. Pea rson correlation analysis was \nused to assess the relationship between cognitive p erformance, spatial \nnavigation performance, plasma biomarkers, and the ALPS index. \nResults or Findings: The ALPS on 5.0TMR and ALPS on 3.0TMR showed \nstrong consistency and correlation. SCD patients ha d significantly lower ALPS \nindex on 5.0TMR and higher navigation errors compar ed to controls. The \nALPS index was positively correlated with spatial n avigation, cognitive \nperformance, and memory performance, and negatively  correlated with plasma \npTau217 levels. \nConclusion: The glymphatic function is impaired in SCD at the p reclinical AD \nstage, which may represent one of the physiological  mechanisms leading to \ndeficits in spatial navigation abilities. DTI-ALPS on 5.0T MR may serve as a \nsensitive neuroimaging biomarker for the preclinica l stage of AD. \nLimitations: Our study is a small-sample cross-sectional study. \nFunding for this study: National Science and Technology Innovation 2030 -- \nMajor program of \"Brain Science and Brain-Like Rese arch\" (2022ZD0211800) \nEthics committee - additional information: The Research Ethics \nCommittees of Nanjing Drum Tower Hospital, the Shen zhen Institute of \nAdvanced Technology of the Chinese Academy of Scien ces, and Peking \nUniversity Shenzhen Hospital. \nAuthor Disclosures:  \nQian Chen: Nothing to disclose \nYe Li: Nothing to disclose \nXiang Fan: Nothing to disclose \nGuanxun Cheng: Nothing to disclose \nBing Zhang: Nothing to disclose \nGe Danni: Nothing to disclose \nFutao Chen: Nothing to disclose \nLixian Zou: Nothing to disclose \n \n \nStatic and Dynamic Functional Connectivity Alternat ions of Medial and \nLateral Entorhinal Cortex with Subjective Cognitive  Decline \n*G. Danni*, Z. Bing; Nanjing/CN \n(gedniii@163.com) \n \nPurpose or Learning Objective: To investigate the static functional \nconnectivity (sFC) and dynamic functional connectiv ity (dFC) of medial \nentorhinal cortex (MEC) and lateral entorhinal cort ex (LEC) in individuals with \nsubjective cognitive decline (SCD) and the associat ions with cognitive \nperformance, spatial navigation and olfactory memor y. \nMethods or Background: Seventy-seven control subjects and 106 SCD \nindividuals were enrolled, and neuropsychological e valuations, 2D \ncomputerized spatial navigation test, olfactory mem ory test and resting-state \nfunctional magnetic resonance imaging (rs-fMRI) wer e collected. Bilateral MEC \nand LEC were selected as seeds to investigate alter nations of the volumes, \nsFC and dFC. \nResults or Findings: Compared to control subjects, SCD individuals exhib ited \ndecreased sFC between bilateral LEC and visual netw ork, between right LEC \nand left posterior cingulate gyrus and sensory moto r network, and between \nright MEC and left hippocampus, visual network and sensory motor network. \nThe dFC between right LEC and right triangular part  of inferior frontal gyrus \n(IFGtriang) decreased, while dFC between left MEC a nd right putamen, and \nbetween right MEC and right middle temporal gyrus i ncreased. In SCD group, \nvolumes of bilateral MEC were positively correlated  with spatial navigation \nability, and sFC between bilateral LEC and visual n etwork was positively \ncorrelated with olfactory memory. The dFC between r ight LEC and right \nIFGtriang was correlated positively with global cog nitive performance. The \ncombination of sFC and dFC as biomarkers to identif y SCD showed an area \nunder curve of 92.1%. \nConclusion: There were functional alternations of EC subregions  in SCD \nindividuals, and we demonstrated the association be tween LEC and spatial \nnavigation, and MEC and olfactory memory. The combi nation of sFC and dFC \nmay be a new neuroimaging biomarker for the early d iagnosis of AD. \nLimitations: The study lacked genetic and biomarker data. Also, we didn't \nhave follow-up data to track pathological progressi on. \nFunding for this study: National Science and Technology Innovation \n2030_Major program of \"Brain Science and Brain-Like  Research\" (No. \n2022ZD0211800) \nEthics committee - additional information: Nanjing Drum Tower Hospital \nEthics Committee \nAuthor Disclosures:  \nGe Danni: Author: Department of Radiology, the Affi liated Drum Tower \nHospital of Nanjing University Medical School \nZhang Bing: Nothing to disclose \n \n \nAutomated cerebral microhemorrhage detection on T2*  GRE for \nAlzheimer’s disease screening \nS. Van Eyndhoven, *R. Magalhaes*, R. Khan, T. V. Ph an, A. Liseune, A. Brys, \nD. M. Sima, J. Verheyden, A. Ribbens; Leuven/BE \n(ricardo.magalhaes@icometrix.com) \n \nPurpose or Learning Objective: Develop a robust automated deep learning-\nbased method for assessment of cerebral microhemorr hages on T2* gradient-\necho (GRE) images. \nMethods or Background: Hypointensities on GRE images can be indicative of \ncerebral microhemorrhages, and serve as exclusionar y criteria for anti-amyloid \n\n \n \nAbstract-based Programme \n \n 46  \nWednesday \ntherapies for Alzheimer’s disease, as they are link ed to increased risk of \nintracerebral hemorrhage. Automated detection tools  could greatly aid \nradiologists in the challenging task of accurately quantifying these findings. A \ndeep learning model was developed to detect microhe morrhages on cross-\nsectional GRE images. Training was done using 247 2 D GRE images and \naccompanying manual microhemorrhage annotations fro m the EMERGE \nclinical trial (NCT02484547). Detection accuracy wa s evaluated on a stratified \nsubset of cases (N=600) of the Alzheimer’s Disease Neuroimaging Initiative, \nwhere up to 10 microhemorrhages were annotated on a  2D GRE sequence by \nexperts. This is representative of the population t hat would be screened via \nMRI before administration of anti-amyloid treatment , for which assessment is \nmore challenging relative to cases with a large num ber of microhemorrhages. \nFor each case, we evaluated the F1 score, i.e., the  harmonic mean between \ndetection sensitivity and positive predictive value , and the absolute error \nbetween the number of microhemorrhages according to  the expert ground truth \nand the automated count. \nResults or Findings: The median F1 score of the trained model was 0.67, and \nits median absolute count error was 1 microhemorrha ge. \nConclusion: Detection of microhemorrhages on 2D GRE images is a  \nchallenging task, which is gaining importance with the advent of novel anti-\namyloid treatments that may lead to hemorrhagic sid e effects. Using a robust, \nvalidated AI tool, as described here, can assist ra diologists in \nmicrohemorrhage detection, providing value especial ly in sparse MR images, \nthough expert assessment remains vital. \nLimitations: None. \nFunding for this study: N/A \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nArne Brys: Employee: icometrix \nRicardo Magalhaes: Employee: icometrix \nAnnemie Ribbens: Employee: icometrix Shareholder: i cometrix \nRafay Khan: Employee: icometrix \nDiana M. Sima: Employee: icometrix \nThanh Vân Phan: Employee: icometrix \nSimon Van Eyndhoven: Employee: icometrix \nArno Liseune: Employee: icometrix \nJan Verheyden: Shareholder: icometrix Employee: ico metrix \n \n \n13:00-14:30 Research Stage 4 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 401 \nEstimation of liver fat and stiffness with \nimaging \n \nModerator \nJ. M. Lee; Seoul/KR  \n(jmlshy2000@gmail.com) \n \n \nThe Diagnostic Accuracy of Quantitative Ultrasound with Fat Fraction \nparameter for the Assessment of Hepatic Steatosis i n Patients with \nMetabolic-Dysfunction Associated Fatty Liver Diseas e \nR. Cannella, A. A. Blandino, A. Tulone, S. Petta, * T. V. Bartolotta*; Palermo/IT \n(tv_bartolotta@yahoo.com) \n \nPurpose or Learning Objective: To investigate the performance of \nquantitative ultrasound with US Fat Fraction (USFF)  in patients with metabolic \ndysfunction-associated steatotic liver disease (MAS LD). \nMethods or Background: This study included consecutive patients with \nMASLD who prospectively underwent MRI and quantitat ive ultrasound on the \nsame day. MRIs were acquired on a 3T scanner and th e fat fraction \nquantification was obtained with mDixonQuant sequen ce. Quantitative \nultrasound consisted of Tissue Attenuation Imaging (TAI), Tissue Scatter-\ndistribution Imaging (TSI), and US Fat Fraction (US FF) acquired by two \noperators (a radiologist and a radiology resident) to evaluate the inter-reader \nreliability. Spearman's rank-order correlation was calculated between USFF \nand MRI fat fraction. The diagnostic performance wa s investigated with the \narea under the receiver operating characteristics c urve (AUC), sensitivity and \nspecificity according to the optimal cutoff. \nResults or Findings: Fifty-nine patients (40 males, median age of 60 yea rs) \nwere enrolled. Among them, 47 (79.7%) had grade ≥1 steatosis and 23 (39.0%) \nhad grade≥2 steatosis. There was a high positive correlation between USFF \nand MRI fat fraction (rho: 0.877, p<0.001). For the  diagnosis of grade≥1 \nsteatosis the AUCs were 0.959 (95%CI: 0.778, 0.999)  for TAI, 0.847 (95%CI: \n0.631, 0.963) for TSI, and 0.988 (95%CI: 0.825, 1.0 00) for USFF. An USFF \n>10.0 had a sensitivity 94.1% and a specificity of 100% for grade≥1 steatosis. \nFor the diagnosis of grade≥2 steatosis, the AUCs were 0.958 (95%CI: 0.777, \n0.999) for TAI, 0.858 (95%CI: 0.757, 0.916) for TSI , and 0.958 (95%CI: 0.777, \n0.999) for USFF. An USFF >11.8 had a sensitivity 10 0% and a specificity of \n83.3% for grade≥2 steatosis. The reproducibility of USFF was excell ent (ICC of \n0.98; 95%CI: 0.95, 0.98). \nConclusion: USFF provides an excellent performance and reproduc ibility for \nthe quantification of hepatic steatosis in patients  with MASLD. \nLimitations: Lack of liver biopsy. \nFunding for this study: Study supported by Samsung. \nEthics committee - additional information: All participants provided written \ninformed consent. \nAuthor Disclosures:  \nAdele Tulone: Nothing to disclose \nAntonino Andrea Blandino: Nothing to disclose \nRoberto Cannella: Research/Grant Support: Co-fundin g by the European \nUnion - FESR or FSE, PON Research and Innovation 20 14-2020 - DM \n1062/2021; research collaboration with Siemens Heal thineers. Other: Support \nfor attending meetings from Bracco and Bayer \nSalvatore Petta: Nothing to disclose \nTommaso Vincenzo Bartolotta: Nothing to disclose \n \n \nMultiparametric ultrasound evaluation of hepatic fi brosis and steatosis in \npatients treated with bempedoic acid: a comparative  study on 100 \npatients \n*G. Daccordi*, P. Sacco, F. Lazzeretti, G. De Filip po, C. Tucci, M. A. Mazzei; \nSiena/IT \n(g.daccordi@student.unisi.it) \n \nPurpose or Learning Objective: This study evaluates the effectiveness of \nbempedoic acid in improving hepatic steatosis using  multiparametric \nultrasound and serum lipid levels, comparing result s with a control group. \nMethods or Background: Background: Hepatic steatosis is associated with \ndyslipidemia and liver dysfunction. Multiparametric  ultrasound, using \ntechniques like shear-wave elastography and attenua tion imaging (ATI), offers \na non-invasive assessment of liver fibrosis and ste atosis. Bempedoic acid is a \npotential therapy to reduce liver fat and improve l ipid profiles. Materials and \nmethods: We enrolled 100 patients with hepatic stea tosis: 50 received \nbempedoic acid therapy, and 50 served as controls. Ultrasound was performed \nusing the Canon Aplio i800, evaluating fibrosis via  shear-wave elastography, \nand steatosis using ATI and dispersion imaging. Two  operators with different \nexperience levels performed the exams at baseline a nd after treatment. Serum \ncholesterol, triglycerides, and liver enzymes were also monitored. \nResults or Findings: Significant improvements in fibrosis, steatosis, an d \ndispersion were observed in the treatment group (p < 0.001). No significant \nchanges were found in the control group (p > 0.05),  highlighting the utility of \nmultiparametric ultrasound for follow-up. Good inte r-operator agreement was \nnoted for ATI (ICC = 0.96), with slight discordance  in fibrosis evaluation (ICC = \n0.88), likely due to experience differences. \nConclusion: Multiparametric ultrasound plays a fundamental role  for the non-\ninvasive follow-up of hepatic fibrosis and steatosi s during therapy. This \napproach is a valuable tool for assessing patient r esponse, aiding in informed \nclinical decisions. \nLimitations: Ongoing single-center study \nFunding for this study: Not addictional funding \nEthics committee - additional information: University of Siena \nAuthor Disclosures:  \nMaria Antonietta Mazzei: Nothing to disclose \nCristina Tucci: Nothing to disclose \nFrancesco Lazzeretti: Nothing to disclose \nPalmino Sacco: Nothing to disclose \nGiorgia Daccordi: Nothing to disclose \nGiovanna De Filippo: Nothing to disclose \n \n \nFunctional Liver Imaging Score: A biomarker for pre diction of acute-on-\nchronic liver failure \n*A. Kristic*, L. Balcar, A. Ba-Ssalamah, T. Reiberg er, M. Mandorfer,  \nN. Bastati-Huber, R. Ambros, L. Beer, S. Pötter-Lan g; Vienna/AT \n \nPurpose or Learning Objective: The Functional Liver Imaging Score (FLIS) \nderived from gadoxetic acid (GA)-enhanced MRI is a prognostic biomarker in \npatients with advanced chronic liver disease (ACLD) . The aim of this study was \nto investigate whether FLIS, as well as quantitativ e imaging parameters, \nincluding the relative liver enhancement (RLE), rel ative enhancement ratio of \nthe biliary system (REB), and liver-to-portal vein contrast ratio (LPC), can \npredict acute-on-chronic liver failure (ACLF; a syn drome defined by \nextrahepatic organ failure and high short-term mort ality) in patients with an \nacute decompensation (AD) of cirrhosis, i.e., the m ain at-risk population. \n\n \n \nAbstract-based Programme \n \n 47  \nWednesday \nMethods or Background: We included 322 ACLD patients with GA-MRI-\nderived, semi-quantitative FLIS, in whom the RLE, R EB, and LPC were also \ncomputed by two independent radiologists. Patients were stratified into stable \nACLD (compensated or non-hospitalized/electively ho spitalized/non-liver-\nrelated hospitalized decompensated ACLD patients) a nd acutely \ndecompensated (AD) patients (non-elective liver-rel ated hospitalization). The \npredictive values of semi-quantitative FLIS and qua ntitative GA-MRI \nparameters for ACLF development were investigated b y Cox regression \nanalyses. \nResults or Findings: The FLIS was lower in AD (vs. stable ACLD) patients . \nFurthermore, low FLIS was an independent risk facto r for ACLF \ndevelopment/liver-related death in AD patients (adj usted hazard ratio [aHR]: \n2.26; 95%CI: 1.08-4.71; P=0.03), as well as in clin ically stable ACLD patients \n(aHR: 2.35; 95%CI: 1.21-4.55; P=0.01). Conversely, while RLE, REB, and LPC \ndistinguished between AD and clinically stable ACLD  patients (P<0.001), they \nfailed to predict ACLF/liver-related death. \nConclusion: The FLIS is a simple prognostic imaging biomarker i n AD patients \nin whom ACLF risk stratification is important to id entify patients who may \nbenefit from intensified monitoring or timely liver  transplant evaluation. \nLimitations: The retrospective design could have led to a select ion bias; \nhowever, the study allowed for a long clinical foll ow-up and a high number of \nendpoints. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Received approval EK \n2023/2017. \nAuthor Disclosures:  \nAntonia Kristic: Nothing to disclose \nSarah Pötter-Lang: Nothing to disclose \nLorenz Balcar: Speaker: received speaker fees from Chiesi, and Gilead. \nRaphael Ambros: Nothing to disclose \nMattias Mandorfer: Advisory Board: served as a spea ker and/or consultant \nand/or advisory board member for AbbVie, Collective  Acumen, Echosens, \nGilead, Ipsen, Takeda, and W. L. Gore & Associates.  Other: received travel \nsupport from AbbVie and Gilead. Consultant: served as a speaker and/or \nconsultant and/or advisory board member for AbbVie,  Collective Acumen, \nEchosens, Gilead, Ipsen, Takeda, and W. L. Gore & A ssociates. Grant \nRecipient: received grants from Echosens. Speaker: served as a speaker \nand/or consultant and/or advisory board member for AbbVie, Collective \nAcumen, Echosens, Gilead, Ipsen, Takeda, and W. L. Gore & Associates. \nLucian Beer: Speaker: received speaker fees from Ta keda, and Lilly. \nAhmed Ba-Ssalamah: Speaker: received honoraria for lectures and a \nconsultancy from Bayer without relation to the pres ent article. Consultant: \nreceived honoraria for lectures and a consultancy f rom Bayer without relation \nto the present article. \nNina Bastati-Huber: Nothing to disclose \nThomas Reiberger: Speaker: speaking honoraria from AbbVie, Gilead, W. L. \nGore & Associates, Intercept, Roche, and MSD. Grant  Recipient: received \ngrant support from AbbVie, Boehringer-Ingelheim, Gi lead, Intercept, MSD, Myr \nPharmaceuticals, Philips Healthcare, Pliant, Siemen s, and W. L. Gore & \nAssociates. Other: and travel support from AbbVie, Boehringer-Ingelheim, \nGilead, and Roche. Consultant: consulting/advisory board fees from AbbVie, \nBayer, Boehringer-Ingelheim, Gilead, Intercept, MSD , and Siemens; and travel \nsupport from AbbVie, Boehringer-Ingelheim, Gilead, and Roche. Advisory \nBoard: consulting/advisory board fees from AbbVie, Bayer, Boehringer-\nIngelheim, Gilead, Intercept, MSD, and Siemens. \n \n \nA compound model improves the accuracy of ultrasoun d-estimated fat-\nfraction \n*P. N. Kaposi-Novák*, B. Zsély, Z. Zsombor, M. Hims el, V. Bérczi, G. Györi,  \nP. Maurovich-Horvat, A. D. Rónaszéki; Budapest/HU \n \nPurpose or Learning Objective: The ultrasound-estimated fat-fraction (UEFF) \nis new biomarker that can facilitate the diagnosis and follow-up of liver \nsteatosis. Different models have been devised to ca lculate UEFF, but these \nhave yet to be compared side-by-side. \nMethods or Background: We retrospectively collected ultrasound parameters \nfrom sixty patients with various grades (S0-S3) of metabolic dysfunction-\nassociated steatotic liver disease (MASLD), includi ng attenuation coefficient \n(AC), backscatter-distribution coefficient (BSC), a nd liver capsule-to-skin \ndistance (CSD). The training set had balanced distr ibution of steatosis grades \n(S0, S1, S3 - 16 cases each, S2 - 12 cases). Univar iable and multivariable \nlinear and exponential models were trained to predi ct the MRI proton density \nfat-fraction (PDFF) using repeated cross-validation . The models were tested on \nfifty cases for which the scanner’s application cal culated an ultrasound fat-\nfraction (USFF). We compared the R-squared (R2) and  the one-way random \neffect intraclass correlation coefficients (ICC) am ong the models. \n \n \n \n \nResults or Findings: In low-grade steatosis (≤ S1), the linear model using AC \nand BSC achieved the best fit (R2= 0.379, p<0.002),  and USFF had the best \nagreement (ICC=0.538, p<0.003) with PDFF. In high-g rade steatosis (≥ S2), \nthe ASC and BSC-based multi-exponential model perfo rmed best (R2= 0.252, \np<0.008, ICC=0.422, p<0.012). The fit could be impr oved by including CSD in \nthe model (R2=0.263, p<0.007). The univariable nonl inear AC model resulted \nin a slightly weaker agreement in both low-grade (I CC=0.236, p<0.019) and \nhigh-grade (ICC=0.227, p<0.012) steatosis. A compou nd model using linear \nregression in low-grade or nonlinear regression in high-grade steatosis \noutperformed (R2= 0.6, p<0.001, ICC=0.849, p<0.001)  other models in \npredicting the test cases. \nConclusion: The UEFF is a robust method to diagnose liver steat osis across \nall stages. The compound model substantially improv es the fit and the \nagreement between UEFF and PDFF. \nLimitations: Single center study. \nFunding for this study: Pál Novák Kaposi was recipient of a research grant \nfrom Samsung Medison Ltd. (Contract number: SE4K/20 23/195). \nZita Zsombor was recipent of a resarch scholarship from the New National \nExcellence Program (ÚNKP-23-3-I-SE-23) of the Hunga rian Ministry of Culture \nand Innovation. \nEthics committee - additional information: Semmelweis University’s \nRegional and Institutional Science and Research Eth ics Committee (Protocol \nnumber: SE RKEB 140/2020, 16 July 2020, and SE RKEB  6/2023, 9 February \n2023) \nAuthor Disclosures:  \nPál N. Kaposi-Novák: Research/Grant Support: Samsun g Medison Ltd. \nAladár David Rónaszéki: Nothing to disclose \nPál Maurovich-Horvat: Nothing to disclose \nBoglárka Zsély: Nothing to disclose \nMarco Himsel: Nothing to disclose \nViktor Bérczi: Nothing to disclose \nZita Zsombor: Nothing to disclose \nGabriella Györi: Nothing to disclose \n \n \nDual-Energy CT Liver Fat Quantification as Imaging Biomarker of \nMortality and Morbidity in Intensive-Care Patients \n*J. Erley*, J. Breckow, K. Roedl, A. Duoerkongjiang , G. De Heer, E. Tahir,  \nJ. Yamamura, G. Adam, I. Molwitz; Hamburg/DE \n \nPurpose or Learning Objective: This study aimed to evaluate the association \nbetween liver fat content, assessed using dual-ener gy computed tomography \n(DECT) material decomposition, in immobilized inten sive care unit (ICU) \npatients with in-hospital mortality, length of ICU stay, and indicators of \nmorbidity (need for tracheotomy and renal replaceme nt therapy). \nMethods or Background: ICU patients who received a DECT upon ICU \nadmission between November 2019 and December 2022 w ere retrospectively \ninvestigated. DECT liver fat fraction (DECT-FF) was  determined by material \ndecomposition for fat, liver tissue, and iodine by combining two regions of \ninterest (ROI) in the right and one ROI in the left  liver lobe (min. size 3.6 cm2). \nCox proportional hazard models were employed, inclu ding DECT-FF, sex, age, \nbody mass index, ICU scoring systems for disease pr ediction, reason for \nadmission, pre-existing malignancies, chronic disea ses, and inflammatory \ndiseases. \nResults or Findings: In total 76 patients were included (33 female, mean  age \n61±12 years, of which 59% died in the hospital. DEC T-FF at CT1 was \n3.3±5.4%. DECT-FF was associated with in-hospital m ortality (hazard ratio \n1.09 [95% confidence interval 1.03; 1.15], p=0.004) , with the length of ICU-stay \n(odds ratio (OR) -4.28 [-6.64; -1.92], p=0.001) and  with the need for a \ntracheotomy (OR: 0.90 [0.80; 0.99], p=0.039). No as sociation was observed \nbetween DECT-FF and ICU scoring systems or renal re placement therapy. \nConclusion: A higher liver DECT-FF upon ICU admission was assoc iated with \nin-hospital mortality and, probably as a consecutiv e effect/bias, with a \ndecreased length of ICU stay and a lower likelihood  of tracheotomy in surviving \npatients. The liver DECT-FF may serve as a predicti ve imaging biomarker of \nmortality in critically ill patients. \nLimitations: The sample size and its heterogeneity. \nFunding for this study: Not applicable. \nEthics committee - additional information: The study has been approved by \nthe ethics committee of the medical association in Hamburg. \nAuthor Disclosures:  \nGerhard Adam: Nothing to disclose \nIsabel Molwitz: Nothing to disclose \nAlidan Duoerkongjiang: Nothing to disclose \nGeraldine De Heer: Nothing to disclose \nKevin Roedl: Nothing to disclose \nJin Yamamura: Nothing to disclose \nJennifer Erley: Nothing to disclose \nJulia Breckow: Nothing to disclose \nEnver Tahir: Nothing to disclose \n \n \n\n \n \nAbstract-based Programme \n \n 48  \nWednesday \nAccuracy of estimates of liver fat content based on  2point-Dixon in \ncomparison to multi-echo-Dixon sequences in a popul ation-based cohort \n*M-N. Von Itter*¹, T. Nonnenmacher², T. Norajitra²,  S. Rospleszcz¹,  \nJ. Machann³, F. Bamberg¹, J. Nattenmüller¹; ¹Freibu rg/DE, ²Heidelberg/DE, \n³Tübingen/DE \n \nPurpose or Learning Objective: The rising prevalence of metabolic \ndysfunction-associated steatotic liver disease (MAS LD), formerly NAFLD, \nmakes it a leading liver disease and a risk factor for steatohepatitis, cirrhosis, \nand hepatocellular carcinoma. As it is potentially reversible, early detection is \nkey. \nMethods or Background: A validated nnU-Net liver segmentation model \nprocessed 10,636 MRI scans from NAKO participants u sing 2-point- (2p) and \nmulti-echo- (me) Dixon data. Samples with significa nt mask mismatches were \nexcluded. The segmentation masks were used to calcu late mean liver fat \ncontent (LFC) from 2p-Dixon data, which was then co mpared to the me-Dixon \nproton density fat fraction. \nResults or Findings: The overall mean difference between the fat fractio n \nfrom me-Dixon measurements and the 2p-Dixon fat est imation is -\n1.101±0.012%. 95% of the measured differences are within the range of -3.8% \nand 1.4%. Mean LFCs from 2p-Dixon data overestimate  the fat content \ncompared to me-Dixon, especially for higher fat dep osition. Liver iron content \ndid not lead to a systemic offset in our cohort, th ough only participants with \nmild to moderate liver iron content were part of ou r sample. \nConclusion: As the 2p-Dixon sequence is often acquired for vari ous clinical \nquestions beyond specific liver imaging, assessment  of 2p-Dixon-based LFC \ncan identify individuals with MASLD in possible scr eening programs or MRI \nstudies for other reasons and guide them to the com plementary acquisition of a \nme-Dixon sequence. This could benefit individuals b y initiating preventive \nmeasures and reduce the socio-economic burden of MA SLD caused by its \ncomplications. \nLimitations: No histological confirmation was performed. \nFunding for this study: Funding was received from the German Research \nFoundation (Deutsche Forschungsgemeinschaft, grant number: 428224476). \nThe NAKO is funded by the Federal Ministry of Educa tion and Research \n(BMBF) [project funding reference numbers: 01ER1301 A/B/C and \n01ER1511D], the federal states and the Helmholtz As sociation, with additional \nfinancial support by the participating universities  and the institutes of the \nLeibniz association. \nEthics committee - additional information: The study received institutional \nreview board approval and written informed consent was obtained from all \nparticipants. \nAuthor Disclosures:  \nTobias Nonnenmacher: Nothing to disclose \nSusanne Rospleszcz: Nothing to disclose \nJohanna Nattenmüller: Nothing to disclose \nJürgen Machann: Nothing to disclose \nFabian Bamberg: Nothing to disclose \nMarc-Nicolas Von Itter: Nothing to disclose \nTobias Norajitra: Nothing to disclose \n \n \nDeep learning-based liver volume and fat fraction q uantification from \nDixon-MRI: Reference curves from over 66,000 indivi duals and their \nprognostic value \n*M. Jung*¹, L. Michel², M. Reisert², S. Jäck², S. R ospleszcz², M. T. Lu¹,  \nF. Bamberg², V. Raghu¹, J. Weiß²; ¹Boston, MA/US, ² Freiburg/DE \n(matthias.jung@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: Steatotic liver disease (SLD) is a major \npublic health concern with a global prevalence of 3 2.4% and an independent \nrisk factor for cardiometabolic and liver disease. We used a deep learning \nframework to quantify liver volume and SLD from MRI  in a large Western \nEuropean population to calculate reference curves a nd investigate their \nprognostic value. \nMethods or Background: We developed a deep-learning model that takes an \nMRI as input and outputs liver volume (L) and fat f raction (FF, %) using data \nfrom the UK Biobank (UKBB) and German National Coho rt (NAKO). \nEstablished FF-thresholds were used to define mild( 5%), moderate(15%), and \nsevere(25%) SLD. We computed age-, sex-, and height -normalized reference \ncurves and assessed the prognostic value of liver v olume z-score (z<1; z=1-2; \nz>2) and SLD-categories for incident outcomes (diab etes; liver disease; all-\ncause mortality) in the UKBB (n=35,002). Cox regres sion assessed the \nassociation between volume z-score and SLD categori es with outcomes after \nadjustment for age, sex, BMI, race, and cardiometab olic risk factors (serum \nglucose, Hb1Ac, lipid panel, prevalent hypertension , history of cancer, alcohol \nconsumption, smoking status). \n \n \n \n \nResults or Findings: Among 66,664 individuals from the general populatio n \n(57.7±12.9 years; BMI: 26.2±4.5 kg/m2, 48.3% female), SLD was high with a \nprevalence of 80.4%. In the UKBB, multivariable-adj usted Cox regression \nshowed that severe steatotic liver disease (SLD) wa s associated with an \nincreased risk of incident diabetes (adjusted hazar d ratio [aHR] 2.66) and liver \ndisease (aHR 6.34) compared with no SLD. A liver vo lume z-score >2 was \nassociated with higher all-cause mortality (aHR 2.2 5) compared with a \nz-score <1. \nConclusion: SLD and normalized liver volume categories predicte d outcomes \nbeyond traditional risk factors. We will release op en-source reference curves to \nenhance clinical liver risk assessment and improve comparability in research. \nLimitations: Predominantly white population. \nFunding for this study: This project was conducted with data from the \nGerman National Cohort (NAKO) (www.nako.de). The NA KO is funded by the \nFederal Ministry of Education and Research (BMBF) [ project funding reference \nnumbers: 01ER1301A/B/C, 01ER1511D, and 01ER1801A/B/ C/D], federal \nstates of Germany, and the Helmholtz Association, t he participating \nuniversities and the institutes of the Leibniz Asso ciation. This research has \nbeen conducted using the UK Biobank Resource under Application Number \n80337. We thank all participants who took part in t he NAKO and UKBB study \nand the staff of these research initiatives. MJ was  funded by the Deutsche \nForschungsgemeinschaft (DFG, German Research Founda tion) - 518480401. \nVKR was funded by Norn Group Longevity Impetus Gran t, NHLBI \nK01HL168231, and AHA Career Development Award 93517 6. \nEthics committee - additional information: Informed consent was obtained \nfrom all participants in the UK Biobank and the Ger man National Cohort study. \nIn addition, we received local IRB approval (IRB of  the University of Freiburg: \n23-1316-S1-retro and 24-1099-S1-retro). \nAuthor Disclosures:  \nSusanne Rospleszcz: Nothing to disclose \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nLea Michel: Nothing to disclose \nMatthias Jung: Nothing to disclose \nFabian Bamberg: Nothing to disclose \nSaskia Jäck: Nothing to disclose \nVineet Raghu: Nothing to disclose \nMichael T. Lu: Nothing to disclose \n \n \nIntra-individual quantitative crossover comparision  of liver fat \nmeasurements between free-breathing radial GRE and conventional \ncartesian GRE breath-hold methods \n*T. B. Rodrigues*¹, M. D. Santana¹, V. Hérida¹, N. Almeida¹, T. Castela¹,  \nR. C. Semelka², M. Ramalho¹; ¹Lisbon/PT, ²Chapel Hi ll, NC/US \n(teresabaratarodrigues@gmail.com) \n \nPurpose or Learning Objective: To evaluate and compare liver fat fraction \nmeasurements derived from a free-breathing radial s equence with those \nobtained from reference cartesian breath-hold techn iques in an intra-individual \nfashion. \nMethods or Background: The study included 40 subjects (19 males, 21 \nfemales; mean age 60.5 ± 13.2) who underwent MRI ex aminations.Three T1-\nweighted sequences were used: Cartesian 2D-FLASH du al gradient-echo(2D-\nGRE), Cartesian 3D-GRE with Dixon technique, and fr ee-breathing 3D-GRE \nwith radial data sampling(Radial 3D-GRE). Two indep endent readers \nmeasured the mean region of interest (ROI) values o f the liver for in-phase (IP) \nand out-of-phase (OP) images for each sequence with  an equal ROI and \nsimilar location. Quantitative liver fat fractions (FF) were calculated using \n((SI(IP)−SI(OP))/(2 × SI(IP))×100. The liver FF wer e compared across \nsequences. The inter-reader agreement was assessed using the intraclass \ncorrelation coefficient (ICC).Pearson correlation a nd regression analyses \nexamined relationships among different measurement techniques. P-value of \n<0.05 was considered significant. \nResults or Findings: The study found strong correlations between liver F F \nmeasurements across 2D-GRE, 3D-GRE, and radial 3D-G RE sequences. \nPearson correlation coefficients were 0.9788 for 2D -GRE vs.3D-GRE, 0.9506 \nfor 3D-GRE vs.Radial 3D-GRE, and 0.9478 for 2D-GRE vs.Radial 3D-GRE \n(p<0.0001). Regression analyses confirmed strong ag reement between \nmethods. ICC was 0.9489 (95%CI=0.9279 to 0.9637). \nConclusion: Our findings underscore the promising potential of the new IP/OP \nfree-breathing Radial 3D-GRE sequence as a reliable  alternative to traditional \nbreath-hold techniques for liver FF in patients who  cannot hold their breath. \nThe excellent correlation between Radial 3D-GRE and  the cartesian methods \nshould recommend its use for patients incapable of suspending breathing. \nLimitations: Measurement values may be affected by variations in  fat \ndistribution within the liver and small sample size .These factors may impact \naccuracy and reproducibility and introduce selectio n bias. However, identical \nROI and matched locations were used in all patients . \nFunding for this study: No funding was received for this study \n \n \n\n \n \nAbstract-based Programme \n \n 49  \nWednesday \nEthics committee - additional information: Retrospective study \nAuthor Disclosures:  \nVasco Hérida: Nothing to disclose \nMiguel Ramalho: Nothing to disclose \nTiago Castela: Nothing to disclose \nNuno Almeida: Nothing to disclose \nMariana Domingues Santana: Nothing to disclose \nRichard C. Semelka: Nothing to disclose \nTeresa Barata Rodrigues: Nothing to disclose \n \n \nMagnetic resonance elastography of the liver: are t he results reliable? \n*V. Atamaniuk*, M. Obrzut, L. Hanczyk, M. Cholewa, B. Obrzut; Rzeszów/PL \n(vitaliyacera500@gmail.com) \n \nPurpose or Learning Objective: Magnetic resonance elastography (MRE) \nhas become the gold standard for non-invasive asses sment of liver stiffness, \nespecially in patients with liver fibrosis, replaci ng liver biopsy. While guidelines \nfor MRE protocols and interpretation have been stan dardized by the \nQuantitative Imaging Biomarkers Alliance (QIBA), th e impact of vibration \namplitude and slice thickness on hepatic stiffness measurements remains \nuncertain. This study aims to evaluate the reliabil ity of MRE under varying \nvibration amplitudes and slice thicknesses. \nMethods or Background: Twenty volunteers (10 men, 10 women), aged 18–\n68, underwent MRE on a 1.5 T whole-body scanner usi ng a 2D GRE \nsequence. The Resoundant system delivered vibration s at 60 Hz with \namplitudes of 25%, 50%, 75%, and 100%. Standard sli ce thickness was 10 \nmm, and an additional scan was performed at 75% amp litude with a 5 mm \nslice thickness. Regions of interest (ROIs) were ma nually drawn per QIBA \nguidelines to ensure consistent location for stiffn ess measurements across \ntested conditions. Statistical analyses included re peated measures ANOVA \nand intraclass correlation coefficients (ICC). \nResults or Findings: The effects of both vibration amplitude (p = 0.11) and \nslice thickness (p = 0.69) on hepatic stiffness wer e not statistically significant. \nThe ICC for different amplitudes was 0.92, and for slice thickness, it was 0.95, \nindicating excellent agreement across conditions. \nConclusion: MRE provides reliable liver stiffness measurements,  with no \nsignificant influence from changes in vibration amp litude or slice thickness. \nGiven the excellent agreement across varying imagin g parameters, MRE can \nbe considered a stable and reproducible method for assessing liver stiffness. \nFurther multicentre studies with larger samples and  3D MRE may help confirm \nthese findings and expand clinical applications of this technique. \nLimitations: The study's limitations include the small sample si ze and the use \nof 2D MRE. \nFunding for this study: No external funding was obtained for this study. \nEthics committee - additional information: Approved by the Regional \nMedical Chamber ethics committee (Resolution No 60/ 2022/B); informed \nconsent was obtained from all participants. \nAuthor Disclosures:  \nMarzanna Obrzut: Nothing to disclose \nMarian Cholewa: Nothing to disclose \nLukasz Hanczyk: Nothing to disclose \nVitaliy Atamaniuk: Nothing to disclose \nBogdan Obrzut: Nothing to disclose \n \n \nFrequency and distribution of steatotic liver disea se in the NAKO study – \nmagnetic resonance imaging of 30,000 participants \n*M-N. Von Itter*¹, T. Nonnenmacher², E. Grune¹, J. Machann³, J. Weiß¹,  \nJ. Nattenmüller¹, T. Norajitra², S. Rospleszcz¹, N.  Consortium¹; ¹Freiburg/DE, \n²Heidelberg/DE, ³Tübingen/DE \n \nPurpose or Learning Objective: Steatotic liver disease (SLD) and its subtype, \nmetabolic dysfunction-associated steatotic liver di sease (MASLD), are risk \nfactors for cardiometabolic disease, liver cirrhosi s, and hepatocellular \ncarcinoma, and represent a major public health burd en. Using magnetic \nresonance imaging in Germany’s largest population-b ased study (NAKO), we \ndescribe the frequency of SLD and MASLD, and evalua te the distribution \naccording to sex, age, BMI, geographic region, and socio-economic status. \nMethods or Background: A validated nnU-Net liver segmentation model \nprocessed 29,842 MRI scans from NAKO participants ( 44.1% women) using \nT1-weighted 6-point Dixon data. The segmentation ma sks were used to \ncalculate mean liver fat content (LFC). SLD was def ined as LFC ≥ 5.56%, and \nMASLD according to established criteria. \nResults or Findings: Overall frequency of SLD and MASLD was 37.6% and \n31.8% in men, and 20.3% and 18.6% in women, respect ively. Frequency \nincreased with increasing BMI and age, with differe nt patterns in men and \nwomen. Geographically, the highest frequency of SLD  was found in Eastern \nGermany (40.6%). Frequency was higher in individual s with low socio-\neconomic status, and this difference was more prono unced in women \n(frequency for high vs. low socio-economic status: 35.7% vs. 47.2% in men, \nand 17.1% vs. 37.9% in women). \nConclusion: Frequency of SLD and MASLD in Germany is high, with  \npronounced differences according to sex, age, BMI, geographic region, and \nsocio-economic status. Our findings provide a robus t basis to estimate the \npublic health impact of these liver diseases in Ger many. \nLimitations: Ultrasound for diagnosis of SLD was not available. \nFunding for this study: Funding was received from the German Research \nFoundation (Deutsche Forschungsgemeinschaft, grant number: 428224476). \nThe NAKO is funded by the Federal Ministry of Educa tion and Research \n(BMBF) [project funding reference numbers: 01ER1301 A/B/C and \n01ER1511D], the federal states and the Helmholtz As sociation, with additional \nfinancial support by the participating universities  and the institutes of the \nLeibniz association. \nEthics committee - additional information: The study received institutional \nreview board approval and written informed consent was obtained from all \nparticipants. \nAuthor Disclosures:  \nTobias Nonnenmacher: Nothing to disclose \nSusanne Rospleszcz: Nothing to disclose \nJohanna Nattenmüller: Nothing to disclose \nElena Grune: Nothing to disclose \nJakob Weiß: Nothing to disclose \nJürgen Machann: Nothing to disclose \nNako Consortium: Nothing to disclose \nMarc-Nicolas Von Itter: Nothing to disclose \nTobias Norajitra: Nothing to disclose \n \n \nAssessment of using Multimodal Magnetic Resonance I maging (MRI) for \nNoninvasive Evaluation of type 2 diabetic \n*M. W. Yang*, W. J. Shao; KunMing/CN \n(yangmengweid@163.com) \n \nPurpose or Learning Objective: To investigate the capability of intravoxel \nincoherent motion (IVIM), diffusion kurtosis imagin g (DKI)and diffusion tensor \nimaging (DTI) to assess the renal function changes of type 2 diabetes. \nMethods or Background: Prospectively included 46 patients diagnosed with \nT2MD, and divided them into three groups based on e stimated glomerular \nfiltration rate (eGFR) and the presence of diabetic  nephropathy: the simple \ndiabetes group (DM), the early diabetic nephropathy  group (e-DKD), and the \nmiddle-to-late diabetic nephropathy group (m- DKD).  At the same time, 33 \nvolunteers（control group,CG） were recruited and underwent MRI \nexaminations to collect images from various sequenc es. The true diffusion \ncoefficient D, pseudo-diffusion coefficient D*, per fusion fraction f, mean \ndiffusion kurtosis MK, mean diffusivity MD, fractio nal anisotropy FA, and \napparent diffusion coefficient ADC of the renal cor tex and medulla of the \nsubjects were measured and statistically analyzed. \nResults or Findings: In the e-DKD and m-DKD groups, ACR increased while \neGFR decreased. The ACR of the e-DKD group signific antly differed from the \nDM group and CG (P<0.05). IVIM values, DKI (cortica l and medullary MD), and \nDTI (cortical and medullary ADC) values from all fo ur groups showed a \ndeclining trend with disease progression, while cor tical and medullary MK \nvalues from DKI showed an increasing trend. The cor tical MK value of the DM \ngroup significantly differed from the CG (P<0.05). Medullary MK value \neffectively distinguished the CG and e-DKD groups ( AUC=0.881, cutoff=0.593, \nsensitivity=76.9%, specificity=95.0%). \nConclusion: IVIM, DKI, DTI Sequences can be used to supplement renal \ndysfunction assessment. Different magnetic resonanc e parameters (IVIM, DKI, \nand DTI Sequences) identify different renal impairm ent changes in type 2 \ndiabetic patients. Cortical MK value has higher dia gnostic efficiency in the early \ndetection of renal damage in DM patients. \nLimitations: The sample size needs to be increased.Further multi center \nstudies are needed to supplement . \nFunding for this study: The \"SKY Imaging Research Fund\" by the China \nInternational Medical Exchange Foundation. \nEthics committee - additional information: Chinese Ethics Review \nNumber:2020143 \nAuthor Disclosures:  \nWei Ju Shao: Nothing to disclose \nMeng Wei Yang: Nothing to disclose \n \n \nPopulation-scale MRI body composition analysis: ass ociations between \nsingle-slice and volumetric measurements of muscle and adipose tissue \n*M. Nowak*¹, L. M. Nunez¹, C. Hill¹, S. Marriage¹, R. Salvati¹, M. Pansini²,  \nH. B. Thomaides-Brears¹, M. Robson¹; ¹Oxford/UK, ²L ugano/CH \n \nPurpose or Learning Objective: Accurate body composition tools are \nimportant for assessing adipose and muscle tissue i n both clinical and \nresearch settings, including obesity management, sa rcopenia, and weight loss \ntrials. The use of whole-body MRI is limited by cos t and processing demands, \nwhile single-slice MRI offers a more efficient alte rnative. This study evaluated \nthe correlations between single-slice and volumetri c assessments of visceral \n\n \n \nAbstract-based Programme \n \n 50  \nWednesday \nadipose tissue (VAT), subcutaneous adipose tissue ( SAT), and muscle tissue, \nand their associations with cardiometabolic risk fa ctors. \nMethods or Background: We analyzed data from a subset of 67,509 \nindividuals from the UK Biobank with water and fat MRI scans (mean age: 66, \n51% male, BMI 26.7). A single axial slice at the L3  vertebra was used to \nmeasure VAT, SAT, and skeletal muscle via semi-auto matic segmentations. \nThese were compared with volumetric assessments of VAT, SAT, total lean \ntissue, and thigh fat-free muscle. Correlation coef ficients were used to assess \nthe relationship between the two methods, and their  associations with \ncardiometabolic risk factors. \nResults or Findings: Single-slice L3 measurements of SAT and VAT \ndemonstrated very strong correlations with SAT (rho =0.94, p<0.001) and VAT \nvolume (rho=0.97, p<0.001), independent of sex, age , BMI, waist \ncircumference, diabetes status, and liver tissue ch aracteristics (SAT: median \nrho 0.93, VAT: median rho 0.96, all p<0.001). Both body composition \nassessments showed similar correlations with cardio metabolic risk factors (all \np<0.01 for HbA1c, triglycerides, high-density lipop rotein, systolic blood \npressure, liver cT1, and liver fat content). Strong  correlations were also \nobserved between single-slice skeletal muscle and b oth total lean tissue \n(r=0.90, p<0.001) and thigh fat-free muscle volume (r=0.91, p<0.001). \nConclusion: Single-slice L3 measurements of VAT, SAT, and muscl e CSA \nmetrics show robust correlations with volumetric as sessments across \nindividuals with diverse cardiometabolic profiles, while exhibiting comparable \nassociations with cardiometabolic risk factors. \nLimitations: N/A \nFunding for this study: Perspectum Ltd. \nEthics committee - additional information: In UK Biobank, ethical approval \nfor data collection was received from the North-Wes t Multi-centre Research \nEthics Committee and the research was carried out i n accordance with the \nDeclaration of Helsinki of the World Medical Associ ation. \nAuthor Disclosures:  \nLuis Miguel Nunez: Nothing to disclose \nRoberto Salvati: Nothing to disclose \nCharles Hill: Nothing to disclose \nMichele Pansini: Nothing to disclose \nMagdalena Nowak: Nothing to disclose \nHelena B Thomaides-Brears: Nothing to disclose \nMatthew Robson: Nothing to disclose \nScott Marriage: Nothing to disclose \n \n \n13:00-14:30 Room G1 \nResearch Presentation Session: \nRadiographers \nRPS 414 \nAI-driven evolution: enhancing image \nquality, workflow, and professional \nidentity for radiographers \n \nModerators \nF. Doo; Baltimore, MD/US  \n(fdoo@som.umaryland.edu) \nS. McFadden; Newtownabbey/UK \nAuthor Disclosures:  \nFlorence Doo: Equipment Support Recipients: Cloud c redits from Amazon \nAWS, Microsoft Azure, Google Cloud; Grant Recipient : Funded in part by \nAssociation for Academic Radiology (AAR) Clinical E ffectiveness in Radiology \nResearch Academic Fund (CERRAF) in part by GE Healt hcare; Research \nGrant/Support: Funded in part by the Johns Hopkins Mid-Atlantic Center for \nCardiometabolic Health Equity (MACCHE), which is su pported by National \nInstitutes of Minority Health and Health Disparitie s (P50MD017438); Speaker: \nHonoraria from Eli Lilly \n \n \nR-AI-diographers: a European survey to explore the perceived impact of \nAI on professional identity, careers, and roles of radiographers \nN. Stogiannos¹, *G. Walsh*¹, B. K. Ohene-Botwe¹, K.  Mchugh², B. Potts¹,  \nJ. St John-Matthews¹, M. F. Mcentee³, Y. Kyratsis ⁴, C. Malamateniou¹; \n¹London/UK, ²Portsmouth/UK, ³Cork/IE, ⁴Rotterdam/NL \n(gemma.walsh@outlook.com) \n \nPurpose or Learning Objective: Artificial intelligence is changing \nradiographer clinical practice and roles. It is the refore vital to understand its \nimpact on the careers, roles and professional ident ity of these professionals. \nMethods or Background: A European-wide, EFRS-endorsed, cross-sectional, \nmixed methods online survey was designed on qualtri cs. Snowball sampling \nwas used to improve uptake. Survey questions explor ed radiographer \nperceptions for the short-term and long-term impact  of AI implementation on \ntheir roles, responsibilities and professional iden tity. The study was translated \nin 8 languages. Responses were compared between dif ferent demographic \ngroups including gender, age, education and country  digital literacy level. \nResults or Findings: 2206 valid responses were received from 37 differen t \ncountries in Europe. 50.4% reported no AI education , and 26.6% were self-\ntaught in AI. Over half (51.1%) thought patient-cen tered care skills will remain \nthe same. 50.9% agreed radiographers will have more  time to spend with \npatients thanks to AI. 57.8% agreed radiographers w ill have to work closer with \nother MIRT professionals in the future, for efficie nt AI implementation. Men \nappeared slightly more enthused about the developme nt of technological skills \nand women about the honing of patient centered care  skills, similar to previous \nstudies. Radiographers were overall optimistic abou t the use of AI in \nhealthcare, and optimism was higher in those countr ies with high digital \nliteracy, better education levels and with more AI experience. \nConclusion: Radiographers were overall optimistic about the use  of AI in \nhealthcare and strongly believed that AI will advan ce patient-centred care. AI \neducation currently lags for European radiographers , and this should be \nacutely addressed at the scale and pace required to  keep up with current \ntechnological developments. Interprofessional colla boration was seen as \nessential for fostering mutual support among profes sionals. \nLimitations: Snowball sampling can lead to selection-bias, but a llows for many \nrecruits. \nFunding for this study: Funded by the College of Radiographers Industry \nPartnership Scheme (CoRIPS) [grant number: 2018]. \nEthics committee - additional information: Ethics approval was obtained \nfrom City St George’s, University of London School of Health and \nPsychological Sciences Ethics Committee (ETH2223-13 46). \nAuthor Disclosures:  \nGemma Walsh: Nothing to disclose \nMark F. Mcentee: Nothing to disclose \nYiannis Kyratsis: Nothing to disclose \nBenard Kwadwo Ohene-Botwe: Nothing to disclose \nNikolaos Stogiannos: Nothing to disclose \nJanice St John-Matthews: Nothing to disclose \nBen Potts: Nothing to disclose \nChristina Malamateniou: Nothing to disclose \nKevin Mchugh: Nothing to disclose \n \n \nAn investigation into radiographers' perception of quality control \nauditing of radiographic practice and the potential  role of Artificial \nIntelligence \n*L. A. Rainford*, M. Mujaydia Alotaibi, J. Mcnulty,  J. Potočnik; Dublin/IE \n(louise.rainford@ucd.ie) \n \nPurpose or Learning Objective: Quality assurance (QA) of radiographic \ntechnique is an essential part of radiation protect ion, traditionally performed \nthrough Reject Analysis. Digital imaging has increa sed the difficulty in \ncompleting radiographic technique auditing and staf f shortages further \ncompromise QA monitoring. This research aimed to se ek radiography opinion \non the use of Artificial Intelligence (AI) for QA. \nMethods or Background: An online survey was developed (n=30 questions) \nto seek information related to QA monitoring of rad iographic technique. \nParticipant demographics, including area of employm ent and country of work, \nand professional and AI experience, were captured. Current QA auditing \ndetails were requested and participant confidence i n these processes. Their \nopinion was requested on the potential challenges a nd benefits of AI use in QA \nmonitoring. The survey was distributed to affiliate  EFRS academic institutions \nto distribute to their clinical training sites and via Radiography social media. \nResults or Findings: Good representation across all radiography professi onal \ngrades was received from 125 participants (n=22 cou ntries). 19.8% reported \nQA of radiographic images on at least a weekly basi s, 18.8% stated monthly, \nwhilst 60% reported it occurred far less frequently . 20% of responses stated \nstaff were not individually reviewed. Only 26.8% we re very confident in current \nQA processes, 48% were somewhat confident and the r emainder not confident \nor unsure. 80% of participants indicated they perce ived AI as having a role in \nQA, less than 10% demonstrated concern. Improved qu ality standards and \nskills were perceived as benefits however considera tion of difficult patients was \nan identified challenge. \nConclusion: Poor confidence in current QA processes was identif ied and a \nlack of standardisation of practice. Radiographers identified AI as having the \npotential to support radiographic technique audits.  Benefits and challenges \nwere identified in open comments. \nLimitations: Online survey: English language could have limited uptake \nFunding for this study: Self funded \nEthics committee - additional information: University College Dublin, \nHuman Research Ethics Committee – Sciences (HREC-LS ) - LS-LR-24-141-\nAlotaibi-Rainford. \n\n \n \nAbstract-based Programme \n \n 51  \nWednesday \nAuthor Disclosures:  \nMeshal Mujaydia Alotaibi: Nothing to disclose \nLouise A. Rainford: Nothing to disclose \nJaka Potočnik: Nothing to disclose \nJonathan Mcnulty: Nothing to disclose \n \n \nA comparative study assessing the effectiveness of artificial intelligence \nand simulation education on reporting radiographer lung cancer \ndetection \nE. Compton¹, S. Lightfoot¹, R. Shah², S. Ather², P.  Taylor¹, *N. H. Woznitza*¹; \n¹London/UK, ²Oxford/UK \n(nicholas.woznitza@nhs.net) \n \nPurpose or Learning Objective: Chest radiographs (CXRs) are a high-\nvolume test, performed for a broad spectrum of reas ons. Education has been \nshown to improve CXR reporting accuracy, in particu lar for less experienced \nreporters. Similarly, artificial intelligence (AI) as a clinical decision support tool \nprovides novice readers with the most benefit. The aim of this study was to \ncompare the impact of education (SIM) with AI in CX R reporting accuracy. \nMethods or Background: A multi-reader, multi-case diagnostic accuracy \nstudy was conducted to determine the impact of SIM and AI on reporting \nradiographer (RR) CXR accuracy. 64 RR consented and  completed bank 1 \nand were randomised stratified by years’ experience  (n=32,50% to AI). 43 RRs \n(24 AI, 19 education) completed both image banks (n =52 CXRs, 26 abnormal). \nResults or Findings: Similar pre and post intervention accuracy was foun d. \nThe AI cohort decreased sensitivity (74%-65%,p=0.01 5) but increased \nspecificity (63%-77%,p<0.0001), the increase SIM se nsitivity (69%-\n62%,p=0.115) and decrease in specificity (62%-68%,p =0.217) were not \nstatistically significant. Standalone AI sensitivit y and specificity were 54% and \n77% respectively. For the AI arm, when the AI was c orrect specificity improved \n(67%-86%,p<0.001) with no significant difference in  sensitivity (89%-\n86%,p=0.31), however when AI was incorrect there wa s a significant decrease \nin sensitivity (52%-32%,p<0.001) with no difference  in specificity (both \n57%,p=1). There were four CXRs that only had one (n =3) or three (n=1) pre-\nintervention correct decisions, suggesting the bank  selected comprised of very \nchallenging cases. \nConclusion: In a challenging CXR bank, both education and AI im proved RR \nperformance. As AI tools are adopted for CXR interp retation in clinical practice \nfurther work is required to ensure reporters are ed ucation in their use. \nLimitations: The enhanced prevalence (50% abnormal) and single p athology \n(lung cancer) may limit transferability into clinic al practice. \nFunding for this study: This study was conducted as part of a clinical \nfellowship supported by NHS England (London). \nEthics committee - additional information: Canterbury Christ Church \nUniversity ETH2223-0246 21st April 2023 \nAuthor Disclosures:  \nPaul Taylor: Nothing to disclose \nSam Lightfoot: Nothing to disclose \nEmma Compton: Nothing to disclose \nRuchir Shah: Nothing to disclose \nSarim Ather: Founder: RAIQC \nNicholas Hans Woznitza: Consultant: InHealth Grant Recipient: SBRI \nHealthcare Employee: NHS London Consultant: SMR Hea lth & Tech \n \n \nRadiographers’ and students’ perspectives on artifi cial intelligence - \nA cross-sectional online survey \n*M. R. V. Pedersen*¹, M. W. Kusk², S. Lysdahlgaard² , H. Mork-Knudsen³,  \nC. Malamateniou⁴, J. Jensen⁵; ¹Vejle/DK, ²Esbjerg/DK, ³Bergen/NO, \n⁴London/UK, ⁵Odense/DK \n(Malene.Roland.Vils.Pedersen@rsyd.dk) \n \nPurpose or Learning Objective: The integration of artificial intelligence (AI) \ninto radiography offers potential in enhancing work flow efficiency, image \nprocessing, patient positioning, and quality assura nce.. This study aimed to \ninvestigate the perspectives and attitudes towards AI in radiography. \nMethods or Background: An online survey including of 29 items was \ndistributed via social media platforms to Nordic st udents and radiographers \nworking in Denmark, Norway, Sweden, Iceland, Greenl and, and the Faroe \nIslands. The survey included questions on demograph ics, specialization, \neducational background, place of work, and perspect ives and knowledge on \nAI. The items were a mix of closed-type and scaled questions, with options for \nfree-text responses when relevant. \nResults or Findings: The survey received 586 responses from all Nordic \ncountries. The mean age was 37.2 years with a stand ard deviation (SD) of \n±12.1 years,. A total of 43% (n = 254) of the respondents had not received any \nAI training in clinical practice, while 13% (n = 76 ) had received AI training \nduring their radiography undergraduate studies. Add itionally, 77.9% (n = 412) \nexpressed interest in pursuing AI education. The ma jority of respondents \n(82.8%, n = 485) were aware of the potential use of  AI, and 39.1% (n = 204) \nhad no reservations about AI. \nConclusion: Overall, radiographers have a positive attitude tow ards AI. \nHowever, there has been very limited training or ed ucation provided to \nradiographers, despite 82.8% reporting plans to imp lement AI in clinical \npractice. Generally, awareness of AI applications i s high \nLimitations: Limitations include language barriers as this surve y was provided \nin English. Most Nordic radiographers speak, read, and write English very well. \nYet, when it come to complex sentences in English t here is a higher risk of \nskipping items, survey drop out, language misunders tanding or \nmisinterpretation. \nFunding for this study: No funding \nEthics committee - additional information: The study was approved by the \nResearch Ethics Committee at the University of Sout hern Denmark (ID: 22-\n58485) \nAuthor Disclosures:  \nHelene Mork-Knudsen: Nothing to disclose \nMartin Weber Kusk: Nothing to disclose \nMalene Roland Vils Pedersen: Nothing to disclose \nJanni Jensen: Nothing to disclose \nSimon Lysdahlgaard: Nothing to disclose \nChristina Malamateniou: Nothing to disclose \n \n \nAn analysis of the user interface preferences of im aging professionals for \nAI to support clinically relevant decision making \n*A. Gill*, S. L. Mcfadden, C. Rainey, L. Mclaughlin , J. Mcconnell, C. Hughes,  \nR. Bond; Belfast/UK \n(avgill16@gmail.com) \n \nPurpose or Learning Objective: This study investigates the cognitive \nbehaviour of imaging staff when interacting with Ex planation User Interfaces \n(EUI). Data was gathered on user preferences of che st radiograph Artificial \nIntelligence (AI)-based EUIs. \nMethods or Background: Human and machine interaction involves the EUI \nthat clinicians use to link medical diagnosis or re port. However, there is \ncurrently a lack of EUI standardisation within this  field (Schalekamp et al, \n2022). Building on an international questionnaire u ndertaken at ECR 2024, a \nmulti-methods study was undertaken incorporating ey e-tracking, Think-Aloud \nand a questionnaire at UKIO 2024. Diagnostic radiog raphers’, radiologists’, \ntrainee radiologists’ and student radiographers’ id entified visual preferences \nwhen reviewing four different types of chest radiog raph AI EUIs i.e. 1) salience \nmaps, 2) textual reports, 3) area of interest and 4 )abnormality score EUIs. \nParticipants reviewed the images whilst wearing eye -tracking software and \nvoiced their thought processes i.e. the “Think-Alou d” method. The post study \nquestionnaire asked the participants about their pe rceived level of confidence \nagainst the four different interfaces. \nResults or Findings: 24 participants enabled understanding of which \ncomponents of the chest radiograph EUI are focused on and subsequently \npreferred. Eye-tracking data relating to fixations and saccades statistically \ndescribed patterns where maximal attention was dire cted in the interpretation \nprocess. Think-Aloud and post-study questionnaire d ata added further insights \nto participant EUI preferences. The analysis of the  eye-tracking study remains \nongoing, and completion is aimed for January 2025. \nConclusion: Understanding user preference for chest radiograph AI EUI \nsupports appropriate user engagement with the infor mation provided by the \ntechnology. This gives radiographers and radiologis ts the ability to explain this \nfurther to patients, positively impacting their und erstanding and subsequent \ncare. \nLimitations: Small sample size may have affected the wider gener alisability of \nfindings. Eye-tracking software capabilities \nFunding for this study: PhD funded by Department for the Economy \nEthics committee - additional information: FCNUR-23-084 reference \nAuthor Disclosures:  \nJonathan Mcconnell: Nothing to disclose \nLaura Mclaughlin: Nothing to disclose \nClare Rainey: Nothing to disclose \nCiara Hughes: Nothing to disclose \nRaymond Bond: Nothing to disclose \nSonyia Lorraine Mcfadden: Nothing to disclose \nAvneet Gill: Nothing to disclose \n \n \nSystematic review on advanced image post-processing  and workflow \noptimization in cardiovascular MRI \n*V. Tambè*, M. Zanardo, C. Torrito, P. Della Cagnol etta, F. Secchi; Milan/IT \n(valentina.tambe@gmail.com) \n \nPurpose or Learning Objective: Cardiovascular magnetic resonance imaging \n(CMR) is a critical tool for diagnosing heart disea se, but is hindered by long \nacquisition times and manual post-processing. This systematic review \nexamines recent advancements in image post-processi ng and workflow \noptimization, focusing on the integration of artifi cial intelligence (AI). \n\n \n \nAbstract-based Programme \n \n 52  \nWednesday \nMethods or Background: A systematic search was conducted using PubMed \nand EMBASE. Included studies involved the use of AI -based methods to \noptimize CMR workflows and post-processing. Eligibl e articles were those \naddressing any of the following: image reconstructi on, segmentation, workflow \nautomation, and clinical integration of AI tools. S tudies without quantitative \noutcomes related to workflow efficiency or post-pro cessing improvements were \nexcluded. \nResults or Findings: Out of 151 articles screened, 33 studies were inclu ded. \nKey findings included: automated segmentation repor ted in 15/33 (45%) \nstudies; image reconstruction in 10/33 (30%); workf low automation 8/33 (24%); \nclinical efficiency in 7/33 (21%); quality control in 5/33 (15%). In the automated \nsegmentation articles, results showed improvements in segmentation speed \nand accuracy with Dice similarity coefficients exce eding 0.90 in many studies, \nand reducing manual post-processing time by up to 6 6%. Other studies \nfocused on reducing scan times and enhancing image quality, with AI-based \nmethods reducing scan times by up to 40% while main taining image quality. \nArticles showed significant improvements in reporti ng times (by up to 30%), \nwhile 5 articles presented data on AI-based quality  control reducing rescans \n(by up to 20%). \nConclusion: AI integration into CMR has significantly improved workflow \nefficiency, reducing acquisition times and enhancin g diagnostic accuracy. \nAutomated segmentation, image reconstruction, and w orkflow automation have \naccelerated processes, reduced operator dependency,  and minimized rescans. \nLimitations: Further large-scale validation is needed to fully i mplement AI in \nCMR across diverse clinical settings. \nFunding for this study: None \nEthics committee - additional information: Not Applicable \nAuthor Disclosures:  \nFrancesco Secchi: Nothing to disclose \nMoreno Zanardo: Nothing to disclose  \nValentina Tambè: Nothing to disclose \nPaolo Della Cagnoletta: Nothing to disclose \nCarmelo Torrito: Nothing to disclose \n \n \nBlended intensive program for innovative technologi es and deep learning \nmodels (AI) in the radiographer's working environme nt \n*C. Schneckenleitner*, C. Kamp, C. Vogl, A. Raith, G. Guevara; Vienna/AT \n(christian.schneckenleitner@fh-campuswien.ac.at) \n \nPurpose or Learning Objective: The working environment of radiographers is \ncharacterized by permanent technological innovation s. We developed an \ninternational intensive program containing theory a nd hands-on training to \nintroduce bachelor and masters’ students to future technologies. This program \nis designed to prepare students for innovative tech nologies and expand their \nprofessional skills in areas such as deep learning models, Computer Assisted \nSurgery Simulation, 3D Printing, Optical Scanning a nd visualization utilizing \nmixed reality. \nMethods or Background: During a supervised online phase, students learned \nhow to create patient-specific 3D-printed models, C T-data segmentation for \ndeep learning training with 3D-Slicer, computer-ass isted surgical planning \n(CAS), how to acquire optical 3D-scans and mixed re ality visualization of the \nrespective 3D-models they created. All this content  was worked on by the \nstudents in a follow-up international skills lab we ek at the University of Applied \nSciences Vienna. Each skills lab block included 6h hours of hands-on training. \nThe students generated deep learning models with th e platform MONAI, \ncreated surgical plans with Materialise Mimics, sca nned with optical 3D-\nscanners from ARTEC and created mixed reality visua lizations. \nResults or Findings: The results were uploaded to an online platform \n(Moodle) by the students and analyzed based on defi ned criteria. Results, 39 \nout of 42 students were able to create a segmentati on for deep learning \ntraining according to the required criteria. 39 out  of 42 successfully performed \na CAS plan, 20 created a 3D-printable patient speci fic model and 42 of 42 \nwere able to produce an optical 3D-scan of the face . \nConclusion: The results show that radiography students can prod uce results \nin adjacent technology areas and expand their compe tence in future \ntechnologies. \nLimitations: The limited time can generate interest but not a sp ecialization in \nthe profession. Explicit training programs are need ed to deepen radiographers’ \nexpertise in these areas. \nFunding for this study: No funding. \nEthics committee - additional information: The results includes no patient \nspecific information or clinical interventions. \nAuthor Disclosures:  \nChristoph Kamp: Nothing to disclose \nChristian Schneckenleitner: Nothing to disclose \nAlexander Raith: Nothing to disclose \nGodoberto Guevara: Nothing to disclose \nChristoph Vogl: Nothing to disclose \n \n \nEvaluation of ChatGPT as support in image qualitati ve assessment for \ncardiac sonographers \n*K. Tissir*¹, G. R. Bonfitto¹, A. Roletto¹, A. Sign oroni²; ¹Milan/IT, ²Brescia/IT \n(karimatissir@hotmail.it) \n \nPurpose or Learning Objective: The integration of Large Language Models \n(LLM) tools like ChatGPT in clinical settings is ch anging how healthcare \nprofessionals manage diagnoses and workflow. In car diac clinics, the accurate \nand timely interpretation of images is crucial for effective diagnosis and \nmonitoring of cardiac conditions. ChatGPT could be particularly beneficial for \ncardiac sonographers. This study aims to assess whe ther ChatGPT can \neffectively support cardiac sonographers in the qua litative evaluation of \nechocardiographic images. \nMethods or Background: A database of 50 anonymized echocardiographic \nimages was retrospectively analyzed, including 2-ch amber, 4-chamber, and \napical 3-chamber views. Three evaluators, a junior sonographer, a senior \nsonographer and ChatGPT-4o conducted the qualitativ e evaluation of the \nimages by identifying scoring them on a 5-point Lik ert scale. The guidelines of \nthe European Association of Echocardiography served  as references. \nResults or Findings: Junior sonographer correctly identified views in 84 % of \ncases (n=42), while ChatGPT correctly identified 58 % of cases (n=29).In \ncomparison to senior sonographer, the junior sonogr apher overestimated 22% \nof the images (n=11), underestimated 36% (n=18), an d agreed in 42% of the \nimages (n=21). In contrast, ChatGPT overestimated 5 2% of the images (n=26), \nunderestimated 18% (n=9), and agreed in 30% of the images (n=15). \nConclusion: ChatGPT-4o shows limitations in identifying echo ca rdiac views \ncompared to other participants. In addition, ChatGP T is inclined to \noverestimate image quality. This can be explained b y limited training of the \nLLM, mainly done with information from guidelines. As other studies in \nliterature showed, more in-depth training could inc rease the performance of \nChatGPT. LLM can assist cardiac sonographers in qua litative analysis of \nimages and supporting anomaly evaluation, but conce rns remain about its \nreliability and bias. \nLimitations: The small sample of participants and cases limited the strength of \nthe conclusions of this study. \nFunding for this study: N/A \nEthics committee - additional information: University of Brescia \nAuthor Disclosures:  \nAndrea Roletto: Nothing to disclose \nKarima Tissir: Nothing to disclose \nAlberto Signoroni: Nothing to disclose \nGiuseppe Roberto Bonfitto: Nothing to disclose \n \n \nNavigating Artificial Intelligence (AI) Leadership:  Radiographers’ \nReadiness and Challenges in Europe \n*G. Walsh*¹, Y. Kyratsis², A. Goodall¹, J. St John- Matthews¹,  \nC. Malamateniou¹; ¹London/UK, ²Rotterdam/NL \n(gemma.walsh@outlook.com) \n \nPurpose or Learning Objective: This study offers unique insights into the \npreparedness of radiographers to pursue AI leadersh ip roles within healthcare \nand potential barriers preventing radiographers exc elling in the AI ecosystem. \nMethods or Background: A European-wide, cross-sectional study utilising a \nmixed methods online survey. Snowball sampling allo wed qualified \nradiographers, and radiography students, to answer the survey, irrespective of \ntheir current role. The survey explored the followi ng areas of interest: a) \ngeneral radiographer demographics, b) radiographers  preparedness and \nconfidence to lead the implementation of AI in heal thcare, c) suggested day-to-\nday responsibilities of an AI-lead radiographer and  d) motivations for \nconsidering AI leadership roles. \nResults or Findings: There were 1733 valid responses from 37 European \ncountries. The typical respondent was female (64%),  a diagnostic radiographer \n(59.9%) with >20 years’ experience (31.3%). A lack of education, training and \nresources led 72.3% of radiographers to feel the pr ofession is not prepared to \nlead the implementation of AI in healthcare. Those that felt confident to lead in \nan AI-enabled work environment (50.06%), felt they already have the \nnecessary experience and skills, but also admitted additional resources would \nbe needed. The top two motivators to pursue an AI l eadership role included \nchampioning change and the promise of appropriate t raining. Inferential \nStatistics ongoing as of October 2024. \nConclusion: Radiographers have a unique skill-set making them t he ideal \ncandidates for AI leadership roles within healthcar e. Radiographers do not \ncurrently feel confident or prepared to undertake A I leadership roles with \neducation, training and a lack of resources creatin g barriers for this. It is \nreassuring radiographers feel motivated to undertak e AI leadership roles, \nhowever increased training and educational support are needed. \nLimitations: Gives a snapshot view of radiographers perceptions.  \nSnowball sampling can lead to selection-bias, but a llows for many recruits. \nFunding for this study: This research has been funded by the AI special cal l \nof the College of Radiographers Industry Partnershi p Scheme (CORIPS) of the \nCollege of Radiographers (Reference Number 218). \n\n \n \nAbstract-based Programme \n \n 53  \nWednesday \nEthics committee - additional information: Ethics approval (ETH2223-1346) \nwas granted by City, University of London. \nAuthor Disclosures:  \nGemma Walsh: Nothing to disclose \nYiannis Kyratsis Yiannis Kyratsis: Nothing to discl ose \nJanice St John-Matthews: Nothing to disclose  \nAmanda Goodall: Nothing to disclose \nChristina Malamateniou: Nothing to disclose \n \n \nMRI deep learning models for assisted diagnosis of knee pathologies and \ninjuries: A systematic review \n*K. M. Mead*, T. Cross, G. Roger, R. Sabharwal, S. Singh, N. Giannotti; \nSydney/AU \n(keileymead13@gmail.com) \n \nPurpose or Learning Objective: Several studies have demonstrated that \ndeep learning (DL) models can be effectively traine d on MRI data to assist \nclinicians in identifying knee injuries and patholo gies. This systematic review \nwas conducted to explore the current landscape of e xisting DL models \ndeveloped for detecting knee injuries and pathologi es through magnetic \nresonance imaging (MRI) and to assess their potenti al clinical applications. \nMethods or Background: Five databases were systematically searched using \nthe following terms ‘Knee AND 3D AND MRI AND Deep L earning’. The \nCovidence platform was used to screen publications based on title, abstract, \nand full text. Only original research articles focussing on the automatic \ndetection of knee injuries and pathologies using DL  models for MRI were \nincluded. The synthesis of results was performed by  two independent \nreviewers. \nResults or Findings: Fifty-four studies were included. The studies focus ed on \nanterior cruciate ligament injuries (n=19), osteoar thritis (n=9), meniscal injuries \n(n=13), general abnormal knee appearance (n=10), ti bial plateau fractures \n(n=1) and synovial fluid detection (n=1). The follo wing convolutional neural \nnetwork (CNN) infrastructures were used: ResNet, VG G, DenseNet, and \nDarkNet. The averaged performance outcomes of the D L models \ndemonstrated sensitivity, specificity, AUC-ROC, and  accuracy of 87%, 90%, \n92%, and 88%, respectively. The DL models for the d etection of a specific \ninjury or pathology outperformed those for general abnormality detection. \nConclusion: This systematic review underscores that fine-tuned DL models for \nknee pathologies and injuries can effectively suppo rt automatic diagnosis. \nFurther large-scale validation and prospective stud ies are needed to confirm \ntheir clinical utility as assistive diagnostic tool s. \nLimitations: Inconsistent data reporting across the studies anal ysed resulted \nin variations in the reporting of DL model performa nce. Sub-group analyses \nwere performed to minimise bias. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective. Ethics \napproval was deemed unnecessary by the Research Int egrity and Ethics \nCommittee at the University of Sydney, Australia. \nAuthor Disclosures:  \nKeiley Michelle Mead: Nothing to disclose \nNicola Giannotti: Nothing to disclose \nSahaj Singh: Nothing to disclose \nTom Cross: Nothing to disclose  \nRohan Sabharwal: Nothing to disclose \nGreg Roger: Nothing to disclose \n \n \nAn Innovative AI-Based Interactive Tool for Learnin g Chest X-Ray \nAnatomy \n*R. S. T. Ribeiro*, T. Coutaudier, L. Mourot, C. S.  D. Reis, L. Raileanu; \nLausanne/CH \n(ricardo.ribeiro@hesav.ch) \n \nPurpose or Learning Objective: To enhance the learning process of chest X-\nray/(CXR) anatomy for medical imaging/(MI) students  by integrating AI \nsegmentation and classification tools into an educa tional web-\napplication/(webapp). \nMethods or Background: The webapp was designed as an interactive \nplatform where students can identify CXR anatomy. T he platform was \nimplemented in Python using Flask to provide the we b-interface, \nTorchXRayVision to classify and segment key regions  (heart, lungs, clavicles, \nspine, scapula, trachea) with AI. PostgreSQL stored  and managedCXR public \ndatasets allowing students’ practice. The user inte rface allows selection and \noutline regions of interest on the radiographs, tha t are compared to \nsegmentations obtained with AI-algorithms. Feedback  is provided through Dice \ncoefficient that assess the accuracy of the user’s segmentation compared to \nthe AI-based reference. The app’s system architectu re is modular. \nResults or Findings: The developed framework successfully integrates AI for \nfast and accurate segmentation of CXR. Its design a llows users to upload their \nown radiographs or use others supplied by public da tasets, interact with \nradiographs by selecting anatomical regions and rec eiving immediate \nfeedback. This functionality aims to support autono mous learning and reduce \nthe need for constant instructor supervision. The m odular architecture ensures \nscalability, enabling the inclusion of more types o f radiographic images and \nenhancing its potential for broader applications in  medical imaging education. \nConclusion: The web application demonstrates a promising approa ch to \nimprove MI education by providing an interactive an d AI-powered learning tool. \nIts design is intended to be adaptable/accessible t hrough any web-browser, \nwith potential to expand into areas such as quality  assessment. Further \ndevelopment of this framework is planned to test it s impact on the MI students \nlearning process. \nLimitations: The app is limited to CXR and relies on AI segmenta tion \nperformance. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nClaudia Sa Dos Reis: Nothing to disclose \nRicardo Silva Teresa Ribeiro: Nothing to disclose \nLucas Mourot: Nothing to disclose \nThéo Coutaudier: Nothing to disclose \nLaura Raileanu: Nothing to disclose \n \n \n15:00-16:00 Research Stage 1 \nResearch Presentation Session: Breast \nRPS 502 \nMRI and contrast-enhanced \nmammography for treatment planning \n \nModerator \nM. Dorrius; Groningen/NL  \n(m.d.dorrius@umcg.nl) \nAuthor Disclosures:  \nMonique Dorrius: Research Grant/Support: KWF PPS Gr ant \n \n \nRadiomic Shape Features for Assessment of Early The rapy Response to \nNeoadjuvant Chemotherapy of Breast Cancer Patients - Preliminary \nResults of the PREDICTOME-Study \n*D. A. Resch*, O. Lafcı, P. Clauser, Z. Bago-Horvat h, Y. Tan, G. Langs,  \nT. Helbich; Vienna/AT \n(daphne.resch@gmail.com) \n \nPurpose or Learning Objective: To analyze the significance of radiomic \nshape features derived from dynamic contrast-enhanc ed (DCE) breast MRI in \nassessing early treatment response in breast cancer  patients undergoing \nneoadjuvant chemotherapy (NAC). \nMethods or Background: We report on the first 29 breast cancer patients of  a \nprospective study, who underwent NAC and received m ultiparametric 18F-FDG \nPET/MRI imaging at baseline (T0) and after three we eks of NAC (T1). DCE-\nMRI derived radiomic shape features, including SHAP E_Volume (mL), \nSHAPE_Volume (vx), SHAPE_Sphericity, SHAPE_Surface( mm²) and \nSHAPE_Compacity were extracted using LIFEx. The agn ostic shape features \nwere compared to the treatment response assessment by two radiologists. All \ndata were stratified by the pathological complete r esponse status (pCR or non-\npCR). Mean change (Δ) of all investigated features were calculated. Pea rson \nChi-Square Test, T-Test and Mann-Whitney U Test wer e applied. \nResults or Findings: Twelve out of 29 (41%) patients had pCR and 17/29 \n(59%) had non-pCR. Radiomic shape features, includi ng ΔSHAPE_Compacity \nand SHAPE_Volume at T1, were significantly associat ed with pCR (P = 0.015 \nand P = 0.04, respectively). Radiologist's response  assessment (stable \ndisease, partial response, disease progression) was  not significantly \nassociated with pathological outcomes (pCR vs. non- pCR) (Pearson Chi-\nSquare: χ2=3.727χ2=3.727, P = 0.155). Similarly, the likelihood rati o test did \nnot show significant results (P = 0.105). \nConclusion: While the radiologist’s assessment did not signific antly correlate \nwith pathological outcomes, radiomic shape features , particularly \nΔSHAPE_Compacity and SHAPE_Volume, demonstrated sign ificant \nassocitations with pCR. These findings suggest that  DCE derived radiomic \nshape features might be a promising tool for predic tion of early NAC response \nin breast cancer patients and may challenge the val idity of RECIST criteria at \nthis stage. \nLimitations: This study is limited by the small sample size. We will be able to \nreport on more data in March 2025. \nFunding for this study: This study is supported by the WWTF (Vienna \nScience and Technology Fund), grant number LS20-065 . \n\n \n \nAbstract-based Programme \n \n 54  \nWednesday \nEthics committee - additional information: EK Nr. 510/2009 \nAuthor Disclosures:  \nGeorg Langs: Nothing to disclose \nDaphne Ariane Resch: Nothing to disclose \nThomas H. Helbich: Nothing to disclose \nOğuz Lafcı: Nothing to disclose \nZsuzsanna Bago-Horvath: Nothing to disclose \nYen Tan: Nothing to disclose \nPaola Clauser: Nothing to disclose \n \n \nReactive Cutaneous Capillary Endothelial Proliferat ion in Breast Tissue \non MRI during Neoadjuvant Chemoimmunotherapy with C amrelizumab in \nTNBC \n*X. Ma*, Q. Xiao, Y. Huang, Y. Gu; Shanghai/CN \n(maxiaowen9397@163.com) \n \nPurpose or Learning Objective: To describe the Reactive cutaneous capillary \nendothelial proliferation (RCCEP) that occurs withi n the breast tissue of triple-\nnegative breast cancer (TNBC) patients undergoing n eoadjuvant \nchemoimmunotherapy with camrelizumab and to investi gate the potential \nfactors influencing its occurrence. \nMethods or Background: We retrospectively collected 106 cases of TNBC \npatients from March 2021 to August 2023, including 60 cases who received \nneoadjuvant chemotherapy (NAC) and 46 cases who und erwent neoadjuvant \nchemoimmunotherapy (NACI). We analyzed the clinical  data, pathological \ncharacteristics, MRI at baseline and during each tr eatment cycle for all \npatients, then identified abnormal lesions after tr eatment and explored their \ninfluencing factors. \nResults or Findings: Abnormal enhancement was observed in 30 patients (3  \nNAC vs. 27 NACI, P<0.001). In the NACI group, cases  of abnormal \nenhancement included 3 cases of ring enhancement, 2  cases of linear \nenhancement, and 22 cases of mass enhancement, whic h typically appeared \nafter the second cycle and rarely appeared after th e fourth or sixth cycle. \nThese lesions generally decreased in size with cont inued treatment. In the \nNACI group, the development of new enhancement lesi ons was correlated with \nyounger age (P=0.007), premenopausal status (P=0.01 4), a lack of peritumoral \nedema on baseline MRI (P=0.007), and the presence o f mass enhancements \n(P=0.012). \nConclusion: TNBC patients treated with camrelizumab frequently exhibit \nRCCEP in the breast tissue, primarily presenting as  mass enhancements on \nMRI. These lesions often regress in size even witho ut drug withdrawal, \nsuggesting that recognizing this pattern can preven t unnecessary biopsies and \nhelp in adjusting treatment strategies accordingly.  \nLimitations: First, this study is retrospective and includes a s mall sample size; \nsecond, in future studies, it is necessary to exten d the research period to \nexplore the relationship between RCCEP in the breas t and their prognosis. \nFunding for this study: None \nEthics committee - additional information: Fudan University Shanghai \nCancer Center \nAuthor Disclosures:  \nYan Huang: Nothing to disclose \nYajia Gu: Nothing to disclose \nQin Xiao: Nothing to disclose \nXiaowen Ma: Nothing to disclose \n \n \nPatient Perspective: Evaluating Imaging Preferences  in Women \nUndergoing Contrast-Enhanced Breast MRI and Contras t-Enhanced \nMammography (CEM) \nN. Caldarelli, *G. Della Pepa*, G. Irmici, E. D'Asc oli, C. De Berardinis,  \nD. Ballerini, A. Bonanomi, C. Depretto, G. P. Scape rrotta; Milan/IT \n(gianmarco.dellapepa@gmail.com) \n \nPurpose or Learning Objective: The study evaluated patient preferences \nbetween contrast-enhanced MRI (MRI) and contrast-en hanced mammography \n(CEM) for breast cancer imaging. MRI has long been the gold standard for \nbreast cancer staging and monitoring neoadjuvant ch emotherapy, while CEM \nhas gained popularity due to its lower cost, faster  examination times, and \naccessibility, offering comparable diagnostic perfo rmance. \nMethods or Background: The aim of this study was to understand patient \npreferences between these two modalities to support  the broader \nimplementation of CEM in clinical practice. The stu dy included 152 patients \nwho underwent both procedures within a six-month in terval between 2018 and \n2024. A Likert scale questionnaire was used to asse ss patient preferences \nfocusing on three main aspects: breast positioning (compression for CEM and \ncoil positioning for MRI), sensation during contras t injection, and overall \ncomfort (exam duration, machine noise, and environm ental factors). \nResults or Findings: Results showed that 72.4% of patients preferred CEM , \n26.3% preferred MRI, and 1.3% expressed no preferen ce. CEM was \nsignificantly more comfortable (p<0.001), with high er median scores than MRI. \nThe main reasons for preferring CEM included faster  exam time (28%), lack of \nclaustrophobia (17%), and absence of noise (15%). I n terms of breast \npositioning, there was a slight preference for MRI (p=0.04). No significant \ndifferences were found in the sensation during cont rast injection (p=0.07). \nConclusion: In conclusion, CEM was the preferred option for mos t patients, \nindicating its potential as an alternative to MRI i n clinical settings. These \nfindings support further exploration of CEM's role in breast cancer imaging. \nLimitations: none. We have no limitations. \nFunding for this study: none. We don't need any funding for this study. \nEthics committee - additional information: none. We don't have any Ethics \ncommittee. \nAuthor Disclosures:  \nNazarena Caldarelli: Nothing to disclose \nGianmarco Della Pepa: Nothing to disclose \nElisa D'Ascoli: Nothing to disclose \nAlice Bonanomi: Nothing to disclose \nDaniela Ballerini: Nothing to disclose \nGianfranco Paride Scaperrotta: Nothing to disclose \nClaudia De Berardinis: Nothing to disclose \nCatherine Depretto: Nothing to disclose \nGiovanni Irmici: Nothing to disclose \n \n \nNon-invasive imaging of the tumor pH in breast canc er with CEST-MRI:  \nA preclinical study \n*D. Prinz*¹, S. J. Bartsch¹, J. Friske¹, D. Laimer- Gruber¹, T. H. Helbich¹,  \nK. Pinker-Domenig²; ¹Vienna/AT, ²New York, NY/US \n(daniela.a.prinz@meduniwien.ac.at) \n \nPurpose or Learning Objective: Tumor acidosis is a key hallmark of breast \ncancer (BC). The increased glucose consumption trig gers aerobic glycolysis, \nleading to the production of lactic acid which resu lts in therapy resistance. \nCurrently, there is no non-invasive tool available to image tumor pH in vivo. We \nattempted to image the extracellular pH (pHe) with acidoCEST using \nIopamidol, and the intracellular pH (pHi) using the  CEST-derived AACID \n(amine and amide concentration-independent detectio n) metric. Non-invasive \nimaging of tumor pH is of great interest because pH  is one of the first \nbiomarkers which changes during treatment. \nMethods or Background: Female athymic nude mice were inoculated with BC \ncells of HER2+ (SKBR-3, n = 5) and triple-negative (MDA-MB-231, n = 4) \nmolecular subtypes. MRI imaging was performed using  a preclinical 9.4T MRI \nsystem. CEST images were acquired and ratiometric m easurements were \nevaluated for the endogenous AACID from baseline im ages and the \nacidoCEST signal based on post-challenge images. \nResults or Findings: Both acidoCEST and AACID produced reliable and \nstable signals. The lower pHe and the higher pHi co uld be visualized for \nHER2+ and triple-negative BC subtypes. Parametric m aps of AACID and \nacidoCEST revealed differences in the pH gradients between BC subtypes, \nwhich correlated with tumor aggressiveness. \nConclusion: We conclude that the AACID-based measurement of pHi  \nsufficiently quantifies the pH gradient between ext racellular and intracellular \ncompartments and may become a promising non-invasiv e contrast-free \nimaging method to monitor early treatment response in BC. \nLimitations: Due to the preclinical nature of this study, the sm all number of \nmice used presents a limitation. \nFunding for this study: This work was funded by the Vienna Science and \nTechnology Fund (WWTF), grant no. LS19-018. \nEthics committee - additional information: This animal study was approved \nby Austrian Federal Ministry of Education, Science and Research \n[66.009/0284-WF/V/3b/2017; 2020-0.363.124; 2022-0.7 26.820] and the \nIntramural Committee for Animal Experimentation of the Medical University of \nVienna. \nAuthor Disclosures:  \nSilvester Julian Bartsch: Nothing to disclose \nDaniela Laimer-Gruber: Nothing to disclose \nKatja Pinker-Domenig: Nothing to disclose \nThomas H. Helbich: Nothing to disclose \nDaniela Prinz: Nothing to disclose \nJoachim Friske: Nothing to disclose \n \n \nCEM Background Parenchymal Enhancement: Exploring I ts Clinical and \nBiological Correlations \n*C. De Berardinis*¹, C. Depretto¹, G. Della Pepa¹, E. D'Ascoli¹, G. Irmici¹,  \nE. Ancona¹, R. Spiaggia², L. Corradini², G. P. Scap errotta¹; ¹Milan/IT, \n²Mussomeli/IT \n(cla.deberardinis@gmail.com) \n \nPurpose or Learning Objective: To assess the correlation between the \ndegree of background parenchymal enhancement (BPE) on contrast-enhanced \nmammography (CEM) and breast density, menopausal st ate, receptor status, \nHER 2 expression and proliferation index (Ki-67) of  malignant tumors. \n\n \n \nAbstract-based Programme \n \n 55  \nWednesday \nMethods or Background: We retrospectively evaluated all patients who \nunderwent CEM at our Institution from January 2023 to April 2024 for pre-\noperative staging and problem solving. BPE was clas sified as minimal, mild, \nmoderate, or marked, in accordance with the BIRADS lexicon. We assessed \nthe menopausal status and evaluated the receptor st atus and Ki-67 of each \nmalignant lesion. Statistical analysis was performe d using the Spearman’s test \nto evaluate the correlation between density and BPE . Chi-square test was \nused to evaluate the correlation between BPE, menop ausal status, receptor \nstatus, and Ki-67. \nResults or Findings: A total of 194 patients were included. Spearman's t est \nanalysis demonstrated a weak correlation between BP E and breast density \n(ρ=0.353, p<0.001, CI:0.219-0.469). The Chi-Square te st revealed a strong \nassociation between BPE and menopausal state, with lower BPE levels in \npostmenopausal patients compared to premenopausal p atients (X²=30.846, \np<0.001). There was no statistically significant as sociation between BPE and \nreceptor status (X²=14.494, p=0.270) and HER2+ stat us (X²=1.648, p=0.649). \nThere was suggestive but not statistically signific ant association (X²=6.738, \np=0.081) between BPE and Ki67. \nConclusion: A significant correlation was demonstrated between the level of \nBPE on CEM and menopausal status. Instead, the corr elation with breast \ndensity proved to be weak, while no statistically s ignificant correlation was \nfound with tumor receptor status. BPE also appears to have a suggestive \ncorrelation with Ki-67% and consequently with tumor  aggressiveness. \nLimitations: This association needs more data and larger sample size to be \nconfirmed but suggests how BPE might be related to breast cancer risk. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nGianmarco Della Pepa: Nothing to disclose \nElisa D'Ascoli: Nothing to disclose \nClaudia De Berardinis: Nothing to disclose \nEleonora Ancona: Nothing to disclose \nGianfranco Paride Scaperrotta: Nothing to disclose \nCatherine Depretto: Nothing to disclose \nLisa Corradini: Nothing to disclose \nRossana Spiaggia: Nothing to disclose \nGiovanni Irmici: Nothing to disclose \n \n \nRole of Breast MRI to identify patients with lesion s of uncertain malignant \npotential (B3) who might avoid surgery: a systemati c review and \nmetanalysis \n*G. Vatteroni*¹, N. Pötsch², P. Clauser², P. A. Bal tzer²; ¹Milan/IT, ²Vienna/AT \n(giulia.vatteroni@gmail.com) \n \nPurpose or Learning Objective: This systematic review and meta-analysis \ninvestigates the added value of Contrast Enhanced B reast MRI (CE-MRI) to \nrule out malignancy in patients with lesions of unc ertain malignant potential \n(B3) diagnosed at image guided biopsy. \nMethods or Background: A systematic review and meta-analysis were \nconducted using predefined criteria. Eligible artic les, published in English until \nAugust 2024, dealt with CE-MRI in the context of B3  lesions. Two reviewers \nextracted study characteristics, including true pos itives (TP), false positives \n(FP), true negatives (TN), and false negatives (FN) . Sensitivity, specificity, \nnegative likelihood ratio, and positive likelihood ratio were calculated using \nbivariate random effects. Fagan nomograms identifie d the maximum pretest \nprobability at which post-test probabilities of a n egative MRI aligned with the \n2% malignancy rate benchmark for downgrading BI-RAD S 4 to BI-RADS 3. I² \nstatistics and meta-regression explored sources of heterogeneity. P-values \n<0.05 were considered significant. \nResults or Findings: Seven studies investigating 479 patients with 493 h igh \nrisk lesions undergoing CE-MRI were included. The a verage breast cancer \nprevalence was 17% (88/493). Pooled sensitivity was  91.3% (95%-CI: 82.8%-\n95.8%) and pooled specificity was 68.8% (95%-CI 50. 3%-82.8%) using a \nrandom effects model. Overall, CE-MRI missed only 6 /493 lesions, all small \nlow-grade Ductal Carcinoma in Situ. Fagan nomograms  showed that CE-MRI \ncould rule out malignancy in B3 lesions diagnosed a t image guided biopsy up \nto a pre-test probability of 13.1%. \nConclusion: CE-MRI in the assessment of B3 lesions could potent ially identify \npatients who might avoid surgery, saving costs and time as well as reducing \npatient anxiety and morbidity. Breast cancer can be  ruled out up to pre-test \nprobabilities of 13.1%. \nLimitations: n/a \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nPascal A.T. Baltzer: Nothing to disclose \nNina Pötsch: Nothing to disclose \nGiulia Vatteroni: Nothing to disclose \nPaola Clauser: Nothing to disclose \n \n \nContrast-Enhanced Digital Mammography for the Diagn osis and \nDetermination of Extent of Disease in Invasive Lobu lar Carcinoma:  \nOur Experience and Literature Review \n*J. Lipin Margaret*, T. Friehmann, S. Tamir, G. Bac har, A. Grubstein;  \nPetach Tikva/IL \n(margaret.lipinski@gmail.com) \n \nPurpose or Learning Objective: In contrast-enhanced imaging for the precise \nevaluation of Invasive lobular carcinoma (ILC), the  aim was to further validate \nthe use of CEM for evaluation of extent of disease in ILC cases, with Digital \nbreast tomosynthesis (DBT) as an adjunct. \nMethods or Background: A retrospective study was conducted in a tertiary \nmedical center. All CEM examinations of ILC patient s between 2017–2023 \nwere re-read by two breast radiologists. The longes t lesion diameter was \ncorrelated to pathology findings. For each case, re aders provided commentary \non the added value of DBT. \nResults or Findings: Twenty-four CEM examinations were evaluated. The \ncohort were on average older than expected for ILC patients (74 vs. 63 years) \nand were unable to undergo MRI due to advanced age and comorbidities. \nEnhancing lesions, ranging from 4–75 mm, strongly c orrelated to pathology \nresults, with statistical significance, for mass an d non-mass lesions (r = 0.94, p \n< 0.001 and r = 0.99, p = 0.002, respectively). For  most lesions (17/24, 71%), \nreaders remarked that DBT allowed for improved char acterization of lesion \nmargins. \nConclusion: When compared to the pathology findings, ILC was ac curately \ndiagnosed and assessed using CEM. The addition of D BT was reported by the \ninterpreting radiologists as a valuable adjunct for  margin analysis. \nLimitations: Sample size was small, a general issue among studie s of ILC in \nCEM (e.g., 30 subjects in the study by Patel et al.  in 2018, 31 in Amato et al. in \n2019, and 38 in Costantini M et al. in 2022). The d esign was retrospective, and \nfurthermore, since all patients underwent CEM due t o contraindication to MR, \nintroducing selection bias. No patient had extremel y dense breasts or marked \nenhancement, conclusions regarding the accuracy of CEM in the most \nchallenging breast types cannot be reached. \nFunding for this study: None. \nEthics committee - additional information: Helsinki approval \nAuthor Disclosures:  \nTal Friehmann: Nothing to disclose \nJohansson Lipin Margaret: Author: Rabin Medical Cen ter, Mammography Unit \nAhuva Grubstein: Nothing to disclose \nGil Bachar: Nothing to disclose \nShlomit Tamir: Nothing to disclose \n \n \nComparison of additional malignant lesions detectio n in Dense vs. Non-\ndense breasts with Magnetic Resonance Imaging (MRI)  or Contrast-\nEnhanced Mammography (CEM) performed for loco-regio nal staging \n*P. Minichetti*, M. Lorenzon, S. Sanità, L. Nardone , L. Cereser, R. Girometti, \nC. Zuiani; Udine/IT \n(paola.minichetti@libero.it) \n \nPurpose or Learning Objective: To compare Additional Malignant Lesions \n(AML) detection of CEM or MRI (Breast Contrast-Imag ing – BCI) performed for \npreoperative loco-regional staging of Breast Cancer  (BC) in patients with \nDense Breast (DB) or Non-DB (N-DB). \nMethods or Background: We retrospectively included 290 patients (median \nage 62 years) with a biopsy-proven BC who underwent  CEM (n=129) or 1.5T-\nMRI (n=161) at our Institute between January 2022 a nd December 2023. \nStaging was performed based on EUSOMA criteria or c linical requests. A \nradiology resident (with >3 years of experience in breast imaging) reviewed all \nreports written by one-of-four breast radiologists (with 5-25 years of \nexperience). Extracted data included: density accor ding to BI-RADS (C, D \nclassified as DB; A, B as N-DB), background parench ymal enhancement, \nfeatures of index lesions and AML (not identified b efore BCI, pathologically \nproven). The BCI detection rate of AML in DB and N- DB groups was assessed \nand compared using T-test. P-values<0.05 were consi dered statistically \nsignificant. \nResults or Findings: In 201 DB patients, preoperative BCI detected 40 AM L \n(19.9%). In 89 N-DB patients, BCI detected 10 AML ( 8.9%). The difference in \nthe BCI AML detection rate between DB and N-DB grou ps was 11% \n(p=0.0202). In the DB group, the AML detection rate  was found to be more \nthan double that in the N-DB group. Specifically, a mong 71 DB patients, CEM \ndetected 15 AML (21.1%) vs 6 AML in 58 N-DB patient s (10.3%); among 130 \nDB patients, MRI detected 25 AML (19.2%) vs 4 AML i n 31 N-DB patients \n(12.9%). \nConclusion: The usefulness of BCI performed for loco-regional s taging is \nsignificantly higher in DB than in N-DB patients. T he detection rate of AML is \ncomparable between MRI and CEM in DB and N-DB patie nts. \nLimitations: Small cohort, monocentric. \nFunding for this study: Nothing. \nEthics committee - additional information: Institutional Review Board (IRB-\nDMED) \n\n \n \nAbstract-based Programme \n \n 56  \nWednesday \nAuthor Disclosures:  \nChiara Zuiani: Nothing to disclose \nSilvia Sanità: Nothing to disclose \nMichele Lorenzon: Nothing to disclose \nRossano Girometti: Nothing to disclose \nLorenzo Cereser: Nothing to disclose \nPaola Minichetti: Nothing to disclose \nLuigi Nardone: Nothing to disclose \n \n \n15:00-16:00 Research Stage 2 \nResearch Presentation Session: \nGenitourinary \nRPS 507 \nNon-malignant pathology of the female \npelvis: insights and imaging approaches \n \nModerator \nC. Panico; Rome/IT  \n(camilla.panico@guest.policlinicogemelli.it) \n \n \nDiaphragmatic Endometriosis: Correlation with Pelvi c Disease and \nSymptoms \n*N. Bogveradze*, A. Santonocito, J. Heine, T. Helbi ch, P. A. Baltzer, R. Wenzl, \nP. Clauser; Vienna/AT \n(bogveradze.nino@gmail.com) \n \nPurpose or Learning Objective: Diaphragmatic endometriosis (DE) is rare, \nwith limited data on its frequency and management. The deep Pelvic \nEndometriosis Index (dPEI) and #Enzian classificati on systems improve pelvis \nmagnetic resonance imaging (MRI) accuracy for diagn osing endometriosis by \nstratifying disease severity. Our study aimed to as sess the prevalence of DE \nand its correlation with disease severity and assoc iated clinical symptoms. \nMethods or Background: In this IRB-approved retrospective study, \nconsecutive abdominal MRIs performed for endometrio sis (2018-2022) were \nreviewed by three radiologists in consensus. Positi ve cases of diaphragmatic \nendometriosis were defined by consensus among reade rs or available \nhistology. Pelvic disease severity was assessed usi ng dPEI and #Enzian \nscores, classifying endometriosis as mild (score ≤2), moderate (scores 3-4), or \nsevere (scores ≥5). Descriptive statistics were used to analyze the  relationship \nbetween pelvic disease severity, diaphragmatic invo lvement, and symptoms. \nResults or Findings: DE was detected in 16/108 patients (14.8%). Among 9 2 \npatients with clinical data, 7/92 (7.6%) showed upp er abdominal symptoms, \nwith DE confirmed in 4/7 (57.1%). No imaging correl ates were identified for \nthree symptomatic patients. Based on dPEI, 2/92 asy mptomatic patients had \nmoderate disease (2.2%), while symptomatic patients , 17/92 had mild (18.5%), \n36/92 moderate (39.1%), and 37/92 severe (40.2%) di sease. Among 16 DE \npatients by dPEI, 2/16 (12.5%) had mild, 7/16 (43.8 %) moderate, and 7/16 \n(43.8%) severe disease. According to #Enzian, 2/92 asymptomatic patients \nhad severe disease (2.2%), while in symptomatic pat ients, 8/92 had mild \n(8.7%), 11/92 moderate (12.0%), and 71/92 severe (7 7.2%). By #Enzian, of the \n16 DE patients, none had mild, 2/16 (12.5%) had mod erate, and 14/16 (87.5%) \nhad severe disease. \nConclusion: DE detected on MRI was associated with symptoms in 25% of \nthe cases. The presence of diaphragmatic endometrio sis is associated with \nmore severe pelvic disease in both scoring systems.  \nLimitations: Retrospective \nFunding for this study: N/A \nEthics committee - additional information: Number: 2057/2017 \nAuthor Disclosures:  \nPascal A.T. Baltzer: Nothing to disclose \nRene Wenzl: Nothing to disclose \nThomas H. Helbich: Nothing to disclose \nAmbra Santonocito: Nothing to disclose  \nNino Bogveradze: Nothing to disclose \nJana Heine: Nothing to disclose \nPaola Clauser: Nothing to disclose \n \n \n \n \n \nAdenomyosis and Deep Infiltrating Endometriosis: Th e Role of \nAdenomyosis Dominant Side and Type in Implant Distr ibution \nA. Durur Karakaya, *H. Özen Atalay*, V. Samadli, U.  Kalkan; Istanbul/TR \n(handeozen15@gmail.com) \n \nPurpose or Learning Objective: The aims of this study are to evaluate the \ncorrelation between the types/dominant side of the adenomyosis and the \nlocation of deep infiltrating endometriosis (DIE) i mplants, as well as the \nrelationship between the type of adenomyosis and th e presence of \nendometrioma. \nMethods or Background: We retrospectively evaluated 311 patients with \nadenomyosis on MRI examinations between January 202 2 and September \n2024. The poor quality MRI examination and patients  without DIE were \nexcluded from the study, and the final evaluation w as performed with 50 \npatients with adenomyosis and accompanying DIE. Ade nomyosis was \nclassified into 3 based on the type as focal, diffu se, cystic; and based on the \ndominant side in the uterus as anterior, posterior,  no dominance. DIE implants \nwere also categorized based on the location as ante rior, posterior, or \ninvolvement of both compartments. Additionally, the  presence and size of \nendometrioma were evaluated. Statistical significan ce was examined by chi-\nsquare tests (p < 0.05). \nResults or Findings: The median patient age was 39 years. A statisticall y \nsignificant correlation was found between adenomyos is dominant side and DIE \nimplant location (p = 0.019). Adenomyosis with no d ominant side or diffuse \nadenomyosis was not significantly associated with e ffecting of both \ncompartments by DIE implants (p = 0.275 and p = 0.1 02, respectively). The \ncorrelation between adenomyosis type and endometrio ma presence (p = \n0.390) or size >40 mm (p = 0.687) was not statistic ally significant. \nConclusion: To our knowledge, this is the first study evaluatin g the correlation \nbetween the dominant side of adenomyosis and DIE im plant location. Previous \nstudies indicated the relationship between the aden omyosis dominant side's \nand symptom severity, pregnancy loss. Our findings emphasize the need for \nfurther investigation into this subject. \nLimitations: The limitations are the small sample size and retro spective \nnature. \nFunding for this study: Not applicable \nEthics committee - additional information: Koc University Biomedical \nResearch Ethics Committee, Istanbul/Turkey \nAuthor Disclosures:  \nAfak Durur Karakaya: Nothing to disclose \nHande Özen Atalay: Nothing to disclose \nVugar Samadli: Nothing to disclose \nUzeyir Kalkan: Nothing to disclose \n \n \nMRI for endometriosis: ESUR Consensus for indicatio ns, reporting and \nclassifications \n*I. Thomassin-Naggara*¹, M. Dolciami², L. Chamie³, A. Guerra⁴, N. Bharwani⁵, \nS. Freeman⁶, P. Rousset⁷, L. Manganaro²; ¹Paris/FR, ²Rome/IT,  \n³São Paulo/BR, ⁴Lisbon/PT, ⁵Surbiton/UK, ⁶Cambridge/UK, ⁷Lyon/FR \n(isabellethomassin@gmail.com) \n \nPurpose or Learning Objective: The ESUR Research Committee appointed \ntwo chairs to supervise the development of the upda ted guidelines. \nMethods or Background: These guidelines are recommendations developed \nby the European Society of Urogenital Radiology (ES UR). A targeted literature \nsearch was performed to discover recent evidence co ncerning the imaging of \nendometriosis. The guidelines were formulated after  careful consideration of \nthe available literature by a group of internationa l experts. The panel included \n20 experts from 10 different countries, including 1 4 European centers and one \nnon-European institution. The methodology was based  on DELPHI process. \nEach item was classified as follows: “RECOMMENDED” (if agreement ≥ 80%); \n“OPTIONAL” (if agreement ≥ 70 % but < 80 %); or “NOT RECOMMENDED” (if \nconsensus was not reached, with < 70 % agreement). The survey was \ncomposed of three parts: 1) Indications, 2) MR prot ocol and lexicon 3) \nclassification/reporting. \nResults or Findings: Regarding indications, MRI is recommended when \nTVUS is inconclusive in diagnosing endometriosis or  negative, in a \nsymptomatic patient, before surgery or intervention al procedure or after \nsurgical treatment if symptoms persist. Regarding r eporting, MR classification \nis recommended, especially radiological score (dPEI ). “Patient centered care” \nis a key dimension of quality care. Good communicat ion with patients, as well \nas among the healthcare team, has the potential to improve care coordination, \nenhance safety and outcomes, increase patient satis faction, and reduce \nhealthcare costs. Standardized MR report and drawin g are crucial for \nimproving communication with patients and surgeons.  \nConclusion: In conclusion, the ESUR consensus on endometriosis \nemphasizes the importance of standardized reporting  and MR classifications to \nenhance communication between radiologists and the multidisciplinary team, \nas well as between radiologists and their patients.  This is crucial in managing a \ndisease where optimized communication is essential for providing patient-\ncentered and value-based care \n\n \n \nAbstract-based Programme \n \n 57  \nWednesday \nLimitations: Consensus paper \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMiriam Dolciami: Nothing to disclose \nPascal Rousset: Consultant: ZIWIG \nAdalgisa Guerra: Nothing to disclose \nSusan Freeman: Nothing to disclose \nIsabelle Thomassin-Naggara: Speaker: GE, Siemens, G uerbet, Hologic, \nCanon, Guebet, Bracco, GSD, Samsung, Fujifilm, Ince pto, ICAD \nLuciana Chamie: Nothing to disclose \nNishat Bharwani: Nothing to disclose \nLucia Manganaro: Nothing to disclose \n \n \nMRI for endometriosis: ESUR Consensus for protocol and lexicon \n*I. Thomassin-Naggara*¹, M. Dolciami², L. Chamie³, A. Guerra⁴, S. Freeman⁵, \nN. Bharwani⁶, P. Rousset⁷, L. Manganaro²; ¹Paris/FR, ²Rome/IT,  \n³São Paulo/BR, ⁴Lisbon/PT, ⁵Cambridge/UK, ⁶Surbiton/UK, ⁷Lyon/FR \n(isabellethomassin@gmail.com) \n \nPurpose or Learning Objective: The ESUR Research Committee appointed \ntwo chairs to supervise the development of the upda ted guidelines \nMethods or Background: The panel included 20 experts from 10 different \ncountries \nResults or Findings: Pre imaging fasting, the use of antiperistalsic age nts, a \nmoderately filled bladder and bowel preparation bef ore MRI are highly \nrecommended. Vaginal and rectal opacification shoul d be considered as an \noption. MR protocol must include multiplanar T2W an d T1W sequence and a \nsequence that visualizes the kidneys. Superficial e ndometriosis should be \ndescribed on T1FS as high signal intensity foci on the peritoneal surface. \nEndometriomas or implants should be described regar ding multiplicity, signal \nintensity, central or peripheral location and bilat erality. MR evaluation of deep \npelvic endometriosis should be performed using a co mpartmental division \ndefining two horizontal and vertical lines. A bladd er nodule should be \naccurately described according to location measured  and the distance to \nureteric orifice provided. Uterosacral ligament (US L) of ≤3mm is normal. A USL \nmust be considered as abnormal if a nodule or spicu lation is visible in at least \ntwo planes or if a bright T1W spot is detected. A p osterior vaginal wall or pouch \nof Douglas nodule should be described and measured.  The term rectovaginal \nseptum must be accurately used for sub peritoneal n odules. External \nadenomyosis should be described according to locati on and size. The \ndescription of a rectosigmoid nodule includes locat ion, number of nodules, \nlongitudinal extent, distance to the anal verge, an d wall thickening. The lateral \ncompartment includes the anterior distal round liga ment, mediolateral \nparametrium and posterolateral parametrium. Abdomin al wall nodules, \nileocaecal junction and appendiceal nodules, as wel l as sigmoid nodules, must \nbe systematically described. \nConclusion: This lecture will present the 10 ESUR statement reg arding MR \nEndometriosis protocol and lexicon \nLimitations: Consensus \nFunding for this study: Nond \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMiriam Dolciami: Nothing to disclose \nPascal Rousset: Consultant: ZIWIG \nAdalgisa Guerra: Nothing to disclose \nSusan Freeman: Nothing to disclose \nIsabelle Thomassin-Naggara: Speaker: GE, Siemens, G uerbet, Hologic, \nCanon, Guebet, Bracco, GSD, Samsung, Fujifilm, Ince pto, ICAD \nLuciana Chamie: Nothing to disclose \nNishat Bharwani: Nothing to disclose \nLucia Manganaro: Nothing to disclose \n \n \nMRI features of Accessory Cavitated Uterine Mass (A CUM) and Cystic \nAdenomyosis \n*Y. Qi*, Z. Zhao, X. Gao, P. Zhang, H. Xue, Y. He, L. Zhu; Beijing/CN \n(qiyafei19910603@163.com) \n \nPurpose or Learning Objective: To assess MRI features in accessory \ncavitated uterine mass (ACUM) and cystic adenomyosi s by evaluating lesion \nand uterine morphology. \nMethods or Background: The study included 16 ACUM patients (mean age \n25.8 years) and 14 cystic adenomyosis patients (mea n age 35.5 years) with \nsurgically and pathologically confirmed diagnoses. Pre-treatment pelvic MRI \nscans were performed, assessing lesion location, si ze, shape, cystic cavity, \nperipheral myometrial thickness, and uterine struct ure. \n \n \n \nResults or Findings: Patients presented with symptoms like dysmenorrhea \nand abdominal pain. Lesions appeared as layered str uctures with T2 \nhypointensity peripherally and T1 hyperintensity ce ntrally. ACUM lesions were \ntypically single and located at the fundal lateral wall, while cystic adenomyosis \nlesions varied in location, 5 adenomyosis lesions w ere irregular. ACUM \npatients had lesions ranging from 22mm to 45mm, wit h a central cystic cavity \ndiameter of 7mm to 36mm and peripheral myometrial t hickness of 5mm to \n10mm. Fifteen ACUM cases showed a concentric ring s tructure and the \njunctional zone was not significantly thickened. Cy stic adenomyosis lesions \nranging from 20mm to 64mm, with a cystic cavity dia meter of 4mm to 60mm \nand peripheral myometrial thickness of 2mm to 34mm.  5 cases showing \nuneven peripheral myometrium, only 1 case showing a  concentric ring \nstructure. 8 cases showing thickening junctional zo ne and 8 cases showing \nadenomyosis or fibroid presence, and 7 patients wit h ovarian or tubal \nendometriosis. \nConclusion: MRI is valuable for diagnosing ACUM and cystic aden omyosis. \nACUM patients are younger, with single, concentric lesions at the fundal lateral \nwall, rarely showing junctional zone thickening or adenomyosis-like changes. \nCystic adenomyosis is suggested by lesions on the p osterior wall, irregular \nshapes, large size, uneven shape, junctional zone t hickening, deep \nendometriosis, and fibroids. \nLimitations: ACUM and cystic adenomyosis are rare diseases, such  the \nnumber of cases are limited. \nFunding for this study: National High Level Hospital Clinical Research \nFunding, 2022-PUMCH-B-069 \nEthics committee - additional information: Peking Union Medical College \nHospital Ethics committee. \nAuthor Disclosures:  \nXin Gao: Nothing to disclose \nZichen Zhao: Nothing to disclose \nLan Zhu: Nothing to disclose \nHuadan Xue: Nothing to disclose \nYonglan He: Nothing to disclose \nYafei Qi: Nothing to disclose \nPeng Zhang: Nothing to disclose \n \n \nComparison of pregnancy outcomes following hysteros alpingography \n(HSG) with either water soluble contrast media (WSC M) or oil soluble \ncontrast media (OSCM) in routine clinical practice \n*A. P. Hemingway*, K. M. Wourms, S. Sudderuddin, E.  Kashef; London/UK \n \nPurpose or Learning Objective: Hysterosalpingography (HSG) represents \nthe mainstay of the imaging evaluation of the paten cy of fallopian tubes in the \ninvestigation of subfertility. An apparent associat ion of improved conception \nfollowing HSG has been recognised since the early 1 900’s. An randomised \ncontrolled trial (RCT )published in 2017 establishe d a significant difference \nbetween OSCM and WSCM with ongoing pregnancies in 3 9.7% following \nOSCM and 29.1% following WSCM .The purpose of this study was to \ndetermine if these results could be replicated in r outine clinical practice. \nMethods or Background: Imaging and medical records relating to 667 \nconsecutive HSGs performed between January 2020 and  December 2021 \nwere reviewed. For both WSCM and OSCM data was anal ysed with respect to \nage, primary of secondary sub-fertility, normal or abnormal HSGs, number of \npregnancies, mode of conception and pregnancy outco me. \nResults or Findings: Records relating to 667 women were reviewed, 76 los t \nto follow-up. 591 records, 498 following WSCM, 93 f ollowing OSCM were \nanalysed. A total of 354 conceptions were recorded,  291 (58.4%) following \nWSCM, 63 (67.7%) following OSCM. Spontaneous concep tions accounted for \n160 of the WSCM (32.1%) pregnancies and 38 (40.9%) of the OSCM \npregnancies. Time to conception was faster in the O SCM group and the \npregnancy outcomes were similar in both groups. The re were no observed \ndifferences in pregnancy rates following assisted r eproduction in the two \ngroups. \nConclusion: This retrospective study is in accordance with publ ished RCTs \nand meta analyses that an HSG with WSCM or OSCM is advantageous in \nincreasing the chances of spontaneous conception an d that OSCM has a \nsignificant advantage over WSCM. \nLimitations: This is a retrospective study. The OSCM group is sm aller than \nthe WSCM group. The study period and analysis is be ing extended to \nencompass HSGs undertaken in 2022 and 2023 \nFunding for this study: No external funding. HSGs standard of care. \nRetrospective data collection \nEthics committee - additional information: This retrospective study was \napproved IRAS Number 254621 \nAuthor Disclosures:  \nKatherine Marie Wourms: Nothing to disclose \nSiham Sudderuddin: Nothing to disclose \nAnne Patricia Hemingway: Nothing to disclose \nElika Kashef: Consultant: Rocket Medical \n \n \n\n \n \nAbstract-based Programme \n \n 58  \nWednesday \nComparing Inter-reader Agreement Between dPEI and # Enzian \nClassifications: Which Is the Better One? \n*A. Santonocito*, N. Bogveradze, J. Heine, T. H. He lbich, P. A. Baltzer,  \nR. Wenzl, P. Clauser; Vienna/AT \n(santonocitoambra@gmail.com) \n \nPurpose or Learning Objective: The deep pelvic endometriosis index (dPEI) \nand #Enzian classifications have been introduced to  facilitate reporting \nendometriosis patients. While studies show good int er-reader agreement with \n#Enzian, limited data exist for dPEI. This study ai med to assess the inter-\nreader agreement of both classifications. \nMethods or Background: In this retrospective, monocentric, IRB-approved \nstudy the pelvic MRIs performed in consecutive pati ents for suspected \nendometriosis from February 2018 to November 2023 w ere evaluated. Two \nreaders (R1, dedicated radiologist, 5 years of expe rience; R2, general \nradiologist, 2 years of experience) independently a ssessed the presence of \nlesions using dPEI and #Enzian classifications. The  extent of disease was then \ncategorized as mild (score ≤2), moderate (3-4), or severe (≥5). Inter-reader \nagreement was evaluated by using Cohen’s Kappa. \nResults or Findings: A total of 108 patients (mean age 32.7 ±7.1 years; range \n21-50 years old) were included in this study. Using  dPEI, R1 classified 22 \n(20.4%) patients with mild, 43 (39.8%) moderate, an d 43 (39.8%) severe \ndisease; while R2 classified 28 (25.9%) patients wi th mild, 43 (39.8%) \nmoderate, and 37 (34.3%) severe disease. The inter- reader agreement for \ndPEI was k=0.47 (p<0.001). Using #Enzian, R1 classi fied 10 (9.3%) patients \nas mild, 12 (11.1%) as moderate, and 86 (79.6%) as severe disease; while R2 \nclassified 21 (19.4%) patients as mild, 34 (31.5%) moderate, and 53 (49.1%) \nsevere disease. The inter-reader agreement for #Enz ian was k=0.31 \n(p<0.001). \nConclusion: Inter-reader agreement was moderate for dPEI and fa ir for \n#Enzian. This evidence suggests that dPEI, as a rad iologically based \nclassification, could be easier to use by non-speci alist radiologists. However, a \ndedicated training is essential to improve inter-re ader agreement in the \nevaluation of endometriosis. \nLimitations: Retrospective study; lacking of assessment of histo logy \nFunding for this study: None \nEthics committee - additional information: Number: 2057/2017 \nAuthor Disclosures:  \nPascal A.T. Baltzer: Nothing to disclose \nRene Wenzl: Nothing to disclose \nThomas H. Helbich: Nothing to disclose \nAmbra Santonocito: Nothing to disclose \nNino Bogveradze: Nothing to disclose \nJana Heine: Nothing to disclose \nPaola Clauser: Nothing to disclose \n \n \n15:00-16:00 Research Stage 3 \nResearch Presentation Session: Chest \nRPS 504 \nImaging of diffuse lung diseases: old and \nnew \n \nModerator \nM. Occhipinti; Florence/IT  \n(mariaelena.occhipinti@gmail.com) \n \n \nImaging and Clinical Features of Interstitial Lung Abnormalities (ILA) that \nPredict Progression to Idiopathic Pulmonary Fibrosi s (IPF) \n*T. Schnitzler*, J. H. Sohn; San Francisco, CA/US \n(tician.schnitzler@ucsf.edu) \n \nPurpose or Learning Objective: Interstitial lung abnormality (ILA) is often an \nincidental imaging finding, representing early or m ild fibrosis. While most cases \ndo not progress, some advance to idiopathic pulmona ry fibrosis (IPF), leading \nto severe outcomes. Accurate risk stratification of  ILA on non-contrast chest CT \nis crucial for guiding follow-up and early treatmen t. This study aims to improve \nstratification by identifying imaging and clinical features that predict \nprogression from ILA to IPF. \n \n \n \n \nMethods or Background: This retrospective case-control study included \npatients from a longitudinal ILD database: a low-ri sk ILA cohort (n = 525) and a \nhigh-risk ILA cohort (n = 221). Imaging features an alyzed included subpleural \nfibrotic reticulation, cranial extent of fibrosis, anterior lung involvement, and \nemphysema severity. Clinical variables included age  and gender. Statistical \nanalyses were conducted using chi-square tests for categorical variables and \nindependent t-tests for continuous variables. \nResults or Findings: The high-risk ILA cohort had significantly higher r ates of \nsubpleural fibrotic changes (78% vs. 36%, p < 0.001 ), cranial extent of fibrosis \n(61% vs. 14%, p < 0.001), anterior lung involvement  (86% vs. 37%, p < 0.001), \nand severe emphysema (48% vs. 39%, p < 0.001) compa red to the low-risk \ncohort. The high-risk cohort was also older (mean a ge 72.64 years vs. 70.65 \nyears, p = 0.020), with no significant gender diffe rence. \nConclusion: This study identifies key imaging and clinical pred ictors of ILA \nprogression to IPF, such as subpleural fibrotic cha nges, cranial extent of \nfibrosis, and older age. These findings could impro ve risk stratification, guiding \ntimely monitoring and interventions to enhance pati ent outcomes. Future \nresearch should validate these findings in larger, multi-center cohorts. \nLimitations: The main limitation is the retrospective single-cen ter design. \nFunding for this study: RSNA Research Fellow Grant 2024  \nSwiss Society for Radiology Research Grant 2023 \nBangerter-Rhyner Foundation, Basel, Switzerland \nEthics committee - additional information: This study is IRB approved  \n(17-22317). \nAuthor Disclosures:  \nTician Schnitzler: Nothing to disclose  \nJae Ho Sohn: Nothing to disclose \n \n \nIdentifying progressive pulmonary fibrosis on seria l CT: An international \nmulti-observer study \n*L. Sun*¹, M. A. Mestas Nuñez², J. Jacob¹, S. Piciu cchi³, L. Calandriello⁴,  \nA. Carvalho⁵, R. E. Ledda⁶, M. Chen¹, A. Devaraj¹; ¹London/UK, \n²Barcelona/ES, ³Forlì/IT, ⁴Rome/IT, ⁵Porto/PT, ⁶Parma/IT \n(logan.sun@doctors.org.uk) \n \nPurpose or Learning Objective: To evaluate the performance and agreement \nof thoracic radiologists and interstitial lung dise ase (ILD) physicians in \nidentifying progressive pulmonary fibrosis on seria l CT scans in patients \nwithout idiopathic pulmonary fibrosis (IPF). \nMethods or Background: 100 patients with various non-IPF fibrotic lung \ndiseases (median age, 64 years [range, 36 to 85]; m ale, n=40) had serial CTs \nobtained 6 to 24 months apart, which were reviewed independently by 12 ILD \nphysician and thoracic radiologist readers blinded to clinical data. CTs were \nreviewed side-by-side and categorised as one of two  groups: Stable Disease \nor Progressive Fibrosis. Groups were compared using  contemporary relative \nchange in percentage predicted forced vital capacit y (FVC), per reader and \nacross the cohort, and analysed by Mann-Whitney U t est and mixed-effects \nmodelling. Interobserver agreement was assessed usi ng intraclass correlation \ncoefficient (ICC). \nResults or Findings: Preliminary data are presented. Mean FVC change for  \nall patients was -6.28% (SD, 14.9). For individual readers, there was a \nsignificant difference in median FVC decline betwee n corresponding \nProgressive Fibrosis versus Stable Disease CT group s (range, -6.84% to -\n11.44%, p=<0.001–0.015). For the whole reader cohor t, mean FVC decline \nwas significantly greater in Progressive Fibrosis v ersus Stable Disease on CT \n(-10.70%, 95% CI [-11.89%, -9.50%] versus -1.47%, 9 5% CI [-2.78%, -0.15%]). \nInterobserver agreement was moderate (ICC = 0.501, 95% CI [0.420, 0.588]). \nConclusion: Among specialist thoracic radiologists and ILD phys icians, visual \nevaluation of serial CT scans is a valuable method for determining progressive \nfibrosis in non-IPF fibrotic lung diseases, judged against contemporary FVC \ndecline, though interobserver agreement remains mod erate. \nLimitations: Single-centre retrospective study \nFunding for this study: Nil sought \nEthics committee - additional information: Prior IRAS approval for \nretrospective research within the Royal Brompton Ho spital radiology \ndepartment \nAuthor Disclosures:  \nSara Piciucchi: Nothing to disclose \nRoberta Eufrasia Ledda: Nothing to disclose  \nAndre Carvalho: Nothing to disclose \nJoseph Jacob: Nothing to disclose  \nMitchell Chen: Nothing to disclose \nAnand Devaraj: Nothing to disclose \nLogan Sun: Nothing to disclose \nLucio Calandriello: Nothing to disclose \nMarcos Alejandro Mestas Nuñez: Nothing to disclose \n \n \n \n \n \n\n \n \nAbstract-based Programme \n \n 59  \nWednesday \nDiagnostic Delay of Lung Cancer in Interstitial Lun g Disease \n*T. Schnitzler*, J. H. Sohn; San Francisco, CA/US \n(tician.schnitzler@ucsf.edu) \n \nPurpose or Learning Objective: Interstitial lung disease (ILD) patients have \nan increased risk of lung cancer, but detection is challenging due to \nbackground fibrosis, leading to diagnostic delays. There is limited research on \nlung cancer in ILD, particularly regarding diagnost ic delays. This study aims to \nanalyze delayed lung cancer diagnoses in ILD patien ts, including tumor stage \nat diagnosis, growth rates, treatment regimens, and  outcomes. \nMethods or Background: This retrospective study included ILD patients with  \nconcomitant lung cancer (pathology proven or >50% r adiologically suspected) \nfrom two referral centers. A thoracic radiologist r e-reviewed chest CTs to \ndetermine the earliest visible and callable lesion time, when it was first deemed \nsuspicious, and its growth rate. Tumor staging, tre atment regimens, and \noutcomes were analyzed. Survival curves were genera ted using the Kaplan-\nMeier method, comparing median survival times betwe en delayed and non-\ndelayed cancer cases with the log-rank test. \nResults or Findings: Seventy-seven cases of concurrent ILD and lung canc er \nwere identified (53 pathology proven, 24 radiologic ally presumed). Delayed \ndiagnoses occurred in 47% (36/77) of cases, with an  average delay of 3.42 \nyears. These delayed cases had a mean annual growth  rate of 293% and a \nmean doubling time of 3.3 years. An additional 5% ( 4/61) were diagnosed post-\nlung transplant. The median survival time was 1269 days for early detection \nversus 867 days for delayed detection. However, the  difference in survival was \nnot statistically significant (p = 0.80). \nConclusion: This study found that 52% of lung cancer cases in I LD had \ndelayed diagnoses, with an average delay of 3.42 ye ars. Despite the delays, \nthere was no significant difference in mortality be tween early and delayed \ndetection cases. \nLimitations: The main limitation is the retrospective study desi gn. \nFunding for this study: Swiss Society for Radiology Research Grant 2023 \nBangerter-Rhyner Foundation, Basel, Switzerland \nEthics committee - additional information: This study was approved by the \nlocal IRB (17-22317) \nAuthor Disclosures:  \nTician Schnitzler: Nothing to disclose \nJae Ho Sohn: Nothing to disclose \n \n \nRadiological assessment of bronchial and arterial d imensions and mucus \nplug presence in 640 bronchiectasis patients: insig hts from the EMBARC \nregistry \n*Y. Chen*¹, A. Pieters¹, E-R. Andrinopoulou¹, S. Al iberti², M. Loebinger³,  \nP. Ciet¹, J. Chalmers⁴, H. A. W. M. Tiddens¹, On Belalf Of Embarc Study \nGroup⁴; ¹Rotterdam/NL, ²Humanitas Research Hospital, Mila n/IT, ³London/UK, \n⁴Dundee/UK \n(y.chen.1@erasmusmc.nl) \n \nPurpose or Learning Objective: Key features of bronchi in bronchiectasis \ndisease are irreversible widening, wall thickening and mucus plugging. The \nbronchiectasis registry EMBARC lacks currently obje ctive quantitative metrics \nfor these features. The aim of our study was to ana lyse EMBARC chest CTs \nusing an AI-based algorithm measuring bronchus and artery (BA) dimensions \nand ratios and counting mucus plugs (MP). \nMethods or Background: 885 CTs from eight EMBARC centres were \nretrospectively collected for automatic analysis us ing LungQ (Thirona, The \nNetherlands), which segments the bronchial tree and  identifies segmental (G0) \nand distal (G1,2,3…) generations. For each BA-pair,  the following dimensions \nare computed: diameters of bronchial outer edge (Bo ut), inner edge (Bin), and \nartery (A), and wall thickness (Bwt) and the follow ing BA-ratios: Bout/A, Bin/A, \nBwt/A, and bronchial wall area/outer area (Bwa/Boa) . Cut-offs for mild and \nsevere bronchial widening are Bout/A>1.1 and >1.5, respectively and for \nthickening (Bwt/A>0.14). The MP analysis automatica lly segments the \nbronchial tree, detects the total number and volume  of MP. \nResults or Findings: 640 CTs were successfully analysed, identifying 141 ,978 \nBA-pairs from G0 until G29 (222 BA-pairs per CT). B out/A>1.1 or >1.5 were \nobserved in 73% and 39% of all BA-pairs, respective ly. Bwt/A>0.14 was \nobserved in 49% of all BA-pairs. The median(IQR) Bo ut/A, Bin/A, Bwt/A, and \nBwa/Boa for G1-6 were 1.34(1.07, 1.72), 1.04(0.81, 1.35), 0.13(0.1, 0.2). MP \nwere found in 83% of CTs, with a median number of 8  plugs and a median \nvolume of 0.44mL per CT. \nConclusion: Our study demonstrates the capability of AI-based a lgorithms to \nmeasure BA-dimensions and detect mucus plugs on che st CT scans of \nbronchiectasis patients. Our findings show widespre ad but heterogeneous \nbronchial widening and thickening, along with the p resence of mucus plugs, \nindicative of active infection and/or inflammation.  \nLimitations: Retrospective study \n \n \n \nFunding for this study: Supported by the Innovative Medicines Initiative an d \nThe European Federation of Pharmaceutical Industrie s and Associations \ncompanies under the European Commission–funded Hori zon 2020 Framework \nProgram and by Inhaled Antibiotic for Bronchiectasi s and Cystic Fibrosis (grant \n115721). EMBARC3 is funded by the European Respirat ory Society through \nthe EMBARC3 clinical research collaboration. EMBARC 3 is supported by \nproject partners Armata, AstraZeneca, Boehringer In gelheim, Chiesi, CSL \nBehring, Grifols, Insmed, Janssen, Lifearc, and Zam bon. J.D.C. is supported \nby the GlaxoSmithKline/Asthma and Lung UK Chair of Respiratory Research. \nEthics committee - additional information: The study received central \nethical approval from the Multicentre Research Ethi cs Committee in the UK on \nJan 8, 2015 (14/SS/1101) and the study is sponsored  by the University of \nDundee, Dundee, UK. \nAuthor Disclosures:  \nStefano Aliberti: Grant Recipient: Stefano Aliberti  has received grants or \ncontracts fees from INSMED incorporated, CHIESI, Fi sher&Paykel; and \nreceived consulting fees from GSK, McGRAW HILL, Zam bon, AstraZeneca, \nCSL Behring GmbH, Moderna, Chiesi, MSD Italia, Pysi oassist SAS, \nGlaxoSmithKline \nEleni-Rosalina Andrinopoulou: Nothing to disclose \nJames Chalmers: Grant Recipient: James D Chalmers h as received research \ngrants from AstraZeneca, Boehringer Ingelheim, Glax oSmithKline, Gilead \nSciences, Grifols, Novartis, Insmed and Trudell; an d received consultancy or \nspeaker fees from Antabio, AstraZeneca, Boehringer Ingelheim, Chiesi, \nGlaxoSmithKline, Insmed, Janssen, Novartis, Pfizer,  Trudell and Zambon. \nPierluigi Ciet: Grant Recipient: Pierluigi Ciet has  received grants or contracts \nfees from NOW-Dutch Research Councel and Horizon Pa thfinder; and has \nreceived payment or honoraria for lectures, present ations etc from Chiesi and \nVertex. \nMichael Loebinger: Grant Recipient: Michael Loebing er has received \nconsulting fees from Armata, 30T, AstraZeneca, Pari on, Ismed, Chiesi, \nZambon, Electromed, Recode and Boehringer Ingelheim ; and received \npayment or honoraria for lectures, presentations et c from Ismed. \nHarm A W M Tiddens: Employee: Harm AWM Tiddens has received research \ngrants from Thirona as Chief Medical Officer and Em eritus Professor \nErasmusMC Sophia; and has received payment or honor aria for lectures, \npresentations etc from Vertex. \nYuxin Chen: Nothing to disclose \nOn Belalf Of Embarc Study Group: Nothing to disclos e \nAngelina Pieters: Nothing to disclose \n \n \nComputed Tomography-Derived Quantitative Imaging Bi omarkers enable \nthe prediction of survival and disease severity in patients with Systemic \nSclerosis \n*M. M. Sieren*¹, H. Graßhoff¹, L. Berkel¹, G. Rieme kasten¹, F. Nensa²,  \nR. Hosch², J. Barkhausen¹, R. Klöckner¹, F. Wegner¹ ; ¹Lübeck/DE, ²Essen/DE \n(malte.sieren@uksh.de) \n \nPurpose or Learning Objective: Systemic Sclerosis (SSc) is a complex \nconnective tissue disorder with variable disease pr ogression and outcome. \nWhile chest CT imaging is recommended in all patien ts to evaluate interstitial \nlung disease, AI-driven body composition analysis ( BCA) can further enhance \nradiological assessment by providing quantitative i maging biomarkers. This \nstudy aims to assess BCA's ability to predict survi val, complications, and \ndisease severity on chest CT. \nMethods or Background: CT scans were obtained from a prospectively \nmaintained cohort of 452 SSc patients, including 12 8 with at least one CT scan \nand 35 patients with up to three follow-up exams. T he follow-up period \naveraged 36.5±4.5 months. An AI-based 3D BCA algori thm measured muscle \nvolume, adipose tissue compartments, and bone miner al density. BCA \nParameters were evaluated in relation to clinical, laboratory, and functional \ndata on baseline and follow-up scans. Survival pred iction was performed using \nregression analysis, comparing models based on BCA,  BMI, and clinical \nparameters. \nResults or Findings: The BCA model outperformed BMI and clinical models in \npredicting survival (BCA AUC=0.74, BMI AUC=0.49, cl inical parameters \nAUC=0.53). Including longitudinal BCA data further improved the model's AUC \nto 0.82. Altered BCA parameters were linked to incr eased odds ratios [with \n95% confidence interval] for complications like acr al ulcers (1.7 [1.1-1.9]), \ninterstitial lung disease (2.1 [1.4-4.4]), cardiac (2.0 [1.3-3.0]) and \ngastrointenstinal manifestations (1.6 [1.4-1.9], al l p<0.05). \nConclusion: This study highlights that quantitative body compos ition \nbiomarkers outperform established parameters in pre dicting survival and \nspecific disease manifestations. These findings pro vide a blueprint how \nradiological assessment can transform from primaril y qualitative assessment to \nincluding previously unavailable quantitative data leading to more personalized \npatient care, potentially improving outcomes for SS c patients. \nLimitations: The study's single-center design and small sample s ize may limit \ngeneralizability, and variations in CT quality coul d affect AI-based BCA \naccuracy. \n \n\n \n \nAbstract-based Programme \n \n 60  \nWednesday \nFunding for this study: None. \nEthics committee - additional information: Study/protocol number: AZ 22-\n289 \nAuthor Disclosures:  \nLennart Berkel: Nothing to disclose \nFranz Wegner: Nothing to disclose \nRoman Klöckner: Nothing to disclose \nGabriela Riemekasten: Nothing to disclose \nJörg Barkhausen: Nothing to disclose \nMalte Maria Sieren: Nothing to disclose \nHanna Graßhoff: Nothing to disclose \nRené Hosch: Nothing to disclose \nFelix Nensa: Nothing to disclose \n \n \nDevelopment of imaging protocol and radiomics-based  nomogram for \nassessing lesion reversibility in connective tissue  disease-associated \ninterstitial lung disease \nY. Zhang, *Y. Wang*, X. Yu, J. Wei, H. Wu; Shanghai /CN \n(yu.wang@philips.com) \n \nPurpose or Learning Objective: To develop an imaging protocol for \nassessing lesion reversibility and a radiomics-base d nomogram for predicting \nlesion reversibility in connective tissue disease-a ssociated interstitial lung \ndisease (CTD-ILD). \nMethods or Background: A retrospective study categorized CTD-ILD patients \ninto training, internal and external validation coh orts. An imaging protocol of \nserial chest CT scans for assessing lesion reversib ility was developed, \nclassifying patients as completely reversible (CR) and non-CR groups based \non CT lesion changes. Lesions were evaluated using morphological CT \nfeatures and radiomics signatures at the lung-zone level. Visual, radiomics, \nand combined nomogram models were developed and com pared through \nreceiver operating characteristic (ROC) curve analy sis. \nResults or Findings: Among 153 patients with 575 affected lung zones, a \nfive-feature radiomics signature significantly corr elated with ILD lesion \nreversibility. The radiomics model showed robust di scrimination, comparable to \nthe visual model in the validation cohorts (interna l: 0.77, 95% CI: [0.68, 0.86] \nversus 0.87, 95% CI: [0.81, 0.94], p=0.056; externa l: 0.73, 95% CI: [0.66, 0.79] \nversus 0.78, 95% CI: [0.72, 0.84], p=0.20), and inf erior to the visual model in \nthe training cohort (0.72, 95% CI: [0.66, 0.79] ver sus 0.82, 95% CI: [0.77, \n0.87], p=0.02). The combined nomogram model outperf ormed the visual model \nalone in the training and external validation cohor ts (0.86, 95% CI: [0.81, 0.91], \np=0.03; 0.82, 95% CI: [0.77, 0.87]; p=0.048). \nConclusion: An imaging protocol was established for assessing l esion \nreversibility in CTD-ILD. The radiomics signature p rovided a quantitative \napproach to predict lesion reversibility. The combi ned nomogram improved the \npredictive accuracy beyond morphological features a lone. \nLimitations: First, clinical information were incomplete and not  included in the \npredictive model. Future research incorporating mor e clinical information is \nneeded. Second, the manual segmentation of lung zon es might introduce bias \nacross various scans. \nFunding for this study: None \nEthics committee - additional information: Ethics approval (No. LY2023-\n019-B) was granted by the institutional review boar d (IRB) of Shanghai \nJiaotong University, School of Medicine, Renji Hosp ital. The IRB waived \ninformed consent requirement for this retrospective  study. \nAuthor Disclosures:  \nYu Wang: Nothing to disclose  \nJiaxu Wei: Nothing to disclose \nYing Zhang: Nothing to disclose \nXiao Yu: Nothing to disclose \nHuawei Wu: Nothing to disclose \n \n \nPleural Effusion as a Prognostic Indicator in COVID -19: A nationwide \nMulticenter Analysis \n*A. M. Bucher*¹, E. Frodl¹, F. G. Meinel², M. M. Si eren³, M. A. Fink⁴,  \nM. S. May⁵, M. S. Kim⁶, T. Vogl¹, A. Surov⁷; ¹Frankfurt/DE, ²Rostock/DE, \n³Lübeck/DE, ⁴Heidelberg/DE, ⁵Erlangen/DE, ⁶Essen/DE, ⁷Minden/DE \n \nPurpose or Learning Objective: This study evaluates the prognostic \nsignificance of pleural effusion (PE) in COVID-19 p atients across 13 German \ncenters, part of the RACOON (Radiological Cooperati ve Network) project. We \naimed to assess the relationship between PE and key  clinical outcomes, in a \nlarge multicentre study. \nMethods or Background: In this retrospective study, 1183 COVID-19 patients  \n(29.3% women, 70.7% men) underwent chest CT to asse ss the presence, \nvolume, and density of PE. We analyzed associations  between PE and clinical \noutcomes including 30-day mortality, ICU admission,  and mechanical \nventilation. We used univariable and multivariable regression analyses, \nadjusting for confounders such as the COVID-19 CT s everity score. \nResults or Findings: PE was identified in 31.5% of patients. A significa nt \ncorrelation was found between PE and 30-day mortali ty (47.5% in non-\nsurvivors vs. 27.3% in survivors, p<0.001). PE pres ence independently \npredicted mortality with a hazard ratio (HR) of 2.2 2 (95% CI 1.65-2.99, \np<0.001). However, PE volume and density were not s ignificantly associated \nwith mortality. ICU admission was necessary in 46.8 % of patients, and 26.7% \nrequired mechanical ventilation. PE presence was al so linked to ICU admission \nand ventilation but not its volume or density. \nConclusion: Pleural effusion is a significant independent predi ctor of 30-day \nmortality in COVID-19 patients, irrespective of its  volume or density. These \nfindings underscore the importance of including PE detection in routine CT \nassessments to enhance clinical decision-making and  patient care. \nLimitations: This retrospective study was limited to German tert iary care \ncenters, which may not represent other settings. \nFunding for this study: Funded by „NUM 2.0“ (FKZ: 01KX2121) \nEthics committee - additional information: IRB approval for this \nretrospective multi centre study was obtained (20-7 19). \nAuthor Disclosures:  \nEric Frodl: Nothing to disclose \nAlexey Surov: Nothing to disclose \nThomas Vogl: Nothing to disclose \nMoon Sung Kim: Nothing to disclose \nMatthias Stefan May: Nothing to disclose \nMatthias Alexander Fink: Nothing to disclose \nFelix G. Meinel: Nothing to disclose \nMalte Maria Sieren: Nothing to disclose \nAndreas Michael Bucher: Nothing to disclose \n \n \nQIP Are Chest Radiographs Being Conducted in Accord ance with the \nBritish Thoracic Society Recommendations for Adults  Diagnosed with \nCommunity Acquired Pneumonia? \n*M. Mobini*, A. Nehvi, S. Buckingham, L. Mills; Ste venage/UK \n \nPurpose or Learning Objective: This audit evaluates whether follow-up chest \nradiographs are being performed for adults diagnose d with Community-\nAcquired Pneumonia (CAP) according to British Thora cic Society (BTS) \nguidelines. A gap was identified when many patients  did not receive a follow-\nup chest X-ray within six weeks, prompting an audit  to identify barriers and \ngaps in care. \nMethods or Background: CAP affects 0.5% to 1% of UK adults annually and \ncarries a mortality risk of 5-14%. Follow-up X-rays  are crucial to ensure the \nresolution of pneumonia and to exclude underlying c onditions such as lung \ncancer. The first cycle retrospectively reviewed 50  adult patients diagnosed \nwith CAP between November 2023 and March 2024. The second cycle \nrepeated the review from March to July 2024, after implementing several \ninterventions. Data were collected to assess compli ance with follow-up X-ray \nrecommendations, virtual Pneumonia clinic (VPC) ref errals, and Casualty \n(CAS) alerts. \nResults or Findings: In the first cycle, only 32.6% of patients received  follow-\nup X-rays within six weeks, with 67.4% failing to c omply. Among those \ndischarged from the emergency department, 67.8% did  not have a follow-up X-\nray. Referrals to the Virtual Pneumonia Clinic (VPC ) were low (23.2%), and \nonly 5% of reports included a CAS alert. Following interventions aimed at \nraising awareness among doctors, improving document ation, enhancing \npatient education, and increasing CAS alerts, the s econd cycle showed \nsignificant improvement. Compliance with follow-up X-rays increased to 70%, \nwith VPC referrals rising to 77% and CAS alerts rea ching 73%. The \ncompliance rate for follow-up X-rays among patients  discharged from \nemergency care increased from 32.2% to 65%. \nConclusion: A marked improvement in adherence to BTS guidelines  after \ntargeted interventions, lead to more follow-up X-ra ys, VPC referrals, and CAS \nalerts. \nLimitations: 1) Retrospective Design 2)Small sample size 3) Sing le-Centered \nAudit \nFunding for this study: East and North Hertfordshire NHS Trust. \nEthics committee - additional information: The study was approved by the \ntrust ethics committee and audit department \nAuthor Disclosures:  \nMoein Mobini: Author: second author \nAabid Nehvi: Author: 1  \nSusan Buckingham: Consultant: 2 Author: 2  \nLauren Mills: Author: 3 \n \n \n \n\n \n \nAbstract-based Programme \n \n 61  \nWednesday \n15:00-16:00 Research Stage 4 \nResearch Presentation Session: Cardiac \nRPS 503 \nCardiac CT: plaques and beyond \n \nModerator \nD. Suchá; Utrecht/NL  \nAuthor Disclosures:  \nDominika Suchá: Research Grant/Support: Philips Hea lthcare Research \nSupport received by the Department of Radiology, UM C Utrecht, NL \n \n \nCT coronary calcium scoring to detect obstructive c oronary artery \ndisease in primary care patients with non-typical c hest pain \nM. Y. Koopman¹, R. Willemsen², B. Kietselaer³, P. M . A. Van Ooijen¹,  \nJ-W. Gratama⁴, R. Braam⁴, R. Van Bruggen⁴, P. Van Der Harst⁵,  \n*R. Vliegenthart*¹; ¹Groningen/NL, ²Maastricht/NL, ³Rochester, MN/US, \n⁴Apeldoorn/NL, ⁵Utrecht/NL \n(r.vliegenthart@umcg.nl) \n \nPurpose or Learning Objective: Computed Tomography coronary calcium \nscoring (CT-CCS) has higher sensitivity for detecti on of obstructive coronary \nartery disease (OCAD) than exercise electrography ( x-ECG), but its utility as \nan initial diagnostic test in primary care remains unclear. This pilot study \ncompares CT-CCS results with x-ECG results in prima ry care and assesses \npatients’ perspectives. \nMethods or Background: Thirty-eight primary care offices participated in t his \nstudy. After cluster randomisation, 19 offices refe rred patients with atypical \nangina pectoris or non-specific thoracic complaints  for CT-CCS and 19 offices \nused x-ECG as the primary test (standard care). Cli nical data were collected \nusing electronic patient records, and patients’ per spectives on the diagnostic \ntest were assessed through a questionnaire. Outcome  measures included CAD \ndiagnosis, initiation of cardiovascular risk manage ment (CVRM), and patient \nsatisfaction. \nResults or Findings: In total, 101 patients were included. In 25 patient s \nundergoing X-ECG, one (4%) had a positive test resu lt and received CVRM, \nbut no patients were diagnosed with obstructive CAD . CT-CCS was performed \nin 76 patients. 17 CT-CCS patients (23%) had a posi tive test result (calcium \nscore >100), and 14 (19%) received CVRM. Obstructiv e CAD was diagnosed \nin four CT-CCS patients (5.3%). Of CT-CCS patients,  31 (43%) perceived the \ntest as ‘very easy’ compared to none of the x-ECG p atients. \nConclusion: CT-CCS is a promising diagnostic tool in primary ca re for the \ndetection of obstructive CAD, offering a more patie nt-friendly experience \ncompared to x-ECG. \nLimitations: Small cohort, especially in the x-ECG arm, and low OCAD rate. A \nfew patients received the test result before comple ting the questionnaire. \nBaseline cardiovascular related risk factors were i nconsistently reported in \nelectronic patient records. \nFunding for this study: Funding was received from the Dutch Heart \nFoundation (Hartstichting, grant number: CVON2017-1 4). \nEthics committee - additional information: The Medical Ethical Committee \nof the University Medical Center of Groningen appro ved CONCRETE (number \n2018/404). \nAuthor Disclosures:  \nRobert Willemsen: Nothing to disclose \nPim Van Der Harst: Nothing to disclose \nRichard Braam: Nothing to disclose \nPeter M.A. Van Ooijen: Nothing to disclose \nMoniek Yvonne Koopman: Nothing to disclose \nRozemarijn Vliegenthart: Research/Grant Support: Si emens Healthineers \nBas Kietselaer: Nothing to disclose \nRykel Van Bruggen: Nothing to disclose \nJan-Willem Gratama: Nothing to disclose \n \n \nThin-slice non-contrast CT detects prognostically r elevant calcified \nplaques missed by conventional calcium scoring \n*F. Yavuz*, F. Biavati, K. Schulze, S. Tsogias, B. Föllmer, A-M. Stantien,  \nM. Bosserdt, M. Dewey; Berlin/DE \n \nPurpose or Learning Objective: To evaluate whether thin-slice non-contrast \nCT (NCCT) can detect prognostically relevant corona ry plaques missed by \nconventional 3.0-mm reconstructions. \nMethods or Background: This study included 141 patients from the CAD-Man \ntrial [NCT00844220] (mean age 60.77 ± 11.06 years, 55% female) with \navailable thin-slice NCCT (0.5-mm). The Agatston me thod was used to detect \ncalcified plaques. Sensitivity and specificity for the detection of calcified \nplaques were calculated using CT angiography (CTA) as the reference \nstandard. Lesion- and patient-level statistics were  calculated for plaque volume \nparameters. Prognostic relevance was assessed by ev aluating plaque \nprogression rates for plaques detected only on thin -slice reconstructions, using \nmedian 10-year follow-up data when available. \nResults or Findings: In total 551 calcified plaques were detected. Thin- slice \nNCCT showed a higher sensitivity (91.83%; 506/551) for detecting coronary \ncalcified plaques compared to 3.0-mm reconstruction  (82.76%; 456/551), \nalthough standard reconstructions showed an overall  per-patient increased \nmean calcified plaque volume (197.22 mm3 ± 330.05 m m3) compared to thin-\nslice NCCT (162.65 mm3 ± 284.1 mm3). Conversely, we  observed a slightly \nlower specificity (97.23%; 492/506) for thin-slice NCCT compared to standard \nreconstructions (99.56%; 454/456). Coronary calcifi ed plaques missed in \nstandard reconstructions were smaller in volume (2. 67 mm3 ± 1.47 mm3) \ncompared to all detected plaques (20.68 mm3 ± 25.56  mm3). Missing calcified \nplaques on standard reconstructions would have led to the omission of 9 out of \n141 patients (6.4%). Additionally, plaques only ide ntified on thin-slice NCCT at \nbaseline were clearly visible at follow-up, with an  average 7.9-fold increase in \nvolume. \nConclusion: Coronary calcified plaques detected exclusively on thin-slice \nNCCT reconstructions showed increased plaque progre ssion rates compared \nto plaques detected in conventional calcium scoring . \nLimitations: This study involved patients from a single-centre, and 10-year \nfollow-up data were not available for all patients with missed plaques. \nFunding for this study: This study was funded by a grant of the Heisenberg \nprogramme. \nEthics committee - additional information: The study was approved by \nethics committee at Charité (EA1/124/23). \nAuthor Disclosures:  \nKenrick Schulze: Nothing to disclose \nFerhat Yavuz: Nothing to disclose \nAnne-Marieke Stantien: Nothing to disclose \nMarc Dewey: Board Member: M.D. is European Society of Radiology (ESR) \nPublications Chair (2022-2025); the opinions expres sed in this presentation are \nthe author’s own and do not represent the view of E SR Grant Recipient: EU \n(EC-GA 603266 in HEALTH.2013.2.4.2-2) DFG (DE 1361/ 14-1, DE 1361/18-1, \nBIOQIC GRK 2260/1, Radiomics DE 1361/19-1 [42822292 2] and 20-1 \n[428223139] in SPP 2177/1), GUIDE-IT (DE 1361/24-1) , Berlin University \nAlliance (GC_SC_PC 27), G-BA (01NVF23002), Berlin I nstitute of Health \n(Digital Health Accelerator) Patent Holder: Patent on fractal analysis of \nperfusion imaging (jointly with Florian Michallek, EPO 2022 EP3350773A1, and \nUSPTO 2021 10,991,109, approved) Author: Cardiac CT  (Springer Nature) \nResearch/Grant Support: Siemens, General Electric, Philips, Canon Other: \nHands-on cardiac CT courses (www.ct-kurs.de) Instit utional research \nagreements: Siemens, General Electric, Philips, Can on. Patent on fractal \nanalysis of perfusion imaging (jointly with Florian  Michallek, EPO 2022 \nEP3350773A1, and USPTO 2021 10,991,109, approved) M .D. is European \nSociety of Radiology (ESR) Publications Chair (2022 -2025); the opinions \nexpressed in this presentation are the author’s own  and do not represent the \nview of ESR \nSotirios Tsogias: Nothing to disclose \nFederico Biavati: Nothing to disclose \nMaria Bosserdt: Nothing to disclose \nBernhard Föllmer: Nothing to disclose \n \n \nAssociation of features derived from segment-level coronary artery \ncalcium scoring with major adverse cardiovascular e vents: A multicentre \nstudy \n*S. Tsogias*, B. Föllmer, M. Mohamed, F. Biavati, K . Schulze, M. Bosserdt, \nM. Dewey; Berlin/DE \n(sotirios.tsogias@charite.de) \n \nPurpose or Learning Objective: To investigate the association of segment-\nlevel coronary artery calcium (CAC) scoring derived  features with major \nadverse cardiovascular events (MACE) compared to ve ssel-based and overall \nCAC scoring. \nMethods or Background: This subanalysis of the multicentre DISCHARGE \ntrial (NCT02400229) included a total of (N = 1446) patients (mean age 59.9 ± \n10.2 years) who had received a calcium scoring CT a nd were followed up over \na median timespan of 3.5 years. The definition of M ACE included nonfatal \nstroke, nonfatal myocardial infarction and cardiova scular death. Associations \nwith MACE were examined for proximal (LM and proxim al segments of the \nLAD, LCX and RCA) versus non-proximal calcification s and the total number of \nsegments containing calcifications out of 19 (0: No  calcification; 1: Limited; 2-9: \nModerate, ≥ 10: Extensive). CAC scores were obtained both manu ally and \nusing deep learning-based scoring methods. Analysis  was performed using \nCox proportional hazards regression adjusting for a ge, sex, body-mass-index, \ndiabetes, dyslipidemia, hypertension, family histor y, smoking status and \nAgatston categories (< 400; ≥ 400) with hazard ratios (HR) and 95% \nconfidence intervals (CI). \n\n \n \nAbstract-based Programme \n \n 62  \nWednesday \nResults or Findings: During follow-up a total of 31 MACE occurred. Proxi mal \nvessel calcifications were associated with higher r isk for MACE (HR = 3.9, 95% \nCI [1.02, 14.5], p < .05). A moderate number of cal cified segments [2-9 \nsegments] was also associated with an increased ris k for MACE (HR = 4.2, \n95% CI [1.08, 16.1], p < .05). \nConclusion: Proximal vessel calcification as well as moderate s egment \ncalcification were associated with a higher risk fo r MACE. \nLimitations: Due to the low number of MACE in this study populat ion 2.1% (31 \nof 1446) the overall predictive value of the segmen t level CAC scoring may \nhave been underrepresented. \nFunding for this study: This work was funded by the German Research \nFoundation through the graduate program BIOQIC (GRK 2260, project-ID: \n289347353) and the DISCHARGE project (603266-2, HEA LTH-2012.2.4.-2) \nfunded by the FP7 Program of the European Commissio n. \nEthics committee - additional information: The study was approved by The \nGerman Federal Office for Radiation Protection and the local or national \nauthorities at each trial site. \nAuthor Disclosures:  \nKenrick Schulze: Nothing to disclose \nMahmoud Mohamed: Nothing to disclose \nMarc Dewey: Grant Recipient: EU (EC-GA 603266 in HE ALTH.2013.2.4.2-2) \nDFG (DE 1361/14-1, DE 1361/18-1, BIOQIC GRK 2260/1,  Radiomics DE \n1361/19-1 [428222922] and 20-1 [428223139] in SPP 2 177/1), GUIDE-IT (DE \n1361/24-1), Berlin University Alliance (GC_SC_PC 27 ), Berlin Institute of \nHealth (Digital Health Accelerator) Patent Holder: Patent on fractal analysis of \nperfusion imaging (jointly with Florian Michallek, EPO 2022 EP3350773A1, and \nUSPTO 2021 10,991,109, approved) Other: Hands-on ca rdiac CT courses \n(www.ct-kurs.de) Other: Institutional research agre ements: Siemens, General \nElectric, Philips, Canon. Other: M.D. is European S ociety of Radiology (ESR) \nPublications Chair (2022-2025); the opinions expres sed in this presentation are \nthe author’s own and do not represent the view of E SR. Author: Cardiac CT \n(Springer Nature) \nSotirios Tsogias: Nothing to disclose \nFederico Biavati: Nothing to disclose \nMaria Bosserdt: Research/Grant Support: Received fu nding from EU-FP7 \nFramework Program (DISCHARGE EU FP EC-GA 603266) \nBernhard Föllmer: Nothing to disclose \n \n \nPrognostic value of semi-quantitative cCTA scores \n*E. Bruno*, A. Bettinelli, V. Morrone, A. Colombo, C. Gnasso, F. Pisu,  \nD. Vignale, A. Palmisano, A. Esposito; Milan/IT \n(bruno.elisa@hsr.it) \n \nPurpose or Learning Objective: Coronary artery disease (CAD) is a global \nleading cause of morbidity and mortality, with comp lex pathogenesis. Coronary \ncomputed tomography angiography (cCTA) is a powerfu l non-invasive tool for \ndiagnosing obstructive CAD. However, most patients have non-obstructive \nCAD, and risk stratification data are limited. Many  cCTA-based risk scores \nwere developed, however with low predictive value a nd reproducibility. This \nstudy aims to develop clinical-imaging models to pr edict major adverse cardiac \nevents (MACEs) in patients undergoing cCTA for susp ected CAD. \nMethods or Background: Observational, single-center retrospective study \nincluding 4096 out of 10104 patients undergoing cCT A between 2016 and \n2020. Patients with cardiovascular comorbidities or  terminal cancer were \nexcluded. Demographics, cardiovascular risk-factors , and medical history were \ncollected via phone contact and medical records, to  calculate known \nsemiquantitative cCTA scores (CAD-RADS, Leiden risk  score, Leaman risk \nscore, SSS, SIS, Calcium score). Patients were comp ared after a minimum 4-\nyear follow-up according to the occurrence of MACEs  (cardiovascular death, \nnonfatal myocardial infarction, all-cause mortality , angina-related \nhospitalization, late coronary revascularization). Multivariable Cox regression \nmodels, adjusted for age and sex, were created usin g significant clinical \nvariables and one cCTA score. \nResults or Findings: Among 1933 patients enrolled (65% men, age:63.5±11. 6 \nyear-old), 353/1933(18%) had MACE. Patients with MA CE had higher rates of \nhypertension, dyslipidemia, diabetes, and higher cC TA scores(all p<.001). All \ncCTA scores significantly predicted MACE occurrence  in Kaplan-Meier survival \nanalysis(p<.005). Six multivariable models includin g clinical features (diabetes, \ndyslipidemia, hypertension) and one cCTA score have  been developed: in \neach model cCTA score was the strongest prognostica tor of outcome, with \nCAD-RADS having the highest HR(2.996, 95%CI 2.374-3 .781, p<.001), \nfollowed by CACS(2.103, 95%CI 1.646-2.687, p<.001).  \nConclusion: CCTA scores area all predictors of outcome, in part icular CAD-\nRADS, indicating the highest-grade coronary artery lesion, had the higher \nHazard Ratio. \nLimitations: No prospective data. \nFunding for this study: None \nEthics committee - additional information: Approved by San Raffaele \nhospital ethics committee (124/2023) \n \n \nAuthor Disclosures:  \nDavide Vignale: Nothing to disclose \nAlberto Colombo: Nothing to disclose \nAntonio Esposito: Nothing to disclose \nElisa Bruno: Nothing to disclose \nAnna Palmisano: Nothing to disclose \nFrancesco Pisu: Nothing to disclose \nChiara Gnasso: Nothing to disclose \nVittorio Morrone: Nothing to disclose \nAndrea Bettinelli: Nothing to disclose \n \n \nMyocardial delayed enhancement with first-generatio n dual-source \nphoton-counting detector CT: an image quality compa rison across \navailable spectral acquisition modes \n*B. Longere*¹, R. Cusumano¹, C. V. Gkizas¹, A. Rodr iguez Musso¹, F. Dubus¹, \nC. Croisille², C. Artaud¹, M. Haidar¹, F. A. Pontan a¹; ¹Lille/FR, ²Bordeaux/FR \n(benjamin.longere@chu-lille.fr) \n \nPurpose or Learning Objective: To compare the image quality of myocardial \ndelayed enhancement (CT-MDE) obtained by two differ ent tube voltages and \nthree distinct cardiac synchronization modes on a f irst-generation dual-source \nphoton-counting detector CT (PCD-CT). \nMethods or Background: Ninety patients (43 women) aged 63 years (54–\n73y) referred for cardiac CT with CT-MDE were enrol led. CT-MDE acquisition \nwas performed 5min after injection of 90mL of iodin e contrast medium \n(400mgI/mL). Tube voltage was set to either 120 or 140kVp. Current was \nautomatically adjusted to a predetermined image qua lity level of 50. CT-MDE \nwas acquired using helicoidal retrospective gating (R120; R140), sequential \ntriggering (S120; S140) or prospective high-pitch g ating (F120; F140). \nTriggering was set to an RR delay of 300ms. Signal- to-noise ratio (SNR), \ncontrast-to-noise ratio (CNR) and subjective image quality were assessed on \nvirtual monoenergetic images at 65keV (VMI65) and i odine maps. \nResults or Findings: High-pitch acquisitions provided the lowest CT dose  \nindex (P<0.001) with no differences in body mass in dex between the 6 groups \n(P=0.09). No differences were observed in SNR acros s the six acquisition \ntypes on VMI65 (P=0.07) and iodine map (P=0.22). F1 20 demonstrated a CNR \nthat was equivalent to or better than that of the o ther acquisitions (VMI65, \nP=0.01; iodine map, P=0.04). No difference was obse rved in artifact scores \n(VMI65, P=0.43; iodine maps, P=0.83). Global subjec tive image quality \nprovided by F120-derived VMI65 was better than or e quivalent to that of other \nseries (P=0.04). \nConclusion: High-pitch acquisition at 120kVp provides lower rad iation dose \nwithout compromising image quality. This acquisitio n mode should be \nrecommended for the assessment of CT-MDE with PCD-C T. \nLimitations: It was a single-center study with a limited sample size. The \ndetectability of iodine enhancement of pathological  finding was not assessed \nas it was considered to be outside the scope of thi s study. \nFunding for this study: None \nEthics committee - additional information: IRB number: CRM-2408-417 \nAuthor Disclosures:  \nBenjamin Longere: Nothing to disclose \nRaphael Cusumano: Nothing to disclose \nChristos Vasileiou Gkizas: Nothing to disclose \nFrançois Dubus: Nothing to disclose \nMehdi Haidar: Nothing to disclose \nCedric Croisille: Employee: Siemens Healthineers \nCamille Artaud: Nothing to disclose \nAimée Rodriguez Musso: Nothing to disclose \nFrançois Ascagne Pontana: Nothing to disclose \n \n \nQuantification of Extracellular Volume in Acute Myo carditis Using Dual-\nSource Photon-Counting Detector CT: A Comparative A nalysis with CMR \n*C. V. Gkizas*¹, J. Limousin¹, W. Ben Mansoura¹, B.  Longere¹,  \nA. L. Rodriguez Musso¹, C. Croisille², F. A. Pontan a¹; ¹Lille/FR, ²Bordeaux/FR \n(chgkizas@gmail.com) \n \nPurpose or Learning Objective: The aim of this study was to assess the \nfeasibility and accuracy of myocardial late enhance ment (LE) scanning for \nextracellular volume (ECV) quantification with dual -source photon-counting \ndetector computed tomography (PCD-CT) in acute myoc arditis. \nMethods or Background: Patients with clinical suspicion of myocarditis who  \nwere referred for coronary CT angiography (CCTA) to  exclude CAD were \nincluded in this retrospective study. The CCTA prot ocol using a first-generation \nPCD-CT, included a systematic LE acquisition. ECV w as calculated from the \niodine ratio of the myocardium and blood pool on th e LE scan. A \ncomprehensive CMR protocol was used as the referenc e method to confirm \nmyocarditis according to the Lake Louise 2018 crite ria. All subjects underwent \nCCTA and CMR within 24 hours. \n \n\n \n \nAbstract-based Programme \n \n 63  \nWednesday \nResults or Findings: 32 patients were included (mean age 36 years; 13 \nfemales). The mean dose length product of the LE sc an was 96± 32 mGy.cm. \nThe mean global ECV between CCT and CMR did not sho w significant \ndifference (29.4% ±4.5 vs 30.0 ±4.1, P=0.69). In patients diagnosed with \nmyocarditis confirmed by CMR (n=25), the mean ECV-C T was notably \nelevated compared to individuals with normal CCT an d CMR findings (31.6% \n±3.6 vs 25.6% ±3.2, P<0.01). ECV-CT value showed a strong positive \ncorrelation with LGE mass (r =0.85; p < 0.001). \nConclusion: Calculation of ECV using iodine maps derived from L E cardiac \nCT images is both feasible and accurate at low radi ation dose. PCD-CT offers \na promising non-invasive imaging method in the cont ext of acute myocarditis. \nLimitations: Retrospective, single study \nFunding for this study: None \nEthics committee - additional information: All subjects were informed and \nprovided their consent. \nAuthor Disclosures:  \nBenjamin Longere: Nothing to disclose \nChristos Vasileiou Gkizas: Nothing to disclose \nCedric Croisille: Nothing to disclose \nJean Limousin: Nothing to disclose \nAimée Leilen Rodriguez Musso: Nothing to disclose \nFrançois Ascagne Pontana: Nothing to disclose \nWissem Ben Mansoura: Nothing to disclose \n \n \n16:30-17:30 Research Stage 1 \nResearch Presentation Session: Oncologic \nImaging \nRPS 616 \nHaematologic malignancies: multimodality \nimaging \n \nModerator \nG. Cowell; Glasgow/UK  \n(Gordon.Cowell@ggc.scot.nhs.uk) \n \n \nProspective assessment of 3T Whole-Body MRI and 18F -FDG PET-CT in \ndiagnosing multiple myeloma and its influence on pa tient care \n*A. Rossi*¹, D. Bezzi¹, D. Diano¹, A. Prochowski Ia murri¹, A. Cattabriga²,  \nE. Antognoni¹, G. Feliciani¹, P. Caroli¹, C. Cerchi one¹; ¹Meldola/IT, ²Bologna/IT \n(alice.rossi@irst.emr.it) \n \nPurpose or Learning Objective: This study aims to compare the diagnostic \nefficacy of Whole Body-Magnetic Resonance Imaging ( WB-MRI) and 18F-\nFluorodeoxyglucose Positron Emission Tomography (PE T-CT) in detecting \nbone marrow infiltration (BMI) in myeloma patients and assess their impact on \npatient management. \nMethods or Background: We prospectively enrolled myeloma patients from \nOctober 2020 to January 2024. Within a month, patie nts underwent 3T WB-\nMRI (following MY-RADS criteria) and PET-CT to asse ss BMI, para, and \nextramedullary disease. Clinical and laboratory dat a were collected. Two \nhaematologists determined treatment plans using Int ernational Myeloma \nWorking Group (IMWG) criteria based on all findings , which were then used to \nevaluate imaging performance. \nResults or Findings: The cohort included 137 patients (73 male; mean age , \n66 years), with 39 having High Risk-Smoldering Mult iple Myeloma (SMM) and \n98 with Multiple Myeloma (MM). WB-MRI sensitivity a nd specificity for BMI in \nMM were 100% and 97%, respectively, while PET-CT sh owed 89% sensitivity \nand 97% specificity (p=0.02). In SMM, BMI-positive patients had higher \nparaprotein levels (p=0.01); in MM, they had higher  paraprotein (p=0.007) and \nlower hemoglobin (p=0.002). Clinical management cha nged in 54% of cases \nbased on combined imaging results, with WB-MRI cons istent with management \nchanges in 97% compared to 61% for PET-CT (p < 0.00 1). \nConclusion: WB-MRI and PET-CT play key roles in evaluating myel oma \npatients. WB-MRI demonstrated superior sensitivity in detecting BMI and had a \ngreater influence on therapeutic decision-making. \nLimitations: No limitations were identified \nFunding for this study: This study was partly funded by the Italian Ministr y of \nHealth for Institutional Research (Ricerca Corrente ) within the research line \n\"Innovative therapies, phase I-III clinical trials,  and therapeutic strategy trials \nbased on preclinical models, onco-immunological mec hanisms, and nano \nvectors. \nEthics committee - additional information: the study was approved by \nC.E.R.O.M comitato etico della Romagna (AccuMRI tri al IRST code 100.15) \nAuthor Disclosures:  \nPaola Caroli: Nothing to disclose \nDavide Bezzi: Nothing to disclose \nDanila Diano: Nothing to disclose \nClaudio Cerchione: Nothing to disclose \nEleonora Antognoni: Nothing to disclose \nArrigo Cattabriga: Nothing to disclose  \nAlice Rossi: Nothing to disclose \nAndrea Prochowski Iamurri: Nothing to disclose \nGiacomo Feliciani: Nothing to disclose \n \n \nPrognostic value of maximum tumor spread (Dmax) in lymphoma \npatients treated with CD19-specific CAR-T cell ther apy \n*M. Winkelmann*, P. Achhammer, V. Blumenberg, K. Re jeski, G. Sheikh,  \nM. Brendel, J. Ricke, M. Subklewe, W. G. Kunz; Muni ch/DE \n \nPurpose or Learning Objective: CD19 specific CAR T-cell therapy (CART) is \nan effective treatment for relapsed or refractory ( r/r) lymphoma. The maximum \ndistance (Dmax) of lymphoma lesions holds potential  as imaging biomarker in \nlymphoma treated with conventional therapies, but h as not been studied in \ncontext of CART. We evaluated Dmax at baseline imag ing as a prognostic tool \nfor assessment of metabolic and overall response, p rogression-free survival \n(PFS) and overall survival (OS). \nMethods or Background: Consecutive r/r lymphoma patients with (PET/)CT \nat baseline before CART were included. Dmax was mea sured in cm at BL. \nPatients were divided into three groups according t o Dmax: low, intermediate \nand high. The sum of product diameters (SPD) accord ing to Lugano criteria \nwas used to represent tumor burden (TB). Overall re sponse according to \nLugano criteria and Deauville score were determined  at follow-up imaging. \nResults or Findings: 103 patients were included. Median baseline Dmax wa s \n40.0 cm (IQR: 16.4 – 70.3 cm). Median TB was signif icantly higher in the \nintermediate and high risk group compared to the lo w risk group (p=0.005). \nIntermediate and high risk group showed significant ly higher Ann Arbor stages \n(p<0.001). The survival analysis revealed a signifi cantly (p=0.030) shorter PFS \nin the high-risk group compared to the other patien ts (91 vs 364 days), but no \nrelevant differences in OS (p=0.151). In addition, no significant differences in \nDeauville score and ORR were detected. \nConclusion: Patients with high Dmax showed a shorter PFS, but n o significant \ndifferences in OS. Dmax as an interval-scaled param eter represents a useful \nalternative to the Ann Arbor classification. \nLimitations: Single center study, limited number of subjects. Fe w patients \nwere excluded because of no measurable disease. Som e patients had only CT \nwithout PET at FU, with a possible redistribution o f Deauville score among \nDmax-based groups. \nFunding for this study: The work was supported by funding from the research  \nprogram “Förderung für Forschung und Lehre (FöFoLe)  project number 1147” \nof the Medical Faculty of Ludwig Maximilian Univers ity (LMU) Munich and the \nBavarian Cancer Research Center (BZKF) \nEthics committee - additional information: All medical records and imaging \nstudies underwent review with approval from the LMU  Munich Institutional \nReview Board (LMU Ethics Committee, project number 19-817). \nAuthor Disclosures:  \nMarion Subklewe: Speaker: Amgen; Astra Zeneca; BMS/ Celgene; GSK; Incyte \nBiosciences; Janssen; Novartis; Pfizer; Seattle Gen etics; Takeda Consultant: \nAven Cell; CDR-Life; GSK; Incyte Biosciences; Janss en; Miltenyi Biotec; \nMolecular Partners Novartis; Pfizer; Takeda Researc h/Grant Support: Amgen; \nBMS/Celgene; Gilead; Incyte Biosciences; Janssen; M iltenyi Biotec; \nMorphosys; Novartis; Roche; Seattle Genetics; Taked a \nViktoria Blumenberg: Consultant: Kite/Gilead Resear ch/Grant Support: \nJanssen; BMS/Celgene; Novartis; Takeda; Roche \nPhilipp Achhammer: Nothing to disclose \nMichael Winkelmann: Nothing to disclose \nKai Rejeski: Research/Grant Support: Kite/Gilead; N ovartis Consultant: \nKite/Gilead; MBS/Celgene \nWolfgang Gerhard Kunz: Advisory Board: Bristol Myer s Squibb; Boehringer \nIngelheim; mintMedical; Need, Inc. \nGabriel Sheikh: Nothing to disclose \nMatthias Brendel: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nEarly Whole-Body MRI as a Predictor of long-term Me tabolic Response in \nLarge B-Cell Lymphoma Patients following CAR T-Cell  Therapy \n*C. Neelsen*¹, C. Sachpekidis¹, J. M. E. Jende¹, R.  Gnirs¹, F. Kurz², P. Dreger¹, \nA. Dimitrakopoulou-Strauss¹, H-P. Schlemmer¹; ¹Heid elberg/DE, ²Geneva/CH \n \nPurpose or Learning Objective: To evaluate the utility of early whole-body \nMRI (wbMRI) for the prediction of long-term metabol ic response in patients with \nlarge B-cell lymphoma (LBCL) following chimeric ant igen receptor T-cell \ntherapy (CARTT). \n\n \n \nAbstract-based Programme \n \n 64  \nWednesday \nMethods or Background: In this prospective, IRB-approved study, we \nassessed 9 LBCL patients with target lesions identi fied on baseline wbMRI \naccording to adapted Response Evaluation Criteria i n Lymphoma (RECIL), \nwho were fit to undergo wbMRI within two weeks (mea n 11 days) following \nCARTT in an outpatient setting. Complete remission was defined as the \nabsence of measurable disease and any lesions on di ffusion-weighted \nimaging. Early wbMRI findings were compared with PE T-CT results at 3-month \nfollow-up, with metabolic responses classified acco rding to the Lugano criteria. \nResults or Findings: At the 3-month PET-CT follow-up 4 patients showed a  \ncomplete metabolic response (CMR), 2 patients had a  partial metabolic \nresponse (PMR) and 3 patients demonstrated progress ive disease (PD). Of the \n4 patients with CMR, 2 had already achieved complet e remission on the early \nwbMRI, while the other 2 patients had a minor and p artial response. The 2 \npatients with PMR exhibited minor responses on earl y wbMRI and finally of the \n3 patients with PD, one had a partial and two had m inor responses. \nConclusion: All patients demonstrated some degree of response o n early \nwbMRI within two weeks after CARTT. Early complete remission appeared to \nbe a predictor of long-term metabolic response. How ever, early wbMRI was \ninconclusive in patients with initial minor or part ial responses, as these patients \nmay still achieve complete metabolic remission or p rogress over time. \nLimitations: The lymphodepleting chemotherapy administered prior  to the \nCAR T-cell infusion complicates the differentiation  of the specific effects of \nchemotherapy versus the CAR T-cell therapy itself. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The study was approved by the \ninstitutional review board (S-950/2021) and written  informed consent was \nobtained from all participants. \nAuthor Disclosures:  \nChristos Sachpekidis: Nothing to disclose \nPeter Dreger: Nothing to disclose \nRegula Gnirs: Nothing to disclose \nChristian Neelsen: Nothing to disclose \nJohann Malte Enno Jende: Nothing to disclose \nFelix Kurz: Nothing to disclose \nAntonia Dimitrakopoulou-Strauss: Nothing to disclos e \nHeinz-Peter Schlemmer: Nothing to disclose \n \n \nDiagnostic value Whole-body Magnetic resonance imag ing (WBMRI) \nshort protocols can be useful in Multiple Myeloma p atients \n*C. R. G. L. O. M. Talei Franzesi*, C. Maino, P. N.  Franco, D. Ippolito,  \nR. Corso; Milan/IT \n(ctfdoc@hotmail.com) \n \nPurpose or Learning Objective: To compare the effectiveness and accuracy \nof whole-body magnetic resonance imaging (WBMRI) sh ort protocols for the \noverall assessment of bone marrow involvement in pa tients with multiple \nmyeloma (MM), in comparison with standard whole-bod y MRI protocol \nMethods or Background: Sixty-four patients with biopsy-proven MM, who \nunderwent an WBMRI with full body coverage (from ve rtex to feet) were \nretrospectively enrolled. WBMRI images were indepen dently evaluated, by two \nexpert radiologists. After identifying the infiltra tion pattern (normal, focal, diffuse \nand combined), the whole skeleton was divided into six anatomic districts: \nskull, spine, sternum and ribs, upper limbs, pelvis  and proximal two-third of \nfemur, remaining parts of lower limbs, and patients  were grouped according to \nnumber(< 5, 5-20, and > 20) and location of the les ions \nResults or Findings: Most of patients showed a focal (59%) and combined \n(33%) infiltration patterns with lytic lesions pred ominantly distributed in the \nspine (82%) and pelvis (67%). Locations less freque ntly involved by focal bone \nlesions were skull and lower limbs (12%, respective ly). Excluding both the \nanatomic regions mentioned before from the standard  MRI protocol, a short \nMRI protocol with a shorter execution time (saving about 14 minutes) could be \nobtained, maintaining a good sensitivity (89.9%), s pecificity (66.7%) and \ndiagnostic accuracy (AUROC=0.881; 95%CIs: 0.797-0.9 65) \nConclusion: MRI short protocols could be proposed as an effecti ve and \nreliable approach to reduce the examination time, p reserving a high diagnostic \naccuracy and can be more focused on the main involv ed districts \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nCesare Maino: Nothing to disclose \nCammillo Roberto Giovanni Leopoldo Oreste Massimili ano Talei Franzesi: \nNothing to disclose \nRocco Corso: Nothing to disclose \nPaolo Niccolò Franco: Nothing to disclose \nDavide Ippolito: Nothing to disclose \n \n \n \nDual-Vessel Microcirculation Imaging in Differentia tion of B cell and T cell \nsubtype in intranodal Non-Hodgkin Lymphoma Using Su per-Resolution \nUltrasound: An Exploring Study \n*Y. Dong*; Shanghai/CN \n(dyj11584@rjh.com.cn) \n \nPurpose or Learning Objective: To explore the diagnostic performance of the \nsuper-resolution ultrasound (SRUS) imaging in dual- vessel systems, i.e., the \nmicrovascular system and the microlymphatic system,  for predicting B cell and \nT cell subtypes in intranodal NHL. \nMethods or Background: Forty-two patients with intranodal NHL were \nincluded in this prospective study. All patients un derwent dual-vessel system \nSRUS imaging via intravenous and intra-lymph node r outes. SRUS parameters \nsuch as vessel density, vessel ratio, vessel comple xity level, diameter, \ndistance, tortuosity, and flow velocity were measur ed for the microvascular and \nmicrolymphatic circulations. Principal component an alysis (PCA) was utilized to \ninterpret parameters, and a regression model was de veloped to predict NHL \nsubtypes. The areas under the receiver operating ch aracteristic curve (AUC), \nsensitivity, and specificity were calculated. \nResults or Findings: Among the 42 patients, 35 were diagnosed with B cel l \nNHL and 7 with T cell NHL. Sixty parameters from du al-vessel SRUS image \nanalysis were obtained for each case. PCA extracted  six principal components \naccounting for 93.1% of the total variance. The reg ression model utilizing these \ncomponents to distinguish between B-cell and T-cell  lymphomas achieved an \nAUC of 0.927 (95% CI: 0.841-1.000), a sensitivity o f 85.7% (95% CI: 42.1%-\n99.6%), and a specificity of 94.3% (95% CI: 80.8%-9 9.3%). \nConclusion: Dual-vessel SRUS imaging, in conjunction with quant itative \nanalysis, could effectively differentiate between B -cell and T-cell NHL, offering \na non-invasive diagnostic alternative. \nLimitations: First, due to the relatively low prevalence of lymp homa in the \ngeneral population, the sample size is relatively s mall. Second, the \nmethodology of microlymphatic SRUS imaging is confi ned primarily to intact \nlymph nodes. In cases of 'bulky mass' lymphomas, wh ere there is extensive \ndisruption of the lymph node architecture, as well as in NHL presenting within \nthe trunk region, this approach might not be approp riate. \nFunding for this study: No. \nEthics committee - additional information: Ruijin Hospital Clinical Research \nCenter Shanghai Jiaotong University, School of Medi cine, Ethic No. \n20240116021828990. \nAuthor Disclosures:  \nYijie Dong: Nothing to disclose \n \n \n16:30-17:30 Research Stage 2 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 605 \nArtificial intelligence in cardiovascular \nimaging \n \nModerator \nT. Leiner; Rochester, MN/US  \n(leiner.tim@mayo.edu) \nAuthor Disclosures:  \nTim Leiner: Research Grant/Support: Philips Healthc are; Other: Editor-in-Chief, \nJournal of Cardiovascular Magnetic Resonance (JCMR)  \n \n \nCT Deep learning AI quantified fibrosis predicts pr ognosis in Pulmonary \nHypertension associated with Chronic Lung Disease \n*K. Dwivedi*, M. Sharkey, S. Alabed, A. Maiter, C. S. Johns, S. Rajaram,  \nR. Condliffe, D. Kiely, A. J. Swift; Sheffield/UK \n(k.dwivedi@sheffield.ac.uk) \n \nPurpose or Learning Objective: Pulmonary Hypertension associated with \nChronic Obstructive Pulmonary Disease (PH-COPD) is a heterogenous \ncondition, with a spectrum of predominantly emphyse ma and some \noverlapping fibrosis. All patients undergo CT, but it is not used for \nprognostication. The study aim is to investigate th e prognostic value of an AI \nmodel that quantifies the percentage of fibrosis on  baseline CT, compared to \nradiological assessment. \nMethods or Background: PH-COPD patients with baseline CT between 2001-\n19 were identified from the ASPIRE registry. A vali dated in-house PH specific \ndeep-learning model was run and provided percentage  of fibrosis by \n\n \n \nAbstract-based Programme \n \n 65  \nWednesday \nquantifying ground glass change, ground glass with reticulation, and \nhoneycombing. Scans were scored as none/mild/modera te/severe fibrosis by \nsub-specialist radiologists. Cases with mean pulmon ary arterial pressure ≥ 35 \nmmHg were classified as severe PH-COPD, and fibrosi s was grouped with a \nthreshold of 3%. Scaled cox regression and Kaplan M eier survival analysis \nwas performed. \nResults or Findings: 157 PH-CLD patients (113 severe PH-COPD) were \nincluded. AI quantified fibrosis % was a significan t predictor of mortality (HR \n1.46, p<0.001) .There was a significant difference (p=0.001) in survival \nbetween patients with more and less than 3% fibrosi s. One and five-year \nsurvival was 84% and 35% respectively in those with  <3% fibrosis and 63% \nand 18% respectively in those with ≥3% fibrosis. Radiologist scored mild (HR \n2.05, p=0.36) and moderate (HR 2.82, p=0.045) fibro sis was a significant \npredictor, but not severe fibrosis. In severe PH-CO PD, radiological scoring was \nnot a significant predictor at any level, but AI fi brosis% was a significant \npredictor (HR 1.37, p<0.001). \nConclusion: CT Deep learning AI model quantified fibrosis is pr ognostic in \npredicting survival and treatment response in PH-CO PD and provides \nadditional value over radiological assessment in se vere PH-COPD. \nLimitations: Single registry analysis, but imaging from 21 hospi tals. \nFunding for this study: Research conducted during post funded by UK \nNational Institute for Health and Care Research \nEthics committee - additional information: Ethical approval was granted by \nthe Institutional Review Board and approved by the National Research Ethics \nService (16/YH/0352). \nAuthor Disclosures:  \nDavid Kiely: Consultant: Ferrer, MSD, Janssen, Unit ed Therapeutics, \nAcceleron \nAndrew J. Swift: Consultant: Janssen pharamceutical s Grant Recipient: \nWellcome Trust, National Institute for Health and C are Research and Janssen \npharamceuticals \nAhmed Maiter: Nothing to disclose \nChris S. Johns: Nothing to disclose \nMichael Sharkey: Grant Recipient: Wellcome Trust \nSamer Alabed: Advisory Board: Royal College of Radi ologists AI Working \nGroup Grant Recipient: Wellcome Trust and National Institute for Health and \nCare Research \nSmitha Rajaram: Nothing to disclose \nRobin Condliffe: Consultant: Janssen pharamceutical s and MSD \nKrit Dwivedi: Advisory Board: Royal College of Radi ologists AI Working Group \nmember Grant Recipient: Wellcome Trust and National  Institute for Health and \nCare Research \n \n \nExternal Validation of a Deep Learning Cardiac Meta l Artifact Reduction \nAlgorithm (DL-C-MAR) to reduce Metal Artifacts of T ranscatheter Aortic \nValves in CT: a retrospective cohort and phantom st udy \n*I. H. T. Khargi*¹, M. Selles¹, N. Huber², J. Brown e², B. Kietselaer², T. Leiner², \nM. F. Boomsma¹; ¹Zwolle/NL, ²Rochester, MN/US \n(indirakhargi@gmail.com) \n \nPurpose or Learning Objective: To assess the performance of a novel deep \nlearning-based cardiac metal artifact reduction alg orithm (DL-C-MAR) in a \nretrospective comparison with unedited conventional  computed tomography \nangiograms (CTAs) of transcatheter aortic valve imp lantation (TAVI) valves \nand phantom experiments. \nMethods or Background: DL-C-MAR was trained using multiple simulated \nmetal implants and artifacts in 1000 CTAs. Performa nce of DL-C-MAR was \nquantitatively and qualitatively investigated in 50  TAVI patients and compared \nto unedited conventional CTAs. To quantitatively as sess image quality, noise, \ncontrast-to-noise ratio (CNR), artifact index (AI),  and artifact volume were \ncalculated. Diameters of the valve struts were also  measured. Images were \nqualitatively rated on overall image quality, exten t of metal artifacts and valve \nleaflet definition by two readers on a four-point s cale. Phantom experiments \nwere conducted using four different size steel cyli nders. Diameters of the \ncylinders were measured by two readers and compared  to their conventional \ncounterparts and the ground truth. All images were visually screened for \npresence of hallucinations. \nResults or Findings: In the CTAs, DL-C-MAR resulted in a higher CNR \n(9.1±5.8 vs. 7.9±4.8), and lower noise (57.2±33.9 vs. 82.1±54.0), AI (53.0±36.0 \nvs. 75.7±55.9), and artifact volume (0.02±0.12mL vs. 0.06±0.42mL) compared \nto unedited conventional CTAs (all p<0.001). The st rut diameter also \ndecreased after DL-C-MAR (1.52±0.26mm vs. 2.05±0.46mm, p=0.005). Initial \nresults from the qualitative analysis suggest incre ased valve leaflet definition \nand decreased metal artifact severity after DL-C-MA R. In the phantom scans, \nDL-C-MAR decreased cylinder diameter by 7-67% (p<0. 001), bringing them \ncloser to the ground truth. No hallucinations were observed. \nConclusion: DL-C-MAR increases image quality and reduces metal artifacts in \nCTAs after TAVI implantation and does not seem to h allucinate on clinical or \nphantom images. \nLimitations: This study did not include impact on clinical decis ion-making \noutcomes. \nFunding for this study: No funding was received for this study \nEthics committee - additional information: This study was reviewed and \napproved as exempt with waived informed consent. Re ference no.: RPR -\n2024-00000086 \nAuthor Disclosures:  \nMark Selles: Nothing to disclose  \nTim Leiner: Nothing to disclose \nNathan Huber: Employee: Philips Healthcare \nBas Kietselaer: Nothing to disclose \nIndira Hélène Theodora Khargi: Nothing to disclose \nJacinta Browne: Nothing to disclose \nMartijn Franklin Boomsma: Nothing to disclose \n \n \nDiagnostic confidence in coronary stent evaluation using coronary CT \nangiography. Comparison of Deep Learning Reconstruc tion, Hybrid \nIterative Reconstruction and Model Based Iterative Reconstruction \n*M. Finazzo*¹, M. M. Lagana², F. Graziano³, F. Pint o², C. Duranti¹, F. Finazzo¹; \n¹Palermo/IT, ²Milan/IT, ³Monza/IT \n(mariofinazzo67@gmail.com) \n \nPurpose or Learning Objective: The assessment of coronary stents using \nCoronary CT Angiography (CCTA) can be challenging. Deep Learning \nReconstruction (DLR) is an innovative CT image reco nstruction method that \nreduces noise, enhancing image quality. This study aims to evaluate whether \nDLR improves diagnostic confidence in coronary sten t evaluation using CCTA \nimages, compared to hybrid iterative reconstruction  (HIR) and model-based \niterative reconstruction (MBIR). \nMethods or Background: CCTA images of 20 patients with 35 stents were \nevaluated retrospectively using three reconstructio n methods: HIR, MBIR, and \nDLR. All examinations were conducted using a 320-ro w whole-heart CT \nscanner. The diagnostic confidence of the images ob tained with each \nreconstruction method was evaluated using a Likert score (1=non-diagnostic, \n2=poor, 3=acceptable, 4=good, 5=excellent). Stents were divided into proximal \nand distal according to their location. Stents loca ted in the proximal and \nintermediate segments of the coronary arteries were  considered proximal; \nstents situated in the distal segments of the main coronary arteries and side \nbranches were considered distal. A cumulative liked  mixed model was created \nin Rstudio version 4.3.1 to examine the differences  across reconstruction \nmethods while accounting for the stent position, to  explore its potential effect \non diagnostic confidence. Post-hoc comparisons were  conducted, and the p-\nvalues were adjusted using the Tukey method. \nResults or Findings: The reconstruction method had a significant impact,  \nirrespective of stent position. Likert scores were significantly higher for DLR \nimages compared to those reconstructed using HIR an d MBIR (p<0.001), with \nno significant difference between HIR and MBIR (p=0 .957). \nConclusion: DLR provided the best diagnostic confidence and sig nificantly \nenhanced the evaluation of coronary stents. As a fu rther development, Super \nResolution DLR, a new reconstruction algorithm, cou ld improve spatial \nresolution, thereby increasing diagnostic confidenc e in coronary stents. \nLimitations: Limited number of patients. \nQualitative analysis only \nFunding for this study: No funding \nEthics committee - additional information: It's a non-pharmacological \nretrospective observational study which have been a pproved by the local \nethics committee \nAuthor Disclosures:  \nMaria Marcella Lagana: Investigator: Canon Medical Systems Investigator \nMario Finazzo: Nothing to disclose \nFrancesca Pinto: Investigator: Canon Medical System s Clinical Application \nSpecialist \nFrancesca Graziano: Nothing to disclose \nCristiana Duranti: Nothing to disclose \nFrancesca Finazzo: Nothing to disclose \n \n \nComparing the performance of Large Language Models for automatic \nCAD-RADS 2.0 classification from cardiac-CT reports  \n*P. Arnold*, M. Russe, E. Kotter, M. T. Hagar; Frei burg/DE \n(philipp.arnold@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: The Coronary Artery Disease-Reporting and \nData System (CAD-RADS) 2.0 offers standardized guid elines for interpreting \ncoronary artery disease in cardiac computed tomogra phy (CT). Accurate and \nconsistent CAD-RADS 2.0 scoring is crucial for comp rehensive disease \ncharacterization and clinical decision-making. This  study investigates the \ncapability of large language models (LLMs) to auton omously generate CAD-\nRADS 2.0 scores from cardiac CT reports. \nMethods or Background: A dataset of 200 synthetic cardiac CT reports was \ncreated to evaluate the performance of several stat e-of-the-art LLMs in \ngenerating CAD-RADS 2.0 scores via in-context learn ing. The tested models \nincluded GPT-3.5, GPT-4o, Mistral 7b, Mixtral 8x7b,  LLama3 8b, LLama3 8b \n\n \n \nAbstract-based Programme \n \n 66  \nWednesday \nwith a 64k context length, and LLama3 70b. The gene rated scores from each \nmodel were compared to the ground truth, which was provided by an \nindependent committee of two board-certified cardio thoracic radiologists. \nResults or Findings: The GPT-4o model and Llama3 70b achieved the \nhighest accuracy in generating full CAD-RADS 2.0 sc ores including all \nmodifiers, with a performance rate of 93% and 92.5%  respectively, followed by \nMixtral 8x7b with 78%. In contrast, less advanced L LMs, such as Mistral 7b \nand GPT-3.5 provided poor performance (16%). Llama3  8b demonstrated \nintermediate results, with an accuracy of 41.5%. \nConclusion: Advanced LLMs are capable of generating autonomousl y CAD-\nRADS 2.0 scores for cardiac CT reports with excelle nt accuracy, potentially \nenhancing both the efficiency and consistency of ca rdiac CT report \nevaluations. Open-source models not only deliver co mpetitive accuracy but \nalso present the benefit of local hosting, mitigati ng concerns around data \nprivacy. \nLimitations: To ensure data privacy and avoid ethical concerns, this study was \nconducted using synthetically generated cardiac CT reports. Even though \nthese were deemed indistinguishable from real patie nt reports, further research \nis needed to validate LLM performance in real-world  settings. \nFunding for this study: Hans A. Krebs Medical Scientist Program \n(Uniklinikum Freiburg) German Research Foundation ( DFG) - SFB 1597 - \n499552394 \nEthics committee - additional information: None \nAuthor Disclosures:  \nMuhammad Taha Hagar: Nothing to disclose \nMaximilian Russe: Nothing to disclose \nPhilipp Arnold: Nothing to disclose \nElmar Kotter: Nothing to disclose \n \n \nMulti-stage deep learning architecture for carotid artery segmentation \nand stenosis degree evaluation: a comparative study  with DSA \n*Z. Zheng*, X. Cao, W. Liu; Shanghai/CN \n(23110860044@m.fudan.edu.cn) \n \nPurpose or Learning Objective: HR-MRI provided a non-invasive and \nradiation-free method for assessing atherosclerosis , with strong advantages for \nvessel wall visualization. However, efficient segme ntation and stenosis degree \nevaluation remained a challenging dilemma that is b oth labor- and time-\nconsuming and susceptible to interobserver variabil ity. Thus, a multi-stage \ndeep learning architecture was developed to address  above issues. \nMethods or Background: The method contained three modules: artery \nlocalization, automatic segmentation, and stenosis degree evaluation modules. \nThe 422 scans were retrospectively collected from t wo tertiary hospitals \nbetween 2018 and 2023 with a training-validation se t (372 patients, 545 \nlesions) and an independent test set (50 patients, 96 lesions). An external \nvalidation set (26 patients, 42 lesions) was collec ted prospectively between \n2023 and 2024. Subsequently, the artery segmentatio n and stenosis degree \nevaluation were compared against the ground truth, which was established by \nconsensus among three radiologists and derived from  diagnostic results \nobtained via DSA. \nResults or Findings: The results showed outstanding performance with hig h \nDSC, IOU, and low RVE, ASSD, and HD95. The concorda nce correlation \ncoefficient (CCC) was 0.985(95% CI: 0.981-0.987), 0 .979(95% CI: 0.963-\n0.984), and 0.963(95% CI: 0.944-0.992) for volumes of artery on all datasets. \nStenosis degree was evaluated on the NASCET achieve d Acc of 0.8750, \n0.8571, AUC of 0.89, 0.80, Sens of 0.8611, 0.9333, and Spec of 0.9167, \n0.6667 on the independent test and external validat ion sets, respectively. \nConclusion: The method achieved no less accuracy than manual \nsegmentation by physicians and maintained a high co nsistency with the DSA \ndiagnostic criteria. In addition, by shortening dia gnostic time and minimizing \ninter-observer variability, it offered an efficient  intelligent aid in clinical practice. \nLimitations: The method performed in multi-stage may take up a l arge amount \nof computational resources and modifications to the  architecture are required \nto optimize the inference speed. \nFunding for this study: This work has received funding from the National \nNatural Science Foundation of China (82402393, 8210 2132, 8237071280), the \nScience and Technology Commission of Shanghai Munic ipality (20S31904300, \n22TS1400900, 23S31904100, 22ZR1409500) and the Grea ter Bay Area \nInstitute of Precision Medicine (Guangzhou) (KCH231 0094). \nEthics committee - additional information: All patients or their guardians \ngave informed consent to use their anonymized MRI i mages and clinical data \nfor research purposes. Since all data were obtained  in the course of daily work, \nthe Ethics Committee waived the need for informed c onsent. \nAuthor Disclosures:  \nZhiji Zheng: Nothing to disclose \nWanchen Liu: Nothing to disclose  \nXin Cao: Nothing to disclose \n \n \nAI-driven joint segmentation of myocardium, scar, a nd microvascular \nobstruction in bright-blood late gadolinium enhance ment cardiac \nmagnetic resonance imaging \n*B. Durand*, V. De Villedon De Naide, T. Génisson, M. Stuber, A. Bustin,  \nH. Cochet; Bordeaux/FR \n(baptisted55@gmail.com) \n \nPurpose or Learning Objective: develop and test an AI-driven deep learning \nmodel for joint segmentation of healthy myocardium,  scar tissue, and \nmicrovascular obstruction (MVO) in cardiac MRI usin g bright-blood phase-\nsensitive inversion recovery (PSIR) imaging. \nMethods or Background: Current methods for scar and MVO quantification in \nPSIR imaging are manual or semi-automated, time-con suming, and prone to \nerrors and variability. Using a nnUNET architecture , the model was trained on \n50 PSIR exams with suspected ischemic heart disease  and evaluated on a test \nset of 20 cases. Data augmentations were applied, a nd manual segmentations \nby radiologists were used for comparison. To maximi ze performance, a joint \nsegmentation approach was employed, and both magnit ude and phase maps \nwere used together. \nResults or Findings: The AI model demonstrated excellent performance \ndespite only 50 exams in training, in segmenting he althy myocardium (median \nDice score 0.96) and good results for scar segmenta tion (median Dice score \n0.75). MVO detection was successful in 2 out of 3 c ases. Inference time was \nunder 5 seconds per exam, and no false positives we re identified outside the \nmyocardium. \nConclusion: AI-driven approach showed robust segmentation of my ocardium \nand scar tissue, with promising results in MVO dete ction. It could streamline \nclinical workflows for myocardial infarction assess ment by reducing time-\ncounsuming manual segmentations. \nLimitations: The model was trained on a small dataset from post- ischemic \npatients, which limits its generalizability to othe r cardiac conditions such as \nhypertrophic cardiomyopathy or infiltrative disease s, where scarring patterns \ndiffer. We plan to expand the training population t o improve the model's \nperformance in non-ischemic cardiomyopathy. \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nt ANR-22-CPJ2-\n0009-01, and from the European Research Council (ER C) grant \"SMHEART\" \nunder the European Union’s Horizon 2020 research an d innovation programme \n(grant agreement No101076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \nAuthor Disclosures:  \nVictor De Villedon De Naide: Nothing to disclose  \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nThaïs Génisson: Nothing to disclose  \nBaptiste Durand: Nothing to disclose \nMatthias Stuber: Nothing to disclose \n \n \n16:30-17:30 Research Stage 3 \nResearch Presentation Session: Physics \nin Medical Imaging \nRPS 613 \nMRI spinning for development and \nsustainability \n \nModerator \nT. G. Maris; Iraklion/GR  \n(tmaris@med.uoc.gr) \n \n \nRevisiting TE selection for T2-weighted spin-echo M RI of the prostate \n*S. J. Riederer*¹, R. Pabi¹, E. Borisch¹, A. Froemm ing¹, A. Kawashima²,  \nN. Takahashi¹; ¹Rochester, MN/US, ²Phoenix, AZ/US \n(riederer@mayo.edu) \n \nPurpose or Learning Objective: To study if the long echo trains of fast-spin-\necho (FSE) prostate T2-weighted imaging (T2-WI) cau se the optimum echo \ntime (TE) for distinguishing normal peripheral zone  (PZ) vs. malignancy to \ndeviate from the optimum TE chosen based on standar d T2 decay. \nMethods or Background: All work was done at 3 Tesla. Experiments were \ndone using a standard (NIST) phantom containing ten  vials with known T2 \nrelaxation times. Vials 5 (T2=133.3 msec) and 6 (96 .9 msec) were analyzed, \n\n \n \nAbstract-based Programme \n \n 67  \nWednesday \nhaving values closest to literature-taken T2 measur ements in normal PZ (125 \nto 150 msec) and PZ malignancy and normal transitio n zone (TZ) (75 to 105 \nmsec). The phantom was imaged using conventional sp in-echo (TE 97, 113, \n129, 153 msec; scan time 20 min for each) and a cli nical T2-WI sequence \n(echo-train-length 21; TE-EFF 104, 114, 135, 145 ms ec; scan time 2:30 for \neach). 30 consecutive subjects with suspected prost ate cancer were imaged \nusing the clinical T2-WI sequence at both TE-EFF 10 0 and TE-EFF 150. \nResults were compared visually for relative contras t of PZ to TZ and any \nsuspected lesions. \nResults or Findings: Contrast-to-noise ratio (CNR) between Vials 5 and 6  for \nconventional spin-echo peaked at TE=125 msec, consi stent with theory. \nHowever, CNR between the vials for the long-ETL T2- WI sequence was \nhighest at TE-EFF 145 msec, 10% higher vs. 104 msec . In 18 of 30 patient \nstudies the TE-EFF 150 series had superior contrast  vs. TE-EFF 100, inferior \nin 2/30, and equivalent in 10/30. \nConclusion: Fast-spin-echo acquisition in prostate T2-WI artifa ctually prolongs \nthe apparent T2 relaxation, causing the optimum ech o times for distinguishing \nnormal from malignant tissue to be higher than that  predicted assuming \ntabulated T2 values. TE-EFF 150 consistently provid es improved contrast vs. \nTE-EFF 100 msec. \nLimitations: Limited number of subjects \nFunding for this study: This work was funded by NIH. \nEthics committee - additional information: Informed consent was provided \nby all human subjects. \nAuthor Disclosures:  \nAdam Froemming: Nothing to disclose \nStephen J. Riederer: Nothing to disclose \nEric Borisch: Nothing to disclose \nNaoki Takahashi: Nothing to disclose \nRonard Pabi: Nothing to disclose \nAkira Kawashima: Nothing to disclose \n \n \nComparison Between Conventional and Compressed SENS E Sequences \non MRI Brain in Paediatric Population \n*I. S. Shah*, P. C. P. Joshi, V. Jahanvi; Pune/IN \n(ishshah04@gmail.com) \n \nPurpose or Learning Objective: To compare the quality and image \nacquisition time between conventional and Compresse d SENSE sequences in \nbrain magnetic resonance imaging (MRI) in paediatri c population. \nMethods or Background: Thirty children (below the age of 18 years) \nundergoing MRI brain were included in this study. I n addition to the routine \nsequences, one Compressed SENSE sequence was added.  2D - T1, T2, and \nFLAIR axial sequences were acquired for brain using  conventional and \nCompressed SENSE techniques. One of each sequence w as acquired in 10 \npatients undergoing an MRI brain study on a 3T MRI using coil 32 channel coil \nfor adults and pediatric 8ch head coil for neonates . Two consultant radiologists \n(with 35 years and 5 years experience in radiology)  independently scored the \nimage quality using the 5-point Likert scale based on resolution, visualization of \nanatomical regions, grey-white matter differentiati on, sharpness of the image \nand artefacts. The subjective criteria details for image quality as per the 5-point \nLikert scale were: non-diagnostic (1), poor (2), mo derate (3), good (4) and \nexcellent (5). \nResults or Findings: The time reduction achieved with 2D T1 at 2 reducti on \nfactor were 60 seconds(24%), with 2D T2 at reductio n factor of 2.2 66 \nseconds(47.83%) and with 2D FLAIR at reduction fact or 2 66 seconds(40%). \nInter-rater agreement for overall diagnostic confid ence was rated higher for \nCompressed SENSE (k – 0.632) than conventional (k –  0.464). Nonsignificant \nstatistical difference was found regarding image qu ality and image contrast \nratio between both techniques. \nConclusion: Compressed SENSE has potential in reducing the imag e \nacquisition time without compromising the image qua lity and diagnostic \nconfidence. Motion artefacts are also reduced with reduction in time with the \nuse of Compressed SENSE sequence. \nLimitations: A small sample size. \nFunding for this study: None \nEthics committee - additional information: Institutional ethics committee \napproval was obtained \nAuthor Disclosures:  \nPriscilla Col Priscilla Joshi: Nothing to disclose \nVandana Jahanvi: Nothing to disclose \nIsha Sandip Shah: Nothing to disclose \n \n \n \n \n \n \n \n \nComparison and optimization of deep learning enhanc ed 2D ATPw-CEST \nMRI at 1.5 Tesla and 3 Tesla: A clinically relevant  phantom study \n*L. Wei*¹, A. Volk¹, S. Campana Tremblay², J. Poujo l³, S. Ammari¹, G. Garcia¹, \nC. Balleyguier¹, N. Lassau¹, F. Bidault¹; ¹Villejui f/FR, ²La Ciotat/FR, ³Buc/FR \n(lecong.wei@gustaveroussy.fr) \n \nPurpose or Learning Objective: The purpose was to compare the CEST \neffect at 1.5T and 3T on high resolution 2D images provided by a deep-\nlearning algorithm. In addition, the effect of satu ration offset number reduction \ncombined with B0 Mapping on MTRasym values was inve stigated, in order to \nminimize acquisition time for clinical use. \nMethods or Background: The phantom consisted of 12 tubes filled with BSA \nat pH 7 for 3 different physiological concentration s, at 37 °C. Acquisitions were \nperformed on 1.5T (GE Artist) and 3T (GE Signa Prem ier) MR scanners. CEST \ndata were acquired with 2D SSFSE using the AIR Reco n Deep Learning option \n(ARDL) for image reconstruction, CW saturation was used, with 61 offsets. \nSNR was compared with previously acquired data with out ARDL. B0 correction \nwas performed by using the chemical shifts of Z-spe ctra minima, and by using \nthe 2D B0 Mapping GRE sequence. MTRasym was compare d for different \noffset numbers ranging from 6 to 61 using B0 Map GR E. \nResults or Findings: 2D CEST SSFSE using ARDL had better SNR \ncompared to the sequence without ARDL at 1.5T and 3 T. For the two B0 \ncorrection methods, MTRasym values were similar and  increased with protein \nconcentration at 1.5T and 3T. For ΔB0<0.3ppm, MTRasym variations were \nsmall for all offset numbers (max 6%). For ΔB0>0.5ppm, MTRasym were \nsimilar for 16 offsets or more. \nConclusion: This study provided a comparison of 2D CEST SSFSE a t 1.5T \nand 3T on a phantom carried out during the same ima ging session. The \nfindings open up the prospect of high-resolution ti me efficient APTw-CEST \nclinical MRI at 1.5T. \nLimitations: However, the results are preliminary, hence repeata bility studies \nand proof of concept in patients will be considered . \nFunding for this study: This material is based upon work supported by the \nANRT with a CIFRE fellowship granted to Lecong Wei.  \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nGabriel Garcia: Nothing to disclose \nFrançois Bidault: Nothing to disclose \nCorinne Balleyguier: Nothing to disclose \nLecong Wei: Employee: Olea Medical, La Ciotat, Fran ce Research/Grant \nSupport: This material is based upon work supported  by the ANRT with a \nCIFRE fellowship granted to Lecong Wei. \nSamy Ammari: Nothing to disclose \nNathalie Lassau: Nothing to disclose \nSophie Campana Tremblay: Employee: Olea Medical \nJulie Poujol: Employee: GE HealthCare \nAndreas Volk: Nothing to disclose \n \n \nCan we measure extreme brain iron content with Quan titative \nSusceptibility Mapping? \n*C. Birkl*¹, M. Panzer¹, C. Kames², A. Rauscher², B . Glodny¹,  \nE. R. R. Gizewski¹, H. Zoller¹; ¹Innsbruck/AT, ²Van couver, BC/CA \n(christoph.birkl@i-med.ac.at) \n \nPurpose or Learning Objective: Aceruloplasminemia (ACP) is a rare \nautosomal recessive disorder characterized by progr essive iron accumulation \nin multiple organs, including the brain, liver, and  pancreas. Magnetic \nResonance Imaging (MRI) is commonly used to detect iron overload, with \nQuantitative Susceptibility Mapping (QSM) emerging as a promising method for \nassessing brain iron levels. Despite its potential,  QSM faces challenges such \nas susceptibility artifacts and a lack of standardi zation. This prospective study \naimed to evaluate the performance of different QSM algorithms in measuring \nbrain iron in patients with severe iron overload, c ompared to healthy controls. \nMethods or Background: QSM images were acquired using a 3D multi-echo \ngradient echo sequence in three patients with ACP a nd three healthy controls. \nWe evaluated six QSM algorithms: (I) Fast Nonlinear  Susceptibility Inversion \n(FANSI), (II) Improved Sparse Linear Equation and L east-Squares (iLSQR), \n(III) Morphology-Enabled Dipole Inversion (MEDI), ( IV) Streaking Artifact \nReduction (STAR) QSM with Rapid Open-source Minimum  Spanning Tree \n(ROMEO) phase unwrapping, (V) STAR QSM with Laplaci an phase \nunwrapping, and (VI) Multi-Echo Rapid Two-Step (MER TS) QSM. Regional \nsusceptibility values were analyzed in the caudate nucleus, putamen, globus \npallidus, and thalamus. \nResults or Findings: We observed significant variability in susceptibili ty \nvalues across the different algorithms for patients  with ACP. Among the \nalgorithms tested, only one showed consistently ele vated susceptibility values \nin the globus pallidus of ACP patients compared to healthy controls. Many \nsusceptibility maps showed signal dropouts in brain  regions with extreme iron \noverload. \n \n\n \n \nAbstract-based Programme \n \n 68  \nWednesday \nConclusion: Our findings suggest that only a subset of QSM algo rithms \nreliably reflect extreme brain iron deposition. Add itionally, the study highlights \nthat performing echo combination prior to phase unw rapping and background \nfield removal may introduce artifacts, resulting in  lower-than-expected \nsusceptibility values due to signal dropouts. \nLimitations: A limitation is the small sample size. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by the \nlocal ethics committee (number 1270/2021) \nAuthor Disclosures:  \nAlexander Rauscher: Nothing to disclose \nBernhard Glodny: Nothing to disclose \nElke Ruth R Gizewski: Nothing to disclose \nChristoph Birkl: Nothing to disclose  \nMarlene Panzer: Nothing to disclose \nHeinz Zoller: Nothing to disclose \nChristian Kames: Nothing to disclose \n \n \nMRI signal intensity comparison of high relaxivity vs standard \ngadolinium-based contrast agents: Concentration-dep endent effects \nacross different MRI sequences and field strengths \n*L. Widmer*¹, S. Bhumiwat², F. Porões¹, J. M. M. Fr oehlich³, H. Thoeny¹; \n¹Fribourg/CH, ²Phatum Wan/TH, ³Zurich/CH \n \nPurpose or Learning Objective: Gadopiclenol (Elucirem™) is a recent high-\nrelaxivity macrocyclic gadolinium-based contrast ag ent (GBCA), with limited \ndata on its detailed concentration dynamics, essent ial for optimizing its use in \ndiverse clinical scenarios. This study compared sig nal intensity (SI) curves \nfrom three GBCAs across various MRI sequences, fiel d strengths, coils and \nconcentrations. \nMethods or Background: Signal intensity of gadopiclenol, gadoteric acid an d \ngadobutrol vials were measured across 18 MRI sequen ces on 1.5T and 3T \nmachines and 8 concentrations ranging from 0 to 25 mmol/L in an \nexperimental in-vitro setting. Relationship between  SI and concentrations were \ncompared in SE, FSE, GRE and IR sequences. \nResults or Findings: Concentration had no linear correlation with the SI . At \nthe same concentrations, gadopiclenol produced high er maximal SI than the \nother two contrast agents in half of the sequences (50%, 9/18). In most \nsequences (56%, 10/18), gadopiclenol had a left-shi fted curve maximum, \nreflecting higher SI at lower concentrations. Resul ts of identification of curves \npatterns by sequence type are still pending. \nConclusion: Signal intensity curve analysis helps optimize imag ing and \ninjection parameters, though in-vivo application re quires considering vessel \nand tissue distribution. These findings suggest usi ng reduced dose of high-\nrelaxivity agents compared to conventional GBCAs, s upporting sustainable \nradiology. \nLimitations: Experimental \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nHarriet Thoeny: Nothing to disclose \nJohannes Malte Maria Froehlich: Nothing to disclose  \nSiwat Bhumiwat: Nothing to disclose \nFabio Porões: Nothing to disclose  \nLucien Widmer: Nothing to disclose \n \n \nDeep Learning-Based Spatial Resolution Improving Al gorithm for MRI: \nComparison of Capabilities for Scan Time Reduction and Image Quality \nImprovement with Conventional Protocol with and wit hout ZIP \n*D. Takenaka*, H. Nagata, T. Ueda, M. Nomura, T. Yo shikawa, Y. Ozawa,  \nY. Ohno; Toyoake/JP \n(daisuke.takenaka.fr@fujita-hu.ac.jp) \n \nPurpose or Learning Objective: Deep learning reconstruction (DLR) and zero \nfill interpolation (ZIP) technique have been clinic ally applied on routine clinical \nMRIs. Recently, deep learning-based spatial resolut ion improving algorithm \n(Precise IQ Engine: PIQE) is developed to transform  MR data from low-spatial \nresolution data to high-spatial resolution data. Th e purpose of this study was to \ndirectly compare utilities of PIQE for scan time re duction and image quality \nimprovement of MRIs as compared with DLR with and w ithout ZIP techniques. \nMethods or Background: 28 consecutive patients suspected with 17 brain \ntumors, 6 spinal diseases and 5 musculoskeletal dis eases were prospectively \nscanned with conventional MR (224-382×256-512matrix ) and new MR \nprotocols (160-192×192-416matrix). Then, both MR pr otocol data were \nreconstructed by DLR with and without ZIP technique  or PIQE techniques (total \nfive MR data sets). Each standard protocol was dete rmined as conventional \nMR protocol reconstructed by DLR without ZIP techni que. To compare scan \ntime reduction and image quality improvement among all protocols, mean \nexamination time and signal-to-noise ratios (SNRs) were compared between \nstandard protocol and others by Dunnett's test. To evaluate qualitative image \nquality improvement, overall image quality, artifac t and diagnostic confidence \nlevel were assessed by 5-point scales and compared between standard \nprotocol and others by Steel's multiple comparison test. \nResults or Findings: Mean examination times of new MR protocols were \nsignificantly shorter than that of conventional pro tocols (p<0.05), although \nSNRs had no significant differences. As compared wi th standard protocol, \noverall image quality and artifact were significant ly improved by conventional \nprotocol reconstructed by DLR with ZIP and new prot ocol with PIQE (p<0.05). \nConclusion: PIQE is equal to or more useful for reduce examinat ion time and \nimage quality improvements as with DLR with and wit hout ZIP technique. \nLimitations: LImited study number and no diagnostic performance evaluation \nFunding for this study: Canon Medical Systems Corporation \nEthics committee - additional information: Fujita Health University Hospital \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Grant-in-A id for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Smoking Rese arch Foundation \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grant-in-Aid  for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Canon Medica l Systems \nCorporation \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Canon Medic al Systems \nCorporation Research/Grant Support: Smoking Researc h Foundation \n \n \nPower Grid Independent Low Field MRI \n*H-M. Klein*; Burbach/DE \n(mklein@greenscan-imaging.de) \n \nPurpose or Learning Objective: Purpose:  \nDevelop a concept for grid independent, power savin g MRI operation using a \npermanent magnet, solar energy, and a generator sup ported battery system. \nMethods or Background: We installed a 0,4 T MRI system with an open \ndesign permanent magnet. Regenerative energy is pro duced with a 29,8 kWp \nsolar array. To achieve grid independency, we insta lled an ´island solution´ \nusing a 22 kWh LiFePO4 battery. For longer periods of power outage, and \ninsufficient solar energy, a specially designed, di rect current (DC), high voltage \ndiesel generator is used. This generator simulates the power profile of a solar \narray, and is connected to the solar power converte r, feeding the battery. \nResults or Findings: Annual energy uptake of the MRI was 7.022 kWh in \n2023. RIS and PACS components consumed 4.959 kWh. H eating and air \nconditioning consumed 12.500 kWh. Total energy cons umption of the practice \nwas 26.801 kWh. Total energy production was 30.930 kWh. Energy balance \nwas positive with 4.129 kWh. Battery and DC generat or can provide power grid \nindependent operation. Without grid and solar energ y, the practice has an \nenergy consumption rate of max. 1,9 l gasoil/hour. \nConclusion: Grid independent, sustainable MRI operation is poss ible using \npermanent magnet technology, solar energy productio n, battery storage and a \nspecially designed power generator. \nLimitations: Only very few high quality low field MRI with perma nent magnet \ntechnology are available in the market. \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nHans-Martin Klein: Nothing to disclose \n \n \nReduced Energy Consumption with Accelerated MRI Usi ng Deep-\nLearning Reconstruction: A Phantom Study \n*Y. Jung*, R. Alizadeh, M. Corwin, L. Hacein-Bey, A . M. Hernandez; \nSacramento/US \n(yojung@ucdavis.edu) \n \nPurpose or Learning Objective: To quantify image quality, potential cost \nsavings, and greenhouse gas emission reductions in accelerated brain and \nprostate MRI exams using deep learning reconstructi on (DLR) and phantom \nimaging. \nMethods or Background: A data logger and current transformer sensor were \ninstalled upstream of the power distribution unit o n three 3T MRI platforms \n(GE, Siemens, United Imaging) to measure power cons umption at 1-second \nintervals. The ACR phantoms were scanned using the T2 FLAIR sequence \nfrom routine brain MRI and the T2-weighted sequence  from routine prostate \nMRI protocols. Phantom scanning was performed using  three scan times with \ndifferent acceleration factors, and images were rec onstructed using \nconventional inverse Fourier transform (IFT) and DL R at three strength levels: \nLow, Medium, and High. Signal-to-noise ratio (SNR) and low-contrast \n\n \n \nAbstract-based Programme \n \n 69  \nWednesday \ndetectability (LCD) measurements were taken using t he ACR phantom. Total \nenergy consumption was recorded for each acquisitio n. \nResults or Findings: Total energy consumption decreased monotonically wi th \nreduced scan time across all systems and protocols.  SNR and LCD, averaged \nacross all scan times, were generally higher for DL R compared to IFT, and \nSNR increased with increasing DLR strength. Reducin g the scan time by ~ 5 \nminutes with medium-strength DLR resulted in a ~65%  reduction in energy \nconsumption compared to a non-accelerated acquisiti on, while maintaining \ncomparable SNR and LCD. Extrapolating these savings  to all T2 FLAIR and \nT2-weighted sequences performed annually at our ins titution would result in \nestimated total savings of 15,944 USD and 74.1 MTCO 2e, equivalent to 16.6 \ngasoline-powered passenger vehicles driven for one year. \nConclusion: DLR-accelerated MRI exams provide substantial reduc tions in \ncost and greenhouse gas emissions without compromis ing image quality in \nphantom imaging experiments. \nLimitations: Patient data would be required to assess the actual  impact of \nDLR on image quality and energy savings in clinical  practice. \nFunding for this study: RSNA Emerging Issues Environmental Impact and \nSustainability grant \nEthics committee - additional information: None \nAuthor Disclosures:  \nRamsey Alizadeh: Nothing to disclose \nLotfi Hacein-Bey: Nothing to disclose \nYoungkyoo Jung: Nothing to disclose \nMichael Corwin: Nothing to disclose \nAndrew M Hernandez: Nothing to disclose \n \n \n16:30-17:30 Research Stage 4 \nResearch Presentation Session: \nInterventional Radiology \nRPS 609 \nDevelopments in vascular and \nneurovascular interventions \n \nModerator \nV. Bérczi; Budapest/HU  \n(berczi@hotmail.com) \n \n \nThrombolysis in basilar infarction (TIBI): A novel angiographic scale for \nevaluating mechanical thrombectomy in basilar arter y occlusion \n*M. E. Chevasco Hanze*, A. Lopez Rueda, A. Nuñez, E . Ripoll, V. Cuba,  \nS. Aixut, L. Aja, M. A. De Miquel Miquel, O. Chirif e;  \nL'Hospitalet de Llobregat/ES \n \nPurpose or Learning Objective: Acute Basilar Artery Occlusion (BAO) has \nthe highest morbidity and mortality in posterior ci rculation strokes. Current \nreperfusion scores for mechanical thrombectomy (MT)  are based on anterior \ncirculation strokes (mTICI score). This study propo ses a basilar artery-specific \nreperfusion score, based on digital subtraction ang iography (DSA), to assess \nMT efficacy in acute BAO. \nMethods or Background: A retrospective analysis was conducted on a \nprospective database of acute BAO patients treated with MT within 24 hours of \nsymptom onset at a stroke center from January 2014 to December 2023. \nInformed consent was obtained, and institutional re view board approval was \ngranted. Clinical, procedural, and radiological dat a were collected. The \nThrombolysis in Basilar Infarction (TIBI) score was  developed by grading \nposterior circulation territories in DSA post-MT, u sing PC-ASPECTS as \nreference (deducting 1 point for each occipital lob e/cerebellar lobe/thalamus or \n2 points for pons/mesencephalon). Successful recana lization was defined as \nTIBI≥8 and assessed by modified Rankin Scale at 90 days.  \nResults or Findings: Ninety-eight patients were included (median age 70,  56 \nmen). Successful recanalization (mTICI2b/3) was ach ieved in 84.7% and \nTIBI≥8 in 71.4%. Good functional status (mRS≤3) was seen in 50% of patients, \nwith better outcomes in mTICI 2b/3 and TIBI ≥8 groups (p <0.001 and p=0.002). \nAfter adjusting for age, NIHSS socre and Glasgow Co ma Sacle, TIBI≥8 was \nassociated with good outcomes at 90 days (OR = 6.18 ; p =0.001), and TIBI≥7 \nwas also linked to good outcomes (OR = 9.45; p <0.0 01). \nConclusion: The TIBI scale is a novel tool for evaluating MT ef ficacy in acute \nBAO. A TIBI≥8 should be the target for successful MT. \nLimitations: unweighted TIBI (<25% were TIBI<7) \nFunding for this study: No \nEthics committee - additional information: Observational study \n \nAuthor Disclosures:  \nAntonio Lopez Rueda: Nothing to disclose \nSonia Aixut: Nothing to disclose \nVictor Cuba: Nothing to disclose \nMiguel Emilio Chevasco Hanze: Nothing to disclose \nEnric Ripoll: Nothing to disclose \nOscar Chirife: Nothing to disclose \nAna Nuñez: Nothing to disclose \nMaria Angeles De Miquel Miquel: Nothing to disclose  \nLucía Aja: Nothing to disclose \n \n \nMTICI 2b-stopped or continued after first-pass: int erim results of the \nRossetti registry for M1 occlusion \n*J. I. García García*¹, O. Chirife¹, P. Vega Valdés ², E. Gonzalez³, F. Delgado⁴, \nG. Dolz¹, A. López-Frías López-Jurado⁵, F. Aparici Robles⁶, A. Lopez Rueda¹; \n¹Barcelona/ES, ²Oviedo/ES, ³Barcaldo/ES, ⁴Cordoba/ES, ⁵Madrid/ES, \n⁶Valencia/ES \n(juanignaciogarciarx@gmail.com) \n \nPurpose or Learning Objective: The purpose of this study is to analyze the \npredictive factors of safety and efficacy of additi onal mechanical thrombectomy \nin patients with acute ischemic stroke due to M1 oc clusion, who achieve \nmTICI2B recanalization after the first pass of endo vascular treatment. \nMethods or Background: We retrospectively analyzed patients with acute \nischemic stroke due to M1 occlusion from the ROSSET TI registry who \nachieved mTICI 2b recanalization after the first pa ss of endovascular \ntreatment. Patients were divided into two groups: t hose who stopped the \nprocedure with a mTICI 2b result and those who cont inued treatment for a \nbetter angiographic outcome. Among those who contin ued, patients were \nfurther split into two subgroups: those with unchan ged results (mTICI 2b) and \nthose with improved results (mTICI 2c/3). Demograph ic, clinical data, \nprocedure details, and outcomes were compared acros s groups. \nResults or Findings: We included 300 patients with acute ischemic stroke  M1 \nocclusion who achieved mTICI2b recanalization score  after the first pass of \nendovascular treatment. 132 patients underwent no f urther passes (group 1), \nwhile 168 patients underwent additional passes, wit h 65 of them maintaining a \nfinal mTICI 2b score (group 2) and 103 achieving a final mTICI 2c-3 score \n(group 3). Group 3 exhibited a higher incidence of distal embolism to new \nterritories compared to group 1 (7.8% vs. 0%; p < 0 .001). No significant \ndifferences in clinical outcomes were observed betw een the groups and \nsubgroups. \nConclusion: Patients who achieved an improved mTICI score after  the first \nrecanalization attempt (mTICI 2c/3) had a higher in cidence of distal embolisms \nto new territories compared to patients with no fur ther passes. No significant \ndifferences in mRS scores at 3 months were observed  between groups and \nsubgroups. \nLimitations: Inherent limitations of retrospective designs. \nFunding for this study: None. \nEthics committee - additional information: ROSSETTI registry creation was \napproved by an ethics committee. \nAuthor Disclosures:  \nGuillem Dolz: Nothing to disclose  \nPedro Vega Valdés: Nothing to disclose \nEva Gonzalez: Nothing to disclose \nFernando Aparici Robles: Nothing to disclose \nAlfonso López-Frías López-Jurado: Nothing to disclo se \nJuan Ignacio García García: Nothing to disclose \nAntonio Lopez Rueda: Nothing to disclose \nFernando Delgado: Nothing to disclose  \nOscar Chirife: Nothing to disclose \n \n \nComparison of DSA Morphology Parameters In Predicti ng Time To \nRecanalization of Internal Carotid Artery Saccular Aneurysms Treated \nWith Primary Coiling: Does Neck Angle Matter? \n*H. Akkaya*, A. I. Soylu, F. Uzunkaya; Samsun/TR \n(dr.hsynakkaya@gmail.com) \n \nPurpose or Learning Objective: Coil embolization is the most commonly \nused method of endovascular treatment of narrow-nec k saccular aneurysms. \nHowever, recanalization and subsequent aneurysm enl argement and rupture \nare common in aneurysms embolized only with coils. The aim of this study was \nto investigate which of the morphology findings dur ing the treatment of internal \ncarotid artery (ICA) aneurysms treated with primary  coiling is more successful \nin predicting the time to recanalization. \nMethods or Background: In this study, DSA images of 51 ICA aneurysms \ntreated with coiling in our center between January 2016 and July 2024 were \nretrospectively analyzed. Dates of embolization and  recanalization times were \nnoted. The segment of the ICA in which the aneurysm s were located, the \nheight of the aneurysm, the diameter of the neck, t he height/diameter ratios, \nwhether there was a parent artery originating from the aneurysm, and the \n\n \n \nAbstract-based Programme \n \n 70  \nWednesday \nangle of the neck of the aneurysm were noted. The r elationship between these \nparameters and recanalization times was examined. \nResults or Findings: The mean age of the patients was 59.6±13.9 years. 2 4 \n(47.1 %) patients had recanalization in follow-up e xaminations. The mean \nduration of recanalization was 12.73±1.7 months. Th e aneurysm neck angle \nwas found to be higher in patients with recanalizat ion (p<0.001). A negative \n(inverse) moderate correlation was found between th e time of recanalization \nand aneurysm size/neck diameter ratio: aneurysm nec k angles (r=-0.425; r=-\n0.537, respectively). \nConclusion: The angle of the neck of the saccular aneurysm at t he time of \ntreatment is one of the angiographic morphology fin dings that are successful in \npredicting the time to recanalization. \nLimitations: The study has some limitations. First of all, the s tudy was single-\ncentered and the number of patients was small. Anot her limitation is that only \naneurysms localized in the internal carotid artery are evaluated. \nFunding for this study: N/A \nEthics committee - additional information: Ondokuz Mayıs University Ethics \ncommittee approval was obtained for the study . The  requirement for informed \nconsent from the patients was waived due to the ret rospective nature of the \nstudy. \nAuthor Disclosures:  \nAyşegül Idil Soylu: Nothing to disclose \nFatih Uzunkaya: Nothing to disclose \nHüseyin Akkaya: Nothing to disclose \n \n \nRadiation Dose Comparison in Endovascular Clot Retr ieval: General \nAnaesthetic versus Conscious Sedation Approaches \n*F. Taylor*, K. Sehgal, D. Carrion, M. Masterson, M . K. Badawy, L-A. Slater; \nMelbourne/AU \n(fergustay@gmail.com) \n \nPurpose or Learning Objective: Endovascular clot retrieval (ECR) is a time-\nsensitive, adjunct treatment for large vessel occlu sion (LVO) strokes. We \npresent the results of this single high-volume stro ke centre, to establish \ndifferences in radiation doses between ECR cases pe rformed under general \nanaesthetic (GA) and conscious sedation (CS). \nMethods or Background: All ECR cases in adult patients between October \n2018 and June 2023 were included. Procedure records , patient characteristics, \nstroke outcome, and radiation dosimetry measures we re collected and \nanalysed retrospectively. Summative data, i.e. medi an and interquartile range \nfor overall dosimetry measures were described overa ll and by subgroup \naccording to anaesthetic strategy. ANOVA was used t o compare radiation \ndose measures, clot location, stroke severity measu res and endovascular clot \nretrieval times between anaesthetic strategy. \nResults or Findings: Radiation doses were higher in patients receiving a  GA; \nDAP 9,346 cGy.cm2, (4,996, 17,442) vs. 7,052 cGy.cm 2 (4,529, 10,926), \nreference air kerma 627 mGy (325, 1,287) vs 507 (30 0, 789), total fluoroscopy \ntime 34 minutes (19, 63) vs 31 (18, 47). This was a ssociated with a greater \npre-reperfusion morbidity (NIHSS of 13 (7, 19) vs 1 1 (6, 17), more complex \nthrombus location (more tandem, less M2, more ICA, more basilar clots) and \ndelayed presentation to hospital (227 minutes (101,  452) vs. 125 (66, 281), p \n<0.001). \nConclusion: We have demonstrated higher radiation doses under G A versus \nCS cases and some of the factors that may contribut e to this. These values \nserve as benchmark ranges for comparable centres pe rforming ECR. Larger \nmulticentre analysis is required to establish more generalisable dose reference \nlevels. \nLimitations: Single centre cohort. Affected by the COVID pandemi c; \ndepartmental decision to preference GA at several p oints between March \n2020-October 2021. TICI grading wasn't core lab adj udicated. \nFunding for this study: This research did not receive any specific grant fr om \nfunding agencies in the public, commercial, or not- for-profit sectors. \nEthics committee - additional information: All procedures performed in \nstudies involving human participants were in accord ance with the ethical \nstandards of the institutional and/or national rese arch committee and with the \n1964 Helsinki declaration and its later amendments or comparable ethical \nstandards. For this type of study formal consent is  not required. \nEthics approval was obtained from the local Human R esearch Ethics \nCommittee (HREC), reference no. QA/91812/MonH-2022- 343316 and the \nneed for informed consent was waived. \nAuthor Disclosures:  \nFergus Taylor: Nothing to disclose \nMohamed Khaldoun Badawy: Nothing to disclose \nDaniel Carrion: Nothing to disclose \nKunal Sehgal: Nothing to disclose \nLee-Anne Slater: Nothing to disclose \nMaeve Masterson: Nothing to disclose \n \n \n \nLong-term outcome results after endovascular abdomi nal aneurysm \nrepair with Zenith stent-graft \n*J. Reymen*, H. Mufty, A. Laenen, S. Houthoofd, G. Maleux; Leuven/BE \n \nPurpose or Learning Objective: To report the long-term outcome of patients \npresenting with an aortic or aortoiliac aneurysm tr eated with the Zenith \nEndoprosthesis. \nMethods or Background: A retrospective analysis of the collected data of 2 09 \nconsecutive patients who underwent endovascular ane urysm repair (EVAR) \nwith the Zenith Endoprosthesis (Cook Medical) betwe en January 1998 and \nDecember 2009 in an academic, tertiary care centre for aortic disease was \nperformed. Patients’ imaging and clinical follow-up  was performed until \nDecember 2023 to ensure a follow-up time of at leas t 14 years. The primary \nend points were overall survival and reintervention -free survival. Secondary \nend points were endograft-related complications, en doleaks, and \nreinterventions. \nResults or Findings: Overall survival at 2-, 5-, 10-, and 15-year follow -up was \n89.00%, 71.29%, 39.71% and 18.18%, respectively, wi th n=2 (1%) aneurysm-\nrelated deaths. Freedom from type I and III endolea k at 5- and 15-years was \n92.82% and 86.12%, respectively; type I (n=37 ; 17. 7%) and type III (n=4 ; 2%) \nendoleaks occurred in the follow-up period until 8 years postoperatively. \nReintervention-free survival was 83.08%, 75.62%, 68 .66% and 66.17% at 2-, \n5-, 10-, and 15-year follow-up, respectively. Reint erventions occurred meanly \nin the 0- to 8-year follow-up period. \nConclusion: Endovascular aneurysm repair using the Zenith Endop rosthesis \nis effective and durable on long-term follow-up, wi th acceptably low endograft-\nrelated complications and reinterventions. The numb er of adverse events and \nreinterventions is minimal after 8 years of follow- up. \nLimitations: It is a single-centre retrospective study, suscepti ble to selection \nbias. The study also has a relatively small patient  cohort ( n=209) because not \nall patients eligible for EVAR were treated with th e Zenith Endoprosthesis. \nFinally, no comparison with other endograft devices  was made. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The institutional ethics committee \napproved this retrospective analysis (MP024667). \nAuthor Disclosures:  \nAnnouschka Laenen: Nothing to disclose \nSabrina Houthoofd: Nothing to disclose \nGeert Maleux: Nothing to disclose \nHozan Mufty: Nothing to disclose \nJessie Reymen: Nothing to disclose \n \n \nAssessing abdominal aortic aneurysm growth by using  radiomics of \ndifferent radii of perivascular adipose tissue afte r endovascular repair \n*R. Lv*, G. Hu, S. Zhang, Z. Zhang, Z. Wang; Beijin g/CN \n(18763896097@163.com) \n \nPurpose or Learning Objective: To investigate the relationship between \nradiomic features of different radii of perivascula r adipose tissue (PVAT) and \nabdominal aortic aneurysm (AAA) growth after endova scular aneurysm repair \n(EVAR). \nMethods or Background: Patients with sub-renal AAA who underwent regular \nfollow-up after EVAR from September 2014 to Septemb er 2024 were \nretrospectively collected. Two radiologists segment ed the aneurysm and \ndifferent radii of PVAT (PVAT1-7: 5mm, 7.5mm, 10mm,  12.5mm, 15mm, \n17.5mm, 20mm) to evaluate the aneurysm volume chang es during follow-up \nand calculate radiomic features of different PVAT r egions. Univariable and \nmultivariable logistic regression was performed to construct models to evaluate \nthe growth of AAA based on the radiomic features of  PVAT with different radii \n(5mm-20mm for models 1-7). Calculate the area under  the curve (AUC), \nsensitivity, specificity, and accuracy of the model s, and test the significance of \nthe performance differences among the models. \nResults or Findings: A total of 79 patients (67±7 years, 82% men) were \nenrolled in this study, 20 of whom had a growing an eurysm. The AUC for \nModels1-7 are 0.72, 0.77, 0.71, 0.70, 0.69, 0.71, a nd 0.75, respectively. The \nspecificities for Models1-7 are 95%, 95%, 97%, 97%,  93%, 92%, and 97%, \nrespectively. DeLong test and McNemar test: p > 0.0 5 (no statistical \nsignificance). \nConclusion: The models constructed using the radiomic features of PVAT with \ndifferent radii after EVAR showed no significant di fferences in performance for \nevaluating AAA growth. The models achieved an avera ge specificity of 95%, \nindicating their effectiveness in minimizing the mi sclassification of non-growing \nAAA cases as growth cases. \nLimitations: The retrospective research highlights the need for prospective \nand long-term follow-up studies. There are also lim itations in the development \nof imaging segmentation techniques, which still req uire manual intervention for \nthree-dimensional aneurysm segmentation. \nFunding for this study: This study has received funding by the National Hig h \nLevel Hospital Clinical Research Funding (2022-PUMC H-B-068). \n \n\n \n \nAbstract-based Programme \n \n 71  \nWednesday \nEthics committee - additional information: Institutional review board of \nPeking Union Medical College Hospital. \nAuthor Disclosures:  \nGe Hu: Nothing to disclose \nZhe Zhang: Nothing to disclose \nZhiwei Wang: Nothing to disclose \nShenbo Zhang: Nothing to disclose \nRui Lv: Nothing to disclose \n \n \nTowards Clinical Magnetic Particle Imaging: Safety Measurements of \nMedical Implants in an Extracorporeally-Perfused Hu man Cadaver Model \n*F. Wegner*¹, T. Friedrich¹, P. Elfers¹, F. Kleefel dt², D. Peter², P. Gruschwitz², \nT. Kampf², P. Vogel², V. Hartung²; ¹Lübeck/DE, ²Wür zburg/DE \n(franz.wegner@uksh.de) \n \nPurpose or Learning Objective: Magnetic Particle Imaging (MPI) is an \nemerging, tracer-based, 3D imaging modality on the way to clinical application. \nIt offers high temporal resolution and operates wit hout the use of ionizing \nradiation, making it particularly advantageous for cardiovascular imaging and \nreal-time interventional monitoring. However, the p otential heating of metallic \nmedical devices within the magnetic fields of MPI s canners is a critical safety \nconcern. This study aimed to assess the thermal beh avior of commercially \navailable medical implants during MPI-scans in a hu man cadaver model. \nMethods or Background: A fiberoptic thermometer probe was introduced into \nthe superficial femoral artery (SFA) of a human cad aver model via a 7 F \nsheath. A series of commercially available endovasc ular implants (including six \nstents, five coils, and one vascular plug) were the n positioned sequentially \nwithin the SFA. The thermometer probe was retracted  sequentially to ensure \ndirect contact with each implant. Additionally, a h ole was drilled in the femur, \nwhich contained a gamma nail, and the fiberoptic pr obe was externally inserted \nto establish contact with the nail. A custom-built human-sized MPI-scanner for \ninterventional purpose was positioned around the ca davers’ thigh, and an MPI-\nsequence consisting of 40 pulses (4 pulses per seco nd) was applied, with the \nrespective implant centrally located within the sca nner. Throughout the MPI-\nsequence, the cadavers’ thigh was perfused extracor poreally with a blood-\nequivalent fluid using an external flow pump. \nResults or Findings: The gamma nail exhibited a temperature increase of \n0.04 K during the MPI-sequence, while no detectable  heating was observed in \nany of the endovascular devices tested. \nConclusion: Commonly used medical implants do not heat up signi ficantly in a \nhuman-sized MPI-scanner under realistic conditions.  \nLimitations: Only a limited number of commercial devices was tes ted in this \nwork. \nFunding for this study: N/A \nEthics committee - additional information: Protocol Number 20220413 01 \nAuthor Disclosures:  \nFranz Wegner: Nothing to disclose \nViktor Hartung: Nothing to disclose \nFlorian Kleefeldt: Nothing to disclose \nPhilipp Gruschwitz: Nothing to disclose \nThomas Friedrich: Nothing to disclose \nDominik Peter: Nothing to disclose \nPatrick Elfers: Nothing to disclose \nPatrick Vogel: Nothing to disclose \nThomas Kampf: Nothing to disclose \n \n \nElectroporation with local or systemic bleomycin fo r the treatment of \nvascular malformations: early results of a prospect ive study \n*N. Papalexis*, G. Peta, M. Di Carlo, S. Quarchioni , L. Campanacci, M. Carta, \nM. Miceli, G. Facchini; Bologna/IT \n(nicolaspapalexis@gmail.com) \n \nPurpose or Learning Objective: Purpose: To evaluate the safety and efficacy \nof electrochemotherapy with bleomycin for the treat ment of soft tissue vascular \nmalformations. \nMethods or Background: Materials and Methods: This study analyzes the \nearly results of a prospective study “BESVAM”, desi gned to prospectively \nevaluate the safety and efficacy of electrochemothe rapy for vascular \nmalformations. 18 patients were enrolled from Febru ary 2023 to July 2024. \nBleomycin was injected intralesionally for low-flow  vascular malformations or \nsystematically for high-flow vascular malformations . The primary goal was pain \ncontrol, measured in VAS score at 3,6, and 12 month s follow-up. The \nsecondary goal was the size reduction of the lesion  and variations in the QLQ \nquestionnaire. \nResults or Findings: Results: Twelve patients received bleomycin \nintravenously and six patients intralesional. Basel ine VAS scores averaged 6.9 \n(SD 2.1), decreasing to 3,1 (SD 3.2) at three month s. Further reduction was \nobserved at 6 and 12 months with scores of 2.1 (SD 2.0) and 1.6 (SD 2.0) \nrespectively. Size was reduced from a mean of 155.1  cm3 (range 56.7 to 515.3 \ncm3) pre-treatment to a mean of 122.5 cm3 (range 42 .2 to 438.9 cm3) at the 6-\nmonth follow-up (p<0.05). (12 months) Twelve patien ts discontinued pain relief \ntherapy. Ten patients experienced skin discoloratio n at the site of insertion of \nthe needles. \nConclusion: Conclusion: The preliminary results are promising, suggesting \nthat electroporation with local or systemic bleomyc in could be a safe and \neffective tool for the management of vascular malfo rmations. \nLimitations: Small sample size, lack of control group \nFunding for this study: None \nEthics committee - additional information: Prospective study approved by \nthe local ethcis committee of Emilia Romagna, Italy . \nAuthor Disclosures:  \nMarco Miceli: Nothing to disclose \nMichela Carta: Nothing to disclose \nLaura Campanacci: Nothing to disclose \nGiancarlo Facchini: Nothing to disclose \nGiuliano Peta: Nothing to disclose \nSimone Quarchioni: Nothing to disclose \nMaddalena Di Carlo: Nothing to disclose \nNicolas Papalexis: Nothing to disclose \n \n \n \n \n \n \n \n\n \n \n 72  \n \n \n \n  \nThursday, February 27 \n\n \n \nThursday \nAbstract-based Programme \n \n 73  \n \n08:00-09:30 Research Stage 1 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 701 \nImaging of the intestines with focus on \nCrohn's disease \n \nModerator \nS. A. Taylor; London/UK  \nAuthor Disclosures:  \nStuart A. Taylor: Advisory Board: aztrazeneca; Gran t Recipient: takeda; Share \nHolder: Motilent \n \n \nImprovement of diagnostic performance in low-dose C T enterography: \nthe impact of an artificial intelligence iterative reconstruction algorithm \nR. Guo¹, W. Zhou², G. Zhang², *T. Wang*², P. Hu¹, Q . Liang¹, P. Rong¹; \n¹Changsha/CN, ²Shanghai/CN \n(tiantian.wang@cri-united-imaging.com) \n \nPurpose or Learning Objective: To investigate the clinical value of artificial \nintelligence iterative reconstruction (AIIR) in opt imizing image spatial resolution \nand diagnostic performance of low-dose CT enterogra phy (CTE) for patients \nwith Crohn's disease (CD), compared with the routin e hybrid iterative \nreconstruction (HIR). \nMethods or Background: Forty patients with suspected CD were \nprospectively enrolled to receive low-dose CTE (80k Vp, ref 120mAs) with \nilecolonoscopy-guided biopsy as the reference. Imag es were reconstructed \nusing the AIIR and HIR. Diagnosis of CD was made on  a per-segment level. \nThe diagnostic confidence was scored with a five-po int scale (1=insufficient, \n5=definitely confirmed). Signal-to-noise ratio (SNR ), contrast-to-noise ratio \n(CNR), and edge rise slope (ERS) of the bowel wall were measured. \nDiagnostic image quality, including conspicuity of bowel wall enhancement, \nthickness, luminal narrowing, comb signs, and creep ing fat signs, was \nevaluated with a five-point scale (1=poor, 5=excell ent). \nResults or Findings: There were 77 bowel segments confirmed with CD, of \nwhich 57 were detected on HIR images and 71 on AIIR  images. AIIR images \nshowed 124.3% higher SNR, 136.3% higher CNR, 68.4% higher ERS, and \nmore conspicuous diagnostic imaging features (all p <0.001), indicating higher \nimage spatial resolution than HIR. With higher spat ial resolution, AIIR images \nshowed better diagnostic performance for CD detecti on than HIR images \n(sensitivity: 88.7% vs 71.2%; accuracy: 92.9% vs 87 .9%; specificity: 94.5% vs \n94.5%; false-positive-rate: 13.4% vs 16.2%; false-n egative-rate: 4.5% vs \n10.9%). The diagnostic confidence was significantly  improved by AIIR (4.5±0.6 \nvs 3.4±0.6, p<0.001). \nConclusion: AIIR improved the image spatial resolution of low-d ose CTE and \nthus delivered higher diagnostic confidence and bet ter diagnostic performance \nthan HIR. Low-dose CTE with AIIR provides excellent  image quality and \nreliable CD detection, making it a feasible option for follow-up examinations for \nCD patients. \nLimitations: A single-centre study. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study received approval \nfrom the Institutional Review Board (R20230019). \nAuthor Disclosures:  \nPengfei Rong: Nothing to disclose \nTiantian Wang: Nothing to disclose \nWanhui Zhou: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nRui Guo: Nothing to disclose \nQi Liang: Nothing to disclose \nPengzhi Hu: Nothing to disclose \n \n \nComparison of Conventional vs. Abbreviated MR Enter ography in \nCrohn's Disease: Assessment of Inter-Radiologist Ag reement for \nCategorizing Disease Activity and Complications \n*J. R. Rimola Gibert*¹, C. Saavedra¹, N. Capozzi², I. De Kock³, A. R. Radmard⁴, \nM. Scharitzer⁵, M. C. Masamunt¹, I. Ordás¹, J. Dillman ⁶; ¹Barcelona/ES, \n²Bologna/IT, ³Ghent/BE, ⁴Tehran/IR, ⁵Vienna/AT, ⁶Cincinnati, OH/US \n(jrimola@clinic.cat) \n \nPurpose or Learning Objective: Crohn's disease (CD) often requires multiple \nimaging evaluations. An abbreviated MR enterography  (aMRE) protocol, \nwithout IV contrast, could improve patient experien ce and reduce costs. This \nstudy aims to compare inter-observer agreement for detecting CD activity and \ncomplications using conventional MRE protocol (cMRE ) versus an abbreviated \nprotocol (aMRE). \nMethods or Background: Ten radiologists from six countries independently \nreviewed cMRE and aMRE exams from 80 CD patients, w ith imaging \nassessments separated by at least one month. The ex ams included both pre- \nand post-treatment images of CD patients. Interobse rver agreement, Fleiss' \nKappa statistics, and Gwet's concordance 1 (AC1), w hich corrects the \nprevalence dependence in categorizations, were calc ulated at the bowel \nsegment and patient levels for the presence of diff erent findings indicative of \nactive inflammation and complications. \nResults or Findings: Overall (n=80), the concordance for detecting disea se \nactivity was high and comparable between aMRE and c MRE: agreement was \n0.83 (0.80-0.86) for aMRE vs. 0.84 (0.82-0.87) for cMRE; Kappa values were \n0.41 (0.29-0.53) vs. 0.36 (0.23-0.49); and AC1 valu es were 0.76 (0.64-0.88) \nvs. 0.79 (0.69-0.89). Agreement for detecting activ e disease was similar across \nthe small bowel and colon, as well as between pre-t reatment (n=51) and post-\ntreatment (n=29) MREs. For detecting strictures, ag reement was 0.76 (0.73-\n0.78) for aMRE vs. 0.72 (0.68-0.75) for cMRE; Kappa  values were 0.35 (0.24-\n0.46) vs. 0.34 (0.24-0.43); and AC1 values were 0.6 1 (0.47-0.74) vs. 0.50 \n(0.35-0.65). For penetrating complications, agreeme nt was 0.81 (0.79-0.84) for \naMRE vs. 0.85 (0.82-0.87) for cMRE; Kappa values we re 0.47 (0.34-0.59) vs. \n0.57 (0.45-0.69); and AC1 values were 0.71 (0.56-0. 86) vs. 0.76 (0.63-0.90). \nConclusion: The interobserver agreement for detecting active CD  and related \ncomplications using an aMRE protocol was comparable  to that of the cMRE \nprotocol that supports the adoption of abbreviated MRE protocols. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Local ethics committee approved \nthe study with the code HCB/2021/0629 \nAuthor Disclosures:  \nMaria Carme Masamunt: Nothing to disclose \nCarolina Saavedra: Nothing to disclose \nIngrid Ordás: Nothing to disclose \nIsabelle De Kock: Nothing to disclose \nNunzia Capozzi: Nothing to disclose \nJordi Rimola Rimola Gibert: Advisory Board: Janssen  Consultant: \nAstraZeneca, Janssen Alimentiv, Clario, Lument, Ori go Research/Grant \nSupport: Abbvie \nMartina Scharitzer: Nothing to disclose \nAmir Reza Radmard: Nothing to disclose \nJonathan Dillman: Nothing to disclose \n \n \nInflammatory Burden in Crohn’s Disease: Insights fr om PET/MR \nEnterography \n*N. Bogveradze*, K. Kranz, T. Traub-Weidinger, C. P rimas, A. Macher-Beer, \nW. Reinisch, T. Mang, M. Hacker, M. Scharitzer; Vie nna/AT \n(bogveradze.nino@gmail.com) \n \nPurpose or Learning Objective: To evaluate the global and regional \ninflammatory burden in patients with Crohn’s Diseas e (CD) using FDG \nPET/MR enterography (PET/MRE), in correlation with histopathological \nfindings and relevant biomarkers. \nMethods or Background: Patients with CD undergoing PET/MRE and \nileocolonoscopy were included in this retrospective  study between 2016-2021. \nEight intestinal segments were manually segmented, and the uncorrected total \nlesion glycolysis values (TLG) were summed to calcu late global CD activity \nscore (GCDAS). GCDAS and highest SUVmax/ patient we re correlated with \nclinical biomarkers (fecal calprotectin [FC], serum  C-reactive protein [CRP], \nperipheral blood leukocyte counts, Harvey-Bradshaw index [HBI]) to evaluate \nglobal inflammation. For assessing regional inflamm ation, SUVmax was \ncorrelated with histopathological disease activity.  Comparisons were made \nusing Spearman's coefficient and Wilcoxon-W tests. \nResults or Findings: In 41 patients (mean age, 40 years ±14 (SD), 26 men ), \nthe highest segmental SUVmax correlated significant ly with FC (r= 0.443 p < \n0.004) and CRP (r = 0.645, p < 0.001). The GCDAS co rrelated significantly \nwith CRP (r = 0.498, p = 0.01) but not with FC (p =  0.49), leukocyte counts \n(p=0.56) or HBI (p = 0.518). SUVmax values signific antly correlated with \ngrading of active inflammation in corresponding his topathological samples (r = \n0.515, p < 0.001) and were higher in segments with severe histopathological \ninflammation (4.7, 95% CI: 3.8-5.6) compared to seg ments with moderate (3.3, \n95% CI: 2.6-4.1) or without active inflammation (2. 0, 95% CI: 1.8-2.0). \nConclusion: SUVmax was found to be a reliable biomarker for ass essing \nglobal inflammatory burden and correlated well with  histopathological \nsegmental activity. Furthermore, SUVmax enabled gra ding of active \ninflammation. GCDAS showed good correlation with CR P, but not with FC and \nmay therefore be less useful for quantifying global  disease burden. \nLimitations: Retrospective study design \nFunding for this study: N/A \nEthics committee - additional information: Ek 1356/2023 \n \n\n \n \nThursday \nAbstract-based Programme \n \n 74  \nAuthor Disclosures:  \nKerstin Kranz: Nothing to disclose \nThomas Mang: Nothing to disclose \nTatjana Traub-Weidinger: Nothing to disclose \nChristian Primas: Nothing to disclose \nAndrea Macher-Beer: Nothing to disclose \nMartina Scharitzer: Nothing to disclose \nNino Bogveradze: Nothing to disclose \nMarcus Hacker: Nothing to disclose \nWalter Reinisch: Nothing to disclose \n \n \nMultiparametric and multi-score MRI evaluation of p ediatric Crohn’s \ndisease: characterization of the perianal fistulizi ng phenotype \n*A. Valenti*, F. Maccioni, L. Busato, L. Bottino, A . Longhi, C. Catalano; \nRome/IT \n(alessandra.valenti@uniroma1.it) \n \nPurpose or Learning Objective: Perianal fistulas are severe complications of \npediatric Crohn’s disease (CD), due to a high risk of demolitive surgery. MRI is \nthe gold standard for scoring intestinal and perian al CD. The purpose of this \nstudy was to stage severity and activity of both le sions using MRI scores in \npediatric CD. \nMethods or Background: A retrospective study was performed on 186 \npediatric patients. Inclusion criteria were: proven  CD, intestinal and perianal \ndisease, complete MRI. Forty patients were finally included. MR Enterography \n(MRE) and high-resolution MRI (HRMRI) of the perian al region were used to \nanalyze intestinal and perianal disease. The MEGS s core was applied to \nassess the severity of intestinal disease, the Park s’ classification and \nMAGNIFI-CD score to classify perianal disease and q uantify its activity. \nCorrelations between location and activity of intes tinal and perianal lesions \nwere investigated. \nResults or Findings: Simple perianal fistulas (Parks A and B) were found  in \n82.5% of patients, whereas complex fistulas (Parks C-E or a combination of \nthem) in 17.5 %. Jejunal, ileal and colonic lesions  were found in 22,5%, 80% \nand 35%, variably associated. Considering colorecta l lesions only, 70% were \nleft-sided. Disease of the left colon was associate d with a more extensive \ndisease, > 25 cm in length (p < 0.001), and a more severe activity, MEGS \ngrade 3. Correlations were found between left-sided  colonic lesions and severe \nfistulas, and between MAGNIFI-CD and MEGS scores bo th grade 3 (p<0.05). \nConclusion: The perianal fistulizing phenotype in pediatric pat ients is \ncorrelated with the severity of intestinal inflamma tion and with left-sided colonic \ndisease. \nLimitations: We used two differente MRI scanners to obtoin our d ata. \nFunding for this study: No funding. \nEthics committee - additional information: We have received the \napprovation of the ethics commitee of our Hospital \nAuthor Disclosures:  \nAlessandro Longhi: Nothing to disclose \nAlessandra Valenti: Nothing to disclose \nLorenza Bottino: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLudovica Busato: Nothing to disclose \nFrancesca Maccioni: Nothing to disclose \n \n \nMagnetic Resonance Imaging biomarkers in the diagno sis of \ngastrointestinal acute Graft-versus-Host-Disease \n*L. Busato*, F. Maccioni, A. Valenti, L. Bottino, A . Iori, U. La Rocca,  \nC. Catalano; Rome/IT \n(ludovica.busato@uniroma1.it) \n \nPurpose or Learning Objective: Acute gastrointestinal Graft-versus-Host \ndisease (GI-a GVHD) is one of the most severe compl ications stem cell \ntransplantation, occurring when the transplanted im mune cells attack the host's \nintestinal tissues. Aim of this study was to evalua te the efficacy of MRI \nbiomarkers in the diagnosis and staging of acute ga strointestinal GI-aGVHD, \ncurrently based on clinical and endoscopic criteria  only. \nMethods or Background: Thirty-five patients with clinical suspicion of GI-\naGVHD were retrospectively and prospectively analys ed, 21/35 retrospectively, \n14/35 prospectively, both separately and in conjunc tion. In these patients we \ninvestigated 15 MRI biomarkers suggestive of bowel inflammation and GVHD \nseverity. A diagnostic score was tested, based on t he most relevant GVHD \nbiomarkers: small bowel involvement, T2 and post-co ntrast bowel wall \nstratification, ascites , oedema of the retroperito neal and declivous tissues. \nResults or Findings: GI -GVHD was confirmed by biopsy in 13/35 patients \n(37%). Analysing the 6 key biomarkers (diagnostic s core), MRI showed 88.2% \nand sensitivity 100%specificity. In addition to the se siw biomarkers, wall \nstiffness and mesenteric oedema appeared useful for  differentiating GVHD \nfrom non-GVHD patients. Most commonly affected segm ents were the \nproximal, middle and distal ileum (82.3%). Consider ing all the 15 MRI \nBiomarkers, MRI showed high predictive value on dis ease severity and \nmortality, superior to the clinical score. \nConclusion: MRI is a noninvasive and accurate tool for the diag nosis of GI-\nGVHD,which can provide crucial information and impl ement current clinical and \nendoscopic criteria. Limitations: The small number of patients, due to the low \nprevalence of this disease. \nLimitations: The small number of patients, due to the low freque nce of the \ndisease. \nFunding for this study: No funding \nEthics committee - additional information: Our study has been approved by \nthe ethics committee of our hospital \nAuthor Disclosures:  \nAlessandra Valenti: Nothing to disclose \nUrsula La Rocca: Nothing to disclose \nAnnapaola Iori: Nothing to disclose \nLorenza Bottino: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLudovica Busato: Nothing to disclose \nFrancesca Maccioni: Nothing to disclose \n \n \nReevaluating MR-Enterography: Value or Overuse? \n*R. Martín-Márquez*¹, E. Gutiérrez Dorta², D. J. L.  Ruiz¹, J. Mesa¹; \n¹Córdoba/ES, ²Ourense/ES \n(rociomartinmarquez@gmail.com) \n \nPurpose or Learning Objective: To evaluate the profitability of Magnetic \nResonance Enterography (MRE) in different clinical contexts, determining if it is \nboth clinically and cost-effective, optimizing reso urces and improving patient \ncare. To establish a protocol for performing MRE in  various clinical scenarios. \nMethods or Background: A retrospective study of 615 patients who \nunderwent MRE at Reina Sofía Hospital (Córdoba, Spa in) over one year. \nPatients were divided into two groups: patients wit h known inflammatory bowel \ndisease (IBD) and those without IBD history. Variab les included the reason for \nthe examination, clinical unit, patient presentatio n, and findings from MRE, \ncolonoscopy, and intestinal biopsy. A descriptive a nalysis was performed, and \ndifferences were assessed with chi-square or Studen t's t-tests (p<0.05). \nDiagnostic indices of MRE were compared to colonosc opy (sensitivity, \nspecificity, positive predictive value (PPV), negat ive predictive value (NPV). \nResults or Findings: A total of 242 (39.3%) patients had a history of IB D, \nwhile 373(60.7%) did not. Pathological findings wer e seen in 68.2% of IBD \npatients versus 23.3% without IBD (p<0.05). 9.1% of  IBD patients with normal \ncolonoscopy had pathological findings on MRE, compa red to 10.4% without \nIBD (p>0.05). Among those with elevated fecal calpr otectin, MRE detected \nmore findings in IBD patients (86.4% vs. 18.8%; p<0 .05). Diarrhea occurred in \n8.7% of IBD patients and 37.3% of non-IBD patients,  with MRE findings in \n52.4% and 16.5%, respectively (p<0.05). Diagnostic indices of MRE: IBD \ngroup: Sensitivity 70.6%, Specificity 90.9%, PPV 92 .3%, NPV 66.7%. Non-IBD \ngroup: S 43.2%, E 89.5%, PPV 70.7%, NPV 73%. \nConclusion: Our study highlights the importance of MRE in IBD p atients and \nthe need to optimize its use in non-IBD patients, e specially with negative \ncolonoscopies. \nLimitations: The main limitations of this study are its retrospe ctive design, \nwhich may lead to missing information, and the lack  of long-term follow-up. \nFunding for this study: No \nEthics committee - additional information: There is no additional \ninformation. \nAuthor Disclosures:  \nJuan Mesa: Nothing to disclose  \nEduardo Gutiérrez Dorta: Nothing to disclose  \nRocío Martín-Márquez: Nothing to disclose  \nDaniel José López Ruiz: Nothing to disclose \n \n \nThe diagnostic yield of non-contrast versus contras t-enhanced magnetic \nresonance enterography (MRE) for small bowel (SB) e valuation in \nundiagnosed patients: Experience from four centres in the UK \n*S. Martin*, J. Pancholi, S. Liong; Manchester/UK \n(sarahh.martinn93@gmail.com) \n \nPurpose or Learning Objective: This service evaluation compares the \ndiagnostic yield and utilisation trends of non-cont rast versus contrast-enhanced \nmagnetic resonance enterography (MRE) for small bow el (SB) evaluation in a \ncohort of patients without a diagnosis of inflammat ory bowel disease or clear \ngastrointestinal symptom aetiology. \nMethods or Background: A retrospective review of all 1,012 MREs performed \nacross four hospitals in Greater Manchester, United  Kingdom, between 1 \nJanuary and 31 December 2023, identified 208 undiag nosed patients. Among \nthese, 92(44.2%) underwent non-contrast and 116(57. 8%) underwent contrast-\nenhanced MRE. Diagnostic yield was assessed by revi ewing MRE reports and \ncorrelating findings with colonoscopic histopatholo gy results (within 6 weeks of \nMRE) and faecal calprotectin levels (FC, within 12 weeks of MRE). \n\n \n \nThursday \nAbstract-based Programme \n \n 75  \nResults or Findings: SB abnormalities were observed in a similar minorit y of \npatients in both the contrast (14/116 [12.1%]) and non-contrast (5/92 [5.4%], \np=0.15) groups. Of 45 patients with colonoscopies c ompleted within 6 weeks of \nMRE, 22 had terminal ileum (TI) biopsies available (non-contrast: n=8; \ncontrast-enhanced: n=14). Compared with TI biopsies , contrast-enhanced \nMRE has sensitivity of 100% (95%CI 15.8-100%) and s pecificity of \n83.3%(95%CI 51.5-97.9%), whereas non-contrast MRE h as sensitivity of \n100%(95%CI 2.5-100%) and specificity of 100%(95%CI 59-100%). Contrast-\nenhanced MRE has positive predictive value (PPV) of  50%(95%CI 22-78%) \nand accuracy of 85.7(95%CI 57.2-98.2%). Non-contras t MRE has PPV \n100%(95%CI 2.5-100%) and accuracy of 100%(95%CI 63. 1-100%). FC was \navailable for 27 patients, but there was no clear r elationship between FC levels \nand MRE results. \nConclusion: Contrast enhancement did not significantly alter th e diagnostic \nyield of small bowel pathology in our cohort and is  known to require longer \nacquisition and reporting times, impacting service capacity. Non-contrast MRE \nhas high sensitivity and specificity for diagnosis of SB pathology. \nLimitations: Retrospective design. Small number of patients with  \nhistopathology and FC available. \nFunding for this study: None. \nEthics committee - additional information: Ethics approval was not required \nas this was an educational project and retrospectiv e service evaluation. \nAuthor Disclosures:  \nJay Pancholi: Nothing to disclose \nSue Liong: Nothing to disclose \nSarah Martin: Nothing to disclose \n \n \nUltrasound is more effective than MRI for monitorin g the response to \nmedical treatment in patients with active ileocolon ic Crohn's disease – a \nprospective blinded multicenter study \nJ. Brodersen¹, *S. R. Rafaelsen*², M. Agerbæk Jue¹,  T. Knudsen¹, J. Keldsen³, \nM. D. Jensen¹; ¹Esbjerg/DK, ²Vejle/DK, ³Odense/DK \n(soeren.rafael.rafaelsen@rsyd.dk) \n \nPurpose or Learning Objective: The aim of this study was to evaluate \nintestinal ultrasound (IUS), magnetic resonance ima ging enterocolonography \n(MREC), panenteric capsule endoscopy (PCE) and faec al calprotectin (FC) for \ndetermining response to medical treatment in patien ts with ileocolonic CD \nMethods or Background: This prospective, blinded, multicentre study \nincluded patients with endoscopically active CD. Pa tients were scheduled for \nIC, MREC, IUS, PCE and FC before and 12 weeks after  medical treatment. \nThe vascularity within the affected bowel wall area s was assessed according to \nthe Limberg score. The Simple Ultrasound Score for Crohn’s Disease (SUS-\nCD) was used for activity assessment. A > 50% reduc tion of the Simple \nEndoscopic Score for Crohn’s Disease (SES-CD) with IC defined treatment \nresponse as gold standard. \nResults or Findings: From 2018 to 2024, 50 patients completed the pre- a nd \npost-treatment evaluation with IC, and endoscopic r esponse was achieved in \n25 (50.0%). PCE was omitted in 12 (24.0%) patients because of stricturing CD. \nAll activity scores decreased in patients achieving  endoscopic response: The \nSimple Ultrasound Score for Crohn’s Disease 2.2 vs.  6.1 (P < 0.001), Magnetic \nResonance Index of Activity 29.0 vs. 37.1 (P = 0.05 ), SES-CD with PCE 3.1 vs. \n12.8 (P < 0.001) and FC 115.3 vs. 1339.9 mg/kg (P <  0.001). The sensitivity \nand specificity of IUS, MREC, PCE and FC was 80.0% (95% CI 56.3-94.3) / \n77.8% (52.4-93.6), 65.2% (42.7-83.6) / 87.0% (66.4- 97.2), 87.5% (61.7-98.4) / \n86.7% (59.5-98.3) and 90.0% (68.3-98.8) / 86.4% (65 .1-97.1), respectively. \nConclusion: IUS, PCE and FC are equally effective for determini ng \nendoscopic response in patients with active CD. MRE C is insufficient for \ndetermining endoscopic response. \nLimitations: First, the sample size is limited. Second, IC serve d as gold \nstandard for treatment response, which may favour m odalities assessing \nmucosal inflammation (PCE and FC). \nFunding for this study: The study was initiated by the investigators withou t \nfunding from medical imaging companies or the capsu le endoscope \nmanufacturer. \nEthics committee - additional information: The study was approved by the \nLocal Ethics Committee of Southern Denmark (S-20170 188). All patients gave \ninformed consent before participation. The study wa s registered: \nNCT03435016. \nAuthor Disclosures:  \nMie Agerbæk Jue: Nothing to disclose \nMichael D. Jensen: Nothing to disclose \nJacob Brodersen: Nothing to disclose \nJens Keldsen: Nothing to disclose \nTorben Knudsen: Nothing to disclose \nSören R. Rafaelsen: Nothing to disclose \n \n \n \n \nGut feels emotion: Psychological distress is associ ated with alterations \nof magnetic resonance enterography in patients with  Crohn's disease \n*Y. Ke*, R. Zhang, H. Cai, Q. Zeng, S-T. Feng, Z. P eng, X. Li; Guangzhou/CN \n(keyq7077@163.com) \n \nPurpose or Learning Objective: Psychological distress may affect bowel \ndisease activity in patients with Crohn's disease ( CD). However, limited studies \nhave investigated its correlation with trans-/peri- intestinal alterations in CD. \nTherefore, we aimed to investigate the relationship  between psychological \ndistress and intestinal abnormalities identified by  magnetic resonance \nenterography (MRE), and to explore their underlying  association using blood \nneurotransmitters. \nMethods or Background: 105 CD patients and 46 healthy controls (HCs) \nwere prospectively recruited. CD patients underwent  MRE and provided blood \nsamples for 19 serum neurotransmitters measurement.  All participants \ncompleted State-Trait Anxiety Inventory (including STAI-Trait and State \nscores), Beck Depression Inventory (BDI), and Perce ived Stress Scale (PSS) \nquestionnaires to assess psychological distress. Co rrelation analysis, \nmultivariable logistic regression, and causal media tion analyses were \nemployed to investigate the relationship between ps ychological distress and \nMRE features. \nResults or Findings: Psychological scores of CD patients, including STAI -\nTrait, PSS, and BDI scores, were significantly high er than HCs (all P<0.001). \nAmong them, STAI-Trait score was significantly corr elated with stricture \n(r=0.505), mural T2WI hyperintensity (r=0.466), per ianal diseases (r=0.359), \nand perienteric effusion (r=0.340) (all P<0.05). Mu ltivariable logistic regression \nanalysis indicated that STAI-Trait score significan tly influenced the odds of \nperienteric effusion (OR: 1.124; 95% CI: 1.007-1.25 5; P=0.036). In causal \nmediation analysis, a direct effect of STAI-Trait s core on perienteric effusion \n(P=0.04) was observed; STAI-Trait score and tryptop han had a combined \neffect on perienteric effusion (P=0.06), approachin g statistical significance. \nNegative correlation between tryptophan level and p erienteric effusion (r=-\n0.220, P<0.05) was also found. \nConclusion: Psychological state is associated with MRE-detectab le intestinal \nmorphological changes, and neurotransmitters may se rve as mediators in \nestablishing this connection. \nLimitations: This was a single-center study with small sample si ze. To \nenhance reliability and validity, future investigat ions should consider \nconducting multicenter studies with larger sample s izes. \nFunding for this study: This study was financially supported by National \nNatural Science Foundation of China (82070680, 8227 0693, 82271958, \n82471948, and 82072002). \nEthics committee - additional information: The study was approved by the \ninstitutional ethics review board of our hospital ( No. [2021]215-2). \nAuthor Disclosures:  \nShi-Ting Feng: Nothing to disclose \nZhenpeng Peng: Nothing to disclose \nYaoqi Ke: Nothing to disclose \nQiaoling Zeng: Nothing to disclose \nRuonan Zhang: Nothing to disclose \nXuehua Li: Nothing to disclose \nHuasong Cai: Nothing to disclose \n \n \nMRI neurophenotype reflecting brain-gut interaction s to predict intestinal \ndisease progression in patients with Crohn’s diseas e \n*R. Zhang*, X. Shen, Y. Wang, J. Lin, L. Huang, W. He, S-T. Feng, X. Li; \nGuangzhou/CN \n(zhangrn25@mail2.sysu.edu.cn) \n \nPurpose or Learning Objective: There is considerable recent interest in the \nrole of brain-gut axis in the pathogenesis and mani festations of Crohn’s \ndisease (CD). We developed a multimodal neuroimagin g-based model to \ncharacterize the neurophenotype of CD patients and predict intestinal disease \nprogression, using multi-omics data to demonstrate its validity. \nMethods or Background: This prospective study enrolled 109 CD patients \nwho underwent baseline tests (including multimodal neuroimaging, \npsychological scales, MR enterography, ileocolonosc opy) and fecal/blood \nsamples collection within one week. The neurophenot ype of patients with \ndifferent intestinal inflammation levels was charac terized using a radiomics \nmodel, developed from 13 out of 13,870 neuroimaging  features. This \nneurophenotype in predicting disease progression du ring follow-up was \nevaluated using Kaplan-Meier curves and Cox regress ion analysis. Multi-omics \ndata (including fecal microbiome, fecal/blood metab olomics, intestinal/blood-\nbrain-barrier permeability, and blood neurotransmit ter) were used to elucidate \nhow this neurophenotype reflecting brain-gut intera ctions. \nResults or Findings: The model enabled accurate characterization of \nneurophenotypes in patients with different intestin al inflammation levels in \ntraining and test cohorts (AUC=0.824-0.842, both P< 0.05). Neurophenotype \nwas the most important predictor of disease progres sion (HR=29.05, P=0.033), \nsurpassing psychological traits (HR=0.95-1.09, all P>0.05). Multi-omics \nanalysis revealed that elevated intestinal inflamma tion was correlated with \n\n \n \nThursday \nAbstract-based Programme \n \n 76  \nincreased intestinal permeability and specific gut microbiota (e.g., \nEnterococcus) and metabolites (e.g., caproic acid),  which collectively \ncontributed to high-risk neurophenotype (all P<0.05 ). High-risk neurophenotype \nsubsequently associated with intestinal disease pro gression by establishing \ncorrelations with six blood neurotransmitters (e.g. , tryptophan) (all P<0.05). \nConclusion: The neurophenotype varies among CD patients with di fferent \nintestinal inflammation levels and can predict inte stinal disease progression. \nMulti-omics data offer biological evidence to suppo rt its validity. \nLimitations: This was a single-centre study, and the potential m echanisms \nunderlying the brain-gut axis in our study have yet  to be validated. \nFunding for this study: This study was financially supported by National \nNatural Science Foundation of China (82070680, 8227 0693, 82271958, \n82072002, 82170537, and 82222010), Guangdong Basic and Applied Basic \nResearch Foundation (2023B1515020070 and 2023A15150 11097), 2023 SKY \nImaging Research Fund of the Chinese International Medical Foundation (Z-\n2014-07-2301), and National Key R&D Program of Chin a (2023YFC2507300). \nEthics committee - additional information: The study was approved by the \ninstitutional ethics review board of The First Affi liated Hospital of Sun Yat-sen \nUniversity (No. [2021]215-2) \nAuthor Disclosures:  \nShi-Ting Feng: Nothing to disclose \nLi Huang: Nothing to disclose \nWeitao He: Nothing to disclose \nRuonan Zhang: Nothing to disclose \nXuehua Li: Nothing to disclose \nYangdi Wang: Nothing to disclose \nXiaodi Shen: Nothing to disclose \nJinjiang Lin: Nothing to disclose \n \n \nFirst Findings from the BIPOCUS Study: Progress and  Educational \nImpact on Ultrasound Training in the Practical Year  \n*E. Höhne*¹, V. Schäfer², S. Petzinna², A. Wittek²,  J. Gotta¹, P. Reschke¹,  \nF. Recker²; ¹Frankfurt/DE, ²Bonn/DE \n \nPurpose or Learning Objective: Point-of-care ultrasound (POCUS) is \nincreasingly important in clinical settings, leadin g to a growing demand for \ncomprehensive ultrasound training in medical educat ion. This study marks the \nUniversity of Bonn's first attempt to integrate ult rasound courses and handheld \ndevices into the regular curriculum for final-year medical students and assess \ntheir utilization. \nMethods or Background: Forty students in their practical year received a \nhandheld ultrasound device for four months and were  invited to participate in \neight optional ultrasound courses, where they acqui red and rated images using \na developed rating system. At the end of the tertia l, they could complete a \nvoluntary survey on equipment usage. \nResults or Findings: Participation in the optional ultrasound courses wa s \npositive, with the Introduction and FAST module att racting the most \nparticipants (29). Lung images received the highest  average rating (18.82 out \nof possible 23 points, SD ± 4.30), while aorta and vena cava images scored \nlowest (16.62, SD ± 1.55). The overall mean score f or all images was 17.47 \n(SD ± 2.74). Only 21 students responded to the survey, and 67% used the \ndevice independently four times or fewer during the  tertial. \nConclusion: The study aimed to enhance improving students' ultr asound \nskills, but device usage was unexpectedly low, with  most students using it only \nonce a month or less. This raises concerns about re source justification, \nprompting future initiatives to focus on technical improvements, better login \ndata access, and closer monitoring of usage and pro gress to emphasize \npractical ultrasound training in medical education.  \nLimitations: This study's limitations include a small sample siz e, single-\ninstitution focus, incomplete usage data, and low r esponse rates, which hinder \nthe generalizability and reliability of the finding s regarding ultrasound device \nintegration into student routines. \nFunding for this study: None \nEthics committee - additional information: The local ethics committee of the \nUuniversity of Bonn approved the study (253/23-EP) . \nAuthor Disclosures:  \nAgnes Wittek: Nothing to disclose \nValentin Schäfer: Nothing to disclose \nFlorian Recker: Nothing to disclose \nPhilipp Reschke: Nothing to disclose \nJennifer Gotta: Nothing to disclose \nElena Höhne: Nothing to disclose \nSimon Petzinna: Nothing to disclose \n \n \n \n08:00-09:30 Research Stage 2 \nResearch Presentation Session: Cardiac \nRPS 703 \nNew techniques and clinical applications \nof CMR \n \nModerator \nK.-F. Kreitner; Mainz/DE  \n(Karl-Friedrich.Kreitner@unimedizin-mainz.de) \n \n \nFlip angle mapping and “without-gadolinium” enhance ment: A new \napproach to acute myocardial infarction in magnetic  resonance imaging \n*G. Lucchi*, G. D. Aquaro, A. Marcucci, M. Lombardo , L. Faggioni, R. Lencioni, \nD. Cioni, E. Neri; Pisa/IT \n \nPurpose or Learning Objective: To evaluate Flip Angle Mapping (FAM) as a \nnon-contrast alternative to Late Gadolinium Enhance ment (LGE) for assessing \nischemic core in acute myocardial infarction using balanced Steady-State Free \nPrecession (bSSFP) MRI sequences. \nMethods or Background: This study included 11 patients suspected of \nMyocardial Infarction with Non-Obstructive Coronary  Arteries (MINOCA) and \neight healthy controls. Two bSSFP datasets were acq uired with flip angles of \n15° and 60°, respectively. Signal intensity differe nces were mapped and \ncompared to LGE images. Pathological areas were man ually segmented on \nboth FAM and LGE images. A statistical analysis was  conducted to evaluate \nthe correlation and concordance of the two methods in estimating the \npathological area. \nResults or Findings: The diagnosis of MINOCA was confirmed in five \npatients, while four others were diagnosed with myo carditis and two with \nTakotsubo syndrome. Comparison between FAM and LGE images showed a \ncomplete overlap of pathological areas. Linear regr ession analysis revealed a \nstrong positive correlation between the extent of L GE and FAM abnormalities \n(r=0.99; p<0.001). Bland-Altman analysis confirmed good agreement between \nthe two methods (mean difference: -0.3%; 95% limits  of agreement: -4.3 to \n3.6%). \nConclusion: These findings suggest that FAM could potentially r eplace LGE \nfor acute myocardial damage assessment, dispensing with contrast agents. If \nvalidated in larger studies, FAM could represent a significant advancement in \nnon-invasive cardiac imaging. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is a methodological \nproof of principle. \nAuthor Disclosures:  \nGiacomo Lucchi: Nothing to disclose \nEmanuele Neri: Nothing to disclose \nAlessandro Marcucci: Nothing to disclose \nGiovanni Donato Aquaro: Nothing to disclose \nRiccardo Lencioni: Nothing to disclose \nLorenzo Faggioni: Nothing to disclose \nMarilena Lombardo: Nothing to disclose \nDania Cioni: Nothing to disclose \n \n \nOne-shot black-blood late gadolinium enhancement im aging for rapid, \nmotion-free, and diagnostically accurate scar imagi ng \n*V. De Villedon De Naide*¹, K. Narceau¹, B. Durand¹ , T. Küstner²,  \nM. Villegas-Martinez¹, P. Jais¹, M. Stuber³, H. Coc het¹, A. Bustin¹; \n¹Bordeaux/FR, ²Tübingen/DE, ³Lausanne/CH \n(victor.de-villedon@ihu-liryc.fr) \n \nPurpose or Learning Objective: Multi-shot black-blood LGE imaging is \nincreasingly being used to assess myocardial scars and overcome poor scar-\nblood contrast, often observed with conventional br ight-blood LGE imaging. \nHowever, this method is time-consuming, requires mu ltiple breath-holds, and is \nprone to residual motion artifacts. Here, we introd uce a one-shot black-blood \nLGE sequence combined with image denoising to provi de rapid, motion-free, \nand diagnostically accurate scar imaging. \nMethods or Background: The 2D black-blood ECG-triggered LGE sequence \nacquires multiple single-shot short-axis images per  slice using a non-selective \n180° inversion pulse, followed by a T1-rho preparat ion. A dummy heartbeat is \nadded between shots, for magnetization-recovery. Si ngle-shot images are \naveraged to enhance quality. The proposed one-shot sequence eliminates \ndummy heartbeats and employs a patch-based low-rank  denoising algorithm \n(PROST) to achieve image quality comparable to mult i-shot techniques. 19 \n\n \n \nThursday \nAbstract-based Programme \n \n 77  \npatients with ischemic heart disease underwent 1.5T  CMR (Siemens Area) \nusing reference PSIR and five-shot black-blood LGE imaging 12min after \ngadolinium injection. One-shot images were retrospe ctively selected from \nmulti-shot datasets and were PROST-denoised. A blin ded radiologist graded \ndiagnostic confidence, documented eventual residual  motion artefact and \nextracted scar volume and signal intensities (blood , scar, remote myocardium) \nusing Circle CVI42 for the three datasets. \nResults or Findings: Acquisition times were in average 4min shorter for black-\nblood one-shot PROST compared to reference sequence s. No statistically \nsignificant differences were observed between black -blood multi-shot and one-\nshot PROST in signal intensities or in scar detecti on, while scar volume \nagreement was excellent. Diagnostic confidence was rated good or excellent in \n95% of black-blood multi-shot and 89% of one-shot P ROST scans. No residual \nmotion artefacts were found in black-blood one-shot  PROST datasets. \nConclusion: Black-blood one-shot PROST provides rapid, motion-f ree, and \ndiagnostically accurate scar imaging, offering a mo re efficient and patient-\nfriendly solution. \nLimitations: Prospecting testing is now warranted. \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nt ANR-22-CPJ2-\n0009-01, and from the European Research Council (ER C) grant \"SMHEART\" \nunder the European Union’s Horizon 2020 research an d innovation programme \n(grant agreement No101076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \nAuthor Disclosures:  \nVictor De Villedon De Naide: Nothing to disclose \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nKalvin Narceau: Nothing to disclose \nManuel Villegas-Martinez: Nothing to disclose \nPierre Jais: Nothing to disclose \nBaptiste Durand: Nothing to disclose \nMatthias Stuber: Nothing to disclose \nThomas Küstner: Nothing to disclose \n \n \nIncremental Value of Multiparametric Cardiac MRI fo r Non-invasive \nIdentification of Significant Acute Cardiac Allogra ft Rejection: a \nProspective and Biopsy-proven Study \n*P. Zhou*, Z. Dong, S. Zhao; Beijing/CN \n(zhoupengyu996@163.com) \n \nPurpose or Learning Objective: Using endomyocardial biopsy as the \nreference standard, this study aimed to 1) evaluate  the association between \ncardiac MRI (CMR) multiparameters and significant a cute cardiac allograft \nrejection (SR), and 2) assess the incremental value  of CMR multiparameters \nover conventional serum examinations for identifyin g SR in heart \ntransplantation (HTx) recipients. \nMethods or Background: HTx recipients with endomyocardial biopsy and \nhealthy controls were prospectively recruited for C MR assessment. CMR \nfeature tracking (CMR-FT) was performed to evaluate  the left ventricular (LV) \nglobal strain in all three directions. The last ser um examinations including N-\nterminal pro brain natriuretic peptide (NT-proBNP) before anti-rejection therapy \nwere recorded. Participants were divided into 3 gro ups: control, SR (acute \ncellular rejection grade≥2R and/or antibody-mediated rejection [AMR] \ngrade≥pAMR1), and NSR (non-SR). \nResults or Findings: Finally, thirty controls (43.3±13.6 years, 26 male) and 51 \nHTx recipients comprising 23 SRs (48.6±12.6 years, 24 male) and 28 NSRs \n(42.7±14.9 years, 16 male) were enrolled for analysis. Compared with NSRs, \nSRs showed elevated NT-proBNP (7797.0±7527.6pg/ml v s \n3334.6±5935.3pg/ml, p<.001), worse LV global longit udinal strain (GLS) (-\n9.7±3.1% vs -13.1±2.9%, p<.001), and increased native T1 (1384±80.1ms vs \n1321±69.9ms, p<.001) and T2 values (50.9±2.7ms vs 45.7±4.3ms, p<.001). In \nmultivariable analysis, LVGLS (OR=0.76, 95%CI, 0.59  to 0.98, p=.03) and T2 \nvalue (OR=1.35, 95%CI, 1.10 to 1.65, p=.01) were in dependently associated \nwith SR after NT-proBNP adjustment. Furthermore, th e likelihood ratio test \nshowed LVGLS (p=.002) and T2 value (p<.001) had inc remental value over \nNT-proBNP for identifying SR. \nConclusion: LV GLS and T2 value were independently associated w ith SR, \nproviding incremental value for non-invasive identi fication of significant \nrejection in HTx recipients. \nLimitations: Although a relatively small participant sample, thi s is a \nprospective and biopsy-proven study with comprehens ive cardiac \nexaminations, including T1 and T2 mapping of CMR. \nFunding for this study: This study is supported by the National Key R&D \nProgram of China (Nos. 2021YFF0501400 and 2021YFF05 01404) and the Key \nProject of National Natural Science Foundation of C hina (No. 81930044). \n \n \nEthics committee - additional information: The ethics committee from Fuwai \nHospital. \nAuthor Disclosures:  \nPengyu Zhou: Nothing to disclose  \nZhixiang Dong: Nothing to disclose \nShihua Zhao: Nothing to disclose \n \n \nMultiparametric cardiac MRI for the detection of ch imeric antigen \nreceptor T-cell therapy associated myocardial chang es \n*D. Kravchenko*¹, L. Bischoff¹, A. Isaak¹, T. Holde rried¹, T. S. Emrich²,  \nA. Varga-Szemes², N. Mesropyan¹, D. Kütting¹, J. A.  Luetkens¹; ¹Bonn/DE, \n²Charleston, SC/US \n(dmitrij.kravchenko@live.ca) \n \nPurpose or Learning Objective: New chimeric antigen receptor (CAR)-T cell \ntherapy has demonstrated advantages over traditiona l cancer therapies for \ntreatment of highly refractory or relapsing hematol ogical malignancies. \nUnfortunately, there is a paucity of data regarding  cardiotoxic cardiac MRI \n(CMR) findings of therapy associated cytokine relea se syndrome (CRS). \nMethods or Background: Consecutive patients were enrolled for CAR-T cell \ntherapy and received a standard 1.5 T CMR examinati on consisting of \nfunctional cines, parametric mapping, late gadolini um enhancement (LGE), \nand featuring tracking strain, before therapy (base line), during acute cytokine \nrelease syndrome (CRS; as determined by treating on cologist and lab \nparameters), and at 6-month follow-up (mean 184±20 days). Data was \ncompared using RM-ANOVA with Tukey’s posthoc test. \nResults or Findings: 29 patients were available for analysis (mean age 6 0±15 \nyears, 23 males [79%]). The most common malignancy was diffuse large B-cell \nlymphoma (13 [45%]). CRS was observed at a median t ime of one day (IQR 1-\n2 days) after CAR T-cell therapy and reached a medi an degree of 1 (IQR 1-2). \nOne patient passed away due to non-cardiac related CRS. One patient \ndeveloped therapy associated heart failure. No new instances of LGE were \nobserved in any cases. There were no differences fr om baseline to CRS or to \nfollow-up scans for left ventricular ejection fract ion (61±5 vs 60±6 vs 59±7%, \np=0.39), T1 relaxation times (969±18 vs 988±26 vs 972±22 ms, p=0.11), T2 \nrelaxation times (53.3±2.4 vs 54.0±3.0 vs 52.9±1.8 ms, p=0.36), global \nlongitudinal strain (-16.3±2.2 vs -15.6±2.4 vs -14.9±2.8, p=0.18), global \ncircumferential strain (-12.5±2.9 vs -13.0±2.7 vs -12.2±1.8, p=0.51), or global \nradial strain (28.3±12.5 vs 29.2±8.1 vs 30.5±6.0, p=0.67). \nConclusion: CAR T-cell therapy related low degree CRS does not produce \nsignificant myocardial changes on multiparametric C MR from baseline to acute \nCRS or follow-up. \nLimitations: Small study size. \nFunding for this study: None \nEthics committee - additional information: Ethikkommission der \nMedizinischen Fakultät Bonn \nGeb. 74, 4. OG \nVenusberg-Campus 1 \n53127 Bonn \nAuthor Disclosures:  \nJulian Alexander Luetkens: Nothing to disclose \nNarine Mesropyan: Nothing to disclose \nAlexander Isaak: Nothing to disclose \nLeon Bischoff: Nothing to disclose \nDmitrij Kravchenko: Speaker: Philips \nDaniel Kütting: Nothing to disclose \nTilman Stephan Emrich: Grant Recipient: Siemens \nTobias Holderried: Nothing to disclose \nAkos Varga-Szemes: Grant Recipient: Siemens Consult ant: Elucid \n \n \nCardiovascular magnetic resonance–derived upper ven tricular septal \nscar can predict the prognosis of left bundle branc h area pacing \n*Y. Fan*, X. Zhu; Nan Jing/CN \n(13218017502@163.com) \n \nPurpose or Learning Objective: As a novel technique, left bundle branch \npacing (LBBP) can achieve excellent resynchronizati on in patients with left \nbundle branch block (LBBB). The study is to use car diovascular magnetic \nresonance (CMR) to evaluate myocardial scars of dif ferent segments to predict \nthe prognosis of patients with LBBB. \nMethods or Background: Consecutive patients with LBBB, left ventricular \nejection(LVEF)≤35% and who underwent CMR examination and successfu l \nLBBP were retrospectively enrolled. The myocardial scar of different segments \nis assessed by CMR. LVEF response was defined as a 15% increase in LVEF \nassessed by echocardiography at 6 months. \nResults or Findings: Among 68 patients who were included, and 52 patient s \npatients showed a favorable LVEF response . The res ponders had lower \nglobal, septal scar burden by CMR (P<0.001). The sc ar burden of AHA 8 is \nindependently associated with the prognosis of LBBP  (AUC: 0.877 [95% CI:  \n\n \n \nThursday \nAbstract-based Programme \n \n 78  \n0.782, 0.971]) and the linear equation was that ΔLVEF= -0.4161(scar burden) \n+ 23.229 (r=-0.60,P＜0.001), indicating that each 1% increased in scar b urden, \nLVEF decreased by 0.4161%. Moreover, the patterns o f scar of AHA8 and the \nscar morphology are independent of the improvement of LVEF (P＞0.05). \nConclusion: The scar burden of AHA 8, as a common implantation area, can \npredict LVEF improvement. And we recommend to pay m ore attention to the \nextent of mycardial LGE rather than the patterns an d morphology. \nLimitations: The limitations are that firstly, it is a retrospec tive single-center \ninvestigation and it may lead to an inclusion bias in that some patients who did \nnot undergo CMR before surgery were not included in  the examination. \nSecondly, the study only investigated patients with  low LVEF, and further \nconfirmation is needed on the relationship between the prognosis of patients \nwith high LVEF. \nFunding for this study: Project supported by the Young Scientists Fund of t he \nNational Natural Science Foundation of China (No.82 302163 ) \nEthics committee - additional information: Given that the study was \nretrospective, the committee has been waived. \nAuthor Disclosures:  \nXiaomei Zhu: Nothing to disclose \nYin Fan: Nothing to disclose \n \n \nWideband myocardial T2 mapping with implantable car diac device:  \nA preliminary evaluation in healthy volunteers at 1 .5 T \n*P. Gut*¹, D. Kim², H. Cochet³, F. Sacher³, P. Jais ³, M. Stuber¹, A. Bustin⁴; \n¹Lausanne/CH, ²Northwestern/US, ³Pessac/FR, ⁴Bordeaux/FR \n(paulinerose.gut@gmail.com) \n \nPurpose or Learning Objective: Myocardial T2 mapping allows assessment \nof myocardial inflammation and edema, but is impact ed by artefacts related to \nimplantable cardioverter defibrillators (ICDs), lea ding to image artifacts and \ninacurate T2 values. This study aimed to integrate a wideband T2 preparation \ninto a T2 mapping sequence and evaluate its perform ance against \nconventional T2 mapping in healthy subjects with an d without ICDs. \nMethods or Background: Three short-axis slices covering the heart at the \nbasal, mid-ventricular, and apical levels were acqu ired in eight healthy \nvolunteers (2 females, age: 26±6y) at 1.5T (MAGNETOM Aera, Siemens) \nduring end-expiration in mid-diastole using both co nventional and wideband T2 \nmapping, with and without an ICD taped below the le ft clavicle (~10 cm from \nthe heart). The T2 preparation module (duration=0, 27, 55ms) included two \nadiabatic hyperbolic secant refocusing pulses of 1. 6 kHz (conventional) and \n5.0 kHz (wideband). Common parameters included: res olution=1.4mmx1.4mm, \nslice thickness=8mm, FA=15°, GRAPPA x2, partial Fou rier phase 6/8, \nTE/TR=2.09/3.95ms, readout bandwidth=1221Hz/pixel, FOV=360mmx287mm, \nand gradient recalled-echo (GRE) readout. T2 maps w ere reconstructed using \na 2-parameter (M0 and T2) fitting model. Myocardial  T2 values were manually \nextracted in 16 heart segments. Statistical analyse s were performed using \nrepeated measures ANOVA and Bonferroni correction. \nResults or Findings: Without ICD, T2 values were not significantly diffe rent \nbetween conventional (mean: mean: 43.5, SD: 2.21) a nd wideband sequences \n(mean: 44.0, SD: 2.15) (P=0.111). With ICD, convent ional T2 values \nsignificantly decreased (mean: 36.4, SD: 5.91) (P<0 .01), especially in apical \nanterior, apical inferior, mid-ventricular anterior , and basal anterior segments. \nWideband T2 values remained unchanged (mean: 42.7, SD: 1.91) (P=0.377). \nConclusion: Wideband T2 mapping effectively reduces ICD-related  artifacts, \nproviding more accurate myocardial T2 values than c onventional T2 mapping. \nLimitations: The study was conducted solely on healthy individua ls. Validation \nin clinical populations is warrented. \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nts Equipex MUSIC \nANR-11-EQPX-0030, ANR-22-CPJ2-0009-01, ANR-21-CE17- 0034-01, and \nProgramme d’Investissements d’Avenir ANR-10-IAHU04- LIRYC, and from the \nEuropean Research Council (ERC) under the European Union's Horizon 2020 \nresearch and innovation program (grant agreement 10 1076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \nAuthor Disclosures:  \nPauline Gut: Nothing to disclose \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nDaniel Kim: Nothing to disclose \nPierre Jais: Nothing to disclose \nFrederic Sacher: Nothing to disclose \nMatthias Stuber: Nothing to disclose \n \n \n \n \n \nLeft ventricular remodelling index to predict ventr icular tachyarrhythmia \nin nonischemic dilated cardiomyopathy with ejection  fraction <35% \n*X. Jia*, S. Zhao; Beijing/CN \n(jiaxi981014@163.com) \n \nPurpose or Learning Objective: Based on current guidelines, only a few \ndilated cardiomyopathy (DCM) patients with left ven tricular ejection fraction \n(LVEF) <35% receive appropriate implantable cardiov erter-defibrillator therapy, \nleading to increased medical costs and patient comp lications. We explored the \npredictive value of LV remodeling index (LVRI) for ventricular tachyarrhythmia \n(VTA) in nonischemic DCM with LVEF <35%. \nMethods or Background: In this retrospective single-center study, \nconsecutive nonischemic DCM patients with LVEF <35%  (n=271) who \nunderwent cardiac magnetic resonance (CMR) imaging were followed up for \nVTA events, including sustained ventricular tachyca rdia, ventricular \nfibrillation/flutter, sudden cardiac death (SCD), a nd aborted SCD. The newly \nderived LVRI was defined as the cubic root of the L V end-diastolic volume \ndivided by the maximal LV wall thickness. Competing  risk regression analysis \nand Kaplan-Meier analysis were used to evaluate the  association of LVRI with \nVTA. \nResults or Findings: During a median follow-up of 71 months (interquarti le \nrange: 17–134 months), 35 (12.9%, mean age 46.7 yea rs, 27 males) \nparticipants reached VTA events. The presence of la te gadolinium \nenhancement (LGE) (62.9% vs. 60.2%, p=0.761) and LV EF (23.3±6 vs. \n21.9±10.3, p=0. 197) were not significantly differe nt between the patients with \nand without VTA events. Kaplan-Meier curve analysis  showed that participants \nwith LVRI ≥7.5 were more likely to experience VTA (p<0.0001). In the multiple \ncompeting risk analysis, when heart transplantation  and heart failure-related \ndeath were counted as competing risks, LV mass inde x (hazard ratio [HR], \n0.983; 95% confidence interval [CI]: 0.968-0.999; p =0.033) and LVRI ≥7.5 (HR, \n2.496; 95% CI: 1.213-5.138; p=0.013) were observed as the independent \npredictors of VTA after adjusting for age, sex and left bundle branch block. \nConclusion: In the cohort of patients with nonischemic DCM with  LVEF <35%, \nCMR-assessed LVRI ≥7.5 was an independent predictor of VTA events. \nLimitations: Not applicable. \nFunding for this study: Funding was received from the National Key R&D \nProgram of China (No. 2021YFF0501400, 2021YFF050140 4); Key Project of \nNational Natural Science Foundation of China (No. 8 1930044). \nEthics committee - additional information: The study received institutional \nreview board approval and written informed consent was obtained from all \nparticipants. \nAuthor Disclosures:  \nShihua Zhao: Nothing to disclose \nXi Jia: Nothing to disclose \n \n \nThe Effect of Obesity on Cardiac Structure and Func tion: A Magnetic \nResonance Study of the Hamburg City Health Cohort \n*J. Erley*, D. G. Aydemir, K. Muellerleile, E. Cavu s, G. Adam, M. Meyer,  \nE. Tahir; Hamburg/DE \n \nPurpose or Learning Objective: To analyze the effect of waist-to-hip ratio \n(WHR) and body mass index (BMI) on cardiac structur e and function using \nmagnetic resonance imaging (CMR). \nMethods or Background: The Hamburg City Health Study (HCHS) is a \npopulation-based cohort study. Individuals between 45-74 years of age \nunderwent 3T CMR. Subjects with cardiac diseases (e .g., coronary artery \ndisease, myocardial infarction), and previous cardi ac interventions were \nexcluded. Linear regression models were conducted, adjusted for age and sex. \nResults or Findings: 1671 subjects (41% female, mean age 64±8 years) wer e \nanalyzed. Median WHR was 0.95 [interquartile range:  0.88; 1.01] and median \nBMI was 26.2kg/m² [23.8; 29.2]. Concerning BMI cut- off values, 44% of \nsubjects were overweight (BMI 25-29.9kg/m²) and 20%  obese (BMI ≥ 30kg/m²). \nAccording to the WHR, 81% of subjects were obese (W HR ≥0.85 in females \nand 0.90 in males). An increase in WHR was associat ed with a 5.2% [0.1-10.2] \nhigher left ventricular (LV) ejection fraction (p=0 .044), and a 43.6g [27.6; 59.5] \nhigher LV end-diastolic mass (EDM) (p<0.001), but l ower LV and right \nventricular (RV) end-diastolic volumes (EDV) (LV: - 37.3ml [-57.4; -17.3], \np<0.001; RV: -34.0ml [-56.0; -12.1], p=0.002) and e nd-systolic volumes (LV: -\n18.7ml [-28.5; -8.8], p<0.001; RV: -16.0ml [-28.8;- 3.1], p=0.015), leading to \nlower stroke volumes (SV) (LV: -18.5ml [-32.2;-4.7] , p=0.008; RV: -18.8ml [-\n33.6;-4.1], p=0.013). An increase in BMI was associ ated with a 1.9g [0.2; 2.2] \nhigher LVEDM (p<0.001), higher EDV (LV: 0.5ml [0.2;  0.9], p=0.002; RV: 0.4ml \n[0.1; 0.7], p=0.047) and a 0.4ml [0.2; 0.7] higher LVSV (p<0.001). \nConclusion: An increase in WHR is associated with a higher left  ventricular \nmass and lower volumes as a sign of concentric remo deling, while an increase \nin BMI is associated with ventricular dilatation. \nLimitations: Analyses are not adjusted for other cardiovascular risk factors. \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 79  \nFunding for this study: The HCHS is supported by the Innovative medicine \ninitiative (IMI) under Grant No. 116074, by the Fon dation Leducq under Grant \nNumber 16 CVD 03, by the euCanSHare Grant Agreement  No. 825903-\neuCanSHare H2020 and the DFG under project Grant TH 1106/5-1; AA93/2-1. \nThe DIFE provides the licence for the Food Frequenc y and Physical activity. \nTechnical equipment is provided by SIEMENS accordin g to a contract for 12 \nyears as well as by the Schiller AG on a loan basis  for 6 years and by Topcon \non a loan basis from 2017 until 2022. The Hamburg C ity Health Study is \nadditionally supported by an unrestricted Grant (20 17–2022) by Bayer. Project-\nrelated analyses are supported by Amgen, Astra Zene ca, BASF, Deutsche \nGesetzliche Unfallversicherung (DGUV), DKFZ, DZHK, Novartis, Seefried \nStiftung and Unilever. The study is further support ed by donations from the \n“Förderverein zur Förderung der HCHS e.V.”, TEPE (2 014) and Boston \nScientific (2016). A current list of the supporters  is online available on \nwww.uke.de/hchs. \nEthics committee - additional information: The study was approved by the \nlocal ethics committee of the medical association i n Hamburg. \nAuthor Disclosures:  \nGerhard Adam: Nothing to disclose \nErsin Cavus: Nothing to disclose \nMathias Meyer: Nothing to disclose \nJennifer Erley: Nothing to disclose \nKai Muellerleile: Nothing to disclose \nDestina Gizem Aydemir: Nothing to disclose \nEnver Tahir: Nothing to disclose \n \n \nLipomatous hypertrophy of the atrial septum in card iac magnetic \nresonance \n*A. Fortunati*, D. Fazzini, S. Papa, M. Alì, F. Dar vizeh, F. Secchi; Milan/IT \n(alice.fortunati@unimi.it) \n \nPurpose or Learning Objective: Lipomatous hypertrophy of the atrial septum \nis a benign anomaly, consisting in a fatty infiltra tion of the interatrial septum \nwith sparing of the fossa ovalis and typically asso ciated with elderly. This \nretrospective study aim to define the prevalence of  LHAS through cardiac \nmagnetic resonance and its correlation with age or functional biventricular \nparameters. \nMethods or Background: A retrospective analysis of 621 patients who \nunderwent CMR from March 2020 to March 2022 was per formed, with the \nfollowing inclusion criteria: the presence of a 4-c hamber sequence and the \npresence of volume analysis. All images were review ed by a reader to evaluate \nthe presence of LHAS. The atrial septum thickness w as measured, and the \nfunctional biventricular parameters retrieved from the clinical report. The \nstatistical analysis was conducted using the Spearm an’s correlation test. \nResults or Findings: Among the 619 patients included in the study, 241 \npatients were found with LHAS and 150 showed lipoma tous deposition of the \nupper half of the atrial septum with a mean thickne ss of the atrial septum of 3 \nmm. A significant negative correlation was found be tween the degree of LHAS \nand left ventricle end-diastolic volume (r = -0.21,  p<0.001) and systolic volume \n(r = -0.20, p<0.001). A significant negative correl ation was also recognized \nbetween the degree of LHAS and right ventricle end- diastolic volume (r = -0.25, \np<0.001) and systolic volume (r = -0.18, p<0.001). \nConclusion: LHAS reached a 39% prevalence. Findings confirm a s ignificant \nnegative correlation between LHAS and biventricular  end-diastolic volumes \naccording to a consecutive reduction of atrial and ventricular volumes to offset \nthe volumetric increase of atrial septum and a prog ressive increase of LHAS \ndisease prevalence with age. \nLimitations: The study has some limitations due to its monocentr ic nature and \nthe relatively small sample size. \nFunding for this study: None \nEthics committee - additional information: The local Ethics Committee \napproved this retrospective study. \nAuthor Disclosures:  \nFrancesco Secchi: Nothing to disclose  \nSergio Papa: Nothing to disclose \nDeborah Fazzini: Nothing to disclose \nMarco Alì: Nothing to disclose \nAlice Fortunati: Nothing to disclose  \nFatemeh Darvizeh: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \nRisk Stratification of Sudden Cardiac Death in Non- Ischemic \nCardiomyopathy: Towards Arrhythmogenic Substrate As sessment in \nCardiac MRI \nD. Zhou, M. Lu, *Y. Wang*; Beijing/CN \n \nPurpose or Learning Objective: Magnetic resonance imaging (MRI)-derived \narrhythmogenic substrate is indicative of sudden ca rdiac death (SCD) in \npatients with non-ischemic cardiomyopathy (NICM). A  key issue that needs to \nbe addressed is what extent of T1 mapping metric co ntributes to the prognosis \nfor SCD over late gadolinium enhancement (LGE). \nMethods or Background: A total of 837 NICM patients who underwent T1 \nmapping MRI were consecutively enrolled in this stu dy. The primary endpoint \nis a composite of SCD-related events, including SCD , appropriate implantable \ncardioverter-defibrillator shock and resuscitated c ardiac arrest. \nResults or Findings: Over a median follow-up of 58.3 months, 78 patients  \nreached the primary endpoint, and 198 patients reac hed the secondary \nendpoint. In the adjusted analysis, LGE ≥ 7.2%(HR: 4.748, p < 0.001), \nextracellular volume (ECV) fraction ≥ 31.8% (HR: 2.913, p = 0.001), and native \nT1 z-score ≥ 2.1 (HR: 1.686, p = 0.035) were associated with SC D-related \nevents. Patients with LGE (-) and ECV ≥ 31.8% were at higher risk of \nexperiencing SCD events compared to those with ECV < 31.8% and LGE \nbetween 0-7.2% or mid-wall/focal LGE. Patients stra tified by LGE ≥ 7.2% \nexhibited a high risk of experiencing SCD-related e vents with an annual event \nrate of 4.65%, regardless of ECV. Patients with LVE F > 35%, LGE < 7.2%, and \nECV < 31.8 exhibited an actual low risk of SCD with  an annual event rate of \n0.2%. \nConclusion: LGE ≥ 7.2% was strongly associated with high SCD risk, s uperior \nto LGE distribution and pattern. ECV serves a cruci al role in differentiating \npatients at low to moderate risk, particularly thos e with negative LGE or \nfocal/mid-wall LGE. \nLimitations: This is a retrospective study. \nFunding for this study: High-level research projects of the National Health  \nCommission (2022-GSP-QZ-5) \nEthics committee - additional information: Fuwai Hospital \nAuthor Disclosures:  \nMinjie Lu: Nothing to disclose \nYining Wang: Nothing to disclose \nDi Zhou: Nothing to disclose \n \n \nOne-click joint bright- and black-blood late gadoli nium enhancement and \nT2 mapping for advanced myocardial imaging in the a cute STEMI \npopulation \n*V. De Villedon De Naide*¹, E. Gerbaud¹, B. Durand¹ , M. Villegas-Martinez¹,  \nA. I. Schmid², P. Jais¹, M. Stuber³, H. Cochet¹, A.  Bustin¹; ¹Bordeaux/FR, \n²Vienn/AT, ³Lausanne/CH \n(victor.de-villedon@ihu-liryc.fr) \n \nPurpose or Learning Objective: CMR imaging enables post-infarction risk-\nstratification by identifying prognostic markers, s uch as infarct size (IS), the \npresence of microvascular obstruction (MVO), ejecti on fraction, area-at-risk \n(AAR) and myocardial salvage (MS). However, collect ing these markers \nrequires the use of several MRI sequences. Here, we  propose a unified 'one-\nclick' sequence jointly collecting scar, MVO, and M S information for seamless \nplanning, fast acquisition, and enhanced image quan tification and analysis in \nthe acute STEMI population. \nMethods or Background: The proposed 2D whole-heart SPOT-MAPPING \nacquisition is a single-shot breath-held sequence g athering black- and bright-\nblood LGE images, averaged for optimal measurement of IS and MVO. For the \nbright-blood shots, a T2 preparation module with in creasing duration is used to \ngenerate a T2 map for myocardial tissue quantificat ion (MS and AAR). Seven \npatients with acute STEMI underwent CMR (1.5T Sieme ns). Pre-contrast T2 \nmaps and post-contrast PSIR, SPOT and SPOT-MAPPING images were \ncollected in a random order 12 after injection of g adolinium. Left ventricular \nwall and scar contours were drawn by a radiologist using Circle CVI42. \nPrognostic markers were extracted according to lite rature, along with T2 values \nin remote and injured myocardium. Acquisition times  were recorded. \nResults or Findings: Acquisition times for PSIR, SPOT, T2 mapping and \nproposed SPOT-MAPPING were 10, 10, 13 and 10 heartb eats per slice, \nrespectively. No significant differences were found  between SPOT-MAPPING \nand PSIR for IS and between SPOT-MAPPING and T2 map ping for AAR, and \nT2 values. By imaging both IS and AAR in a co-regis tered fashion, SPOT-\nMAPPING enabled the measurement of the MS and the M VO. \nConclusion: SPOT-MAPPING enables easy planning, fast acquisitio n and \nenhanced image quantification and analysis for pati ents with acute STEMI. \nLimitations: Further validation in a larger cohort is warranted,  as SPOT-\nMAPPING clinical application is still in its early stages. \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 80  \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nt ANR-22-CPJ2-\n0009-01, and from the European Research Council (ER C) grant \"SMHEART\" \nunder the European Union’s Horizon 2020 research an d innovation programme \n(grant agreement No101076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \nAuthor Disclosures:  \nVictor De Villedon De Naide: Nothing to disclose \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nAlbrecht Ingo Schmid: Nothing to disclose \nManuel Villegas-Martinez: Nothing to disclose \nPierre Jais: Nothing to disclose \nBaptiste Durand: Nothing to disclose \nEdouard Gerbaud: Nothing to disclose \nMatthias Stuber: Nothing to disclose \n \n \n08:00-09:30 Research Stage 3 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 705 \nArtificial intelligence in breast imaging \n \nModerator \nJ. Teuwen; Nijmegen/NL  \n(j.teuwen@nki.nl) \n \n \nOptimal utilization of an AI diagnostic software in  a mammography \nscreening program in Switzerland \n*M. Blum*, A. Geissler, D. Ehlig, J. Vogel, J. Sube lack, R. Morant; \nSt.Gallen/CH \n(marcel.blum@unisg.ch) \n \nPurpose or Learning Objective: The goal of this study is to evaluate \nProfound AI® (pAI) in the screening process of the organized mammography \nscreening program (MSP) “donna”. We aim to identify  the optimal utilization of \npAI in the MSP regarding its effectiveness (sensiti vity and specificity) and its \ninfluence on required resources. \nMethods or Background: In this retrospective study, we analyse all \nmammographies from one screening round, i.e., the y ears of 2022 and 2023, \nof the MSP “donna” in the Swiss canton of St.Gallen  (approximately 27,600 \nmammographies) using pAI by iCAD, which will assign  each mammography a \ncase and predictive risk score. We use optimization  models, such as the \nreceiver operating characteristics curve, to find t he optimal threshold for case \ndiscussion in a consensus conference. We simulate m ultiple AI implementation \nscenarios within the MSP, including AI as a substit ute for one of the two \nradiologists and AI as a preselection tool to ident ify mammographies for double \nreading. \nResults or Findings: First results of this study are expected in early 2 025 with \nanticipation to determine an optimal threshold when  a mammography should \nbe further discussed in a consensus conference. Thi s threshold is expected to \nincrease the effectiveness by increasing the breast  cancer detection rate. In \nthe simulated scenarios, we expect that the workloa d of radiologists can be \nreduced significantly, thus increasing the efficien cy of the MSP, without loss of \neffectiveness. \nConclusion: Our study will contribute to identifying the optima l implementation \nof AI in the screening process of an MSP, optimize its effectiveness, i.e., \nincreasing the cancer detection rate, and its effic iency, as well as initiate a \ndiscussion about the future of organized screening.  \nLimitations: This study’s limitation lies in its retrospective d esign and the initial \nomission of interval carcinomas. \nFunding for this study: This study is partly funded by the Cancer League of  \nEastern Switzerland. \nEthics committee - additional information: This study has been approved by \nthe Ethics Committee of Eastern Switzerland (EKOS) under the project ID \n2024-01310. \n \n \n \n \n \nAuthor Disclosures:  \nJonas Subelack: Nothing to disclose \nAlexander Geissler: Nothing to disclose \nDavid Ehlig: Nothing to disclose \nMarcel Blum: Nothing to disclose \nJustus Vogel: Nothing to disclose \nRudolf Morant: Nothing to disclose \n \n \nArtificial intelligence mammography interpretation systems are affected \nmore by mammographic image quality issues than radi ologists are \n*S. D. Verboom*, J. M. D. S. Boita, M. Broeders, I.  Sechopoulos; Nijmegen/NL \n(sarah.verboom@radboudumc.nl) \n \nPurpose or Learning Objective: To determine how common image quality \nissues in mammograms affect the performance of arti ficial intelligence (AI)-\nbased mammography interpretation systems compared t o expert breast \nradiologists. \nMethods or Background: Five common image quality issues were simulated \non 80 digital screening mammograms (40:20:20, cance r:benign:normal). Each \nissue was simulated at two levels: the lowest quali ty that was acceptable to \nradiologists, and a realistic quality that was not acceptable. Thirteen expert \nbreast radiologists from five countries and two com mercial AI systems \nassessed all mammograms and scored the mammograms w ith a probability of \nmalignancy (PoM) and a recall decision. The AI reca ll decision was obtained \nby matching the specificity on standard quality ima ges to that of the \nradiologists. The area under the receiver operating  characteristics curve (AUC) \nand recall decisions of radiologists and AI for the  two lower quality levels were \ncompared to those for the standard quality images. \nResults or Findings: The radiologists’ original mean AUC of 0.76 (95%CI \n0.68-0.84) was not affected by the lower image qual ity (p=0.77, 0.46). The \nAUCs of AI system A were 0.72 (0.60-0.83) on the or iginal quality, 0.68 (0.55-\n0.80) (p=0.47) for the lower-acceptable quality, an d 0.61 (0.49-0.74) (p=0.06) \nfor the unacceptable quality. For system B, the AUC  decreased from 0.95 \n(0.90-1.0) to 0.91 (0.84-0.96) (p=0.25) and to 0.87  (0.78-0.95) (p=0.02), \nrespectively. Radiologists gave the same recall dec ision in 83% and 82% of \nthe cases for each quality level. Meanwhile, system  A gave the same recall \ndecision in 75% (p=0.06) and 68% (p=0.001) of the c ases and system B in \n80% (p=0.47) and 78% (p=0.27) of the cases. \nConclusion: Image quality can affect AI performance and recall decision more \nthan radiologists’, even when radiologists’ perform ance is not affected. \nLimitations: Retrospective study with limited sample size. \nFunding for this study: aiREAD financed by the Dutch Research Council \n(NWO), Dutch Cancer Society (KWF), Health Holland ( HH). \nEthics committee - additional information: Approval of the etics committe \nwas not applicable due to the retrospective nature of this study with \nanonymized data that was previously approved for re trospective use. \nAuthor Disclosures:  \nMireille Broeders: Speaker: Hologic and Siemens Hea lthcare Research/Grant \nSupport: Hologic, Screenpoint Medical, Sectra Benel ux, Volpara Healthcare, \nLunit, and iCAD \nSarah Delaja Verboom: Nothing to disclose \nIoannis Sechopoulos: Research/Grant Support: Siemen s Healthcare, Canon \nMedical Systems, ScreenPoint Medical, Sectra Benelu x, Volpara Healthcare, \nLunit Advisory Board: Koning Corp. Speaker: Canon, Siemens Healthcare \nJoana Maria Dos Santos Boita: Employee: After compl eting this project \nemployee at Canon Medical Systems Europe \n \n \nEvaluation of a Digital Breast Tomosynthesis Cancer  Detection AI \nAlgorithm Using the Personal Performance in Mammogr aphic Screening \nScheme (PERFORMS) \nG. Partridge¹, P. Phillips², J. James¹, N. Sharma³,  K. Satchithananda⁴,  \nR. Butler⁵, J. Lewin⁵, M. Michell⁴, *Y. Chen*¹; ¹Nottingham/UK, ²Lancaster/UK, \n³Leeds/UK, ⁴London/UK, ⁵New Haven, CT/US \n(yan.chen@nottingham.ac.uk) \n \nPurpose or Learning Objective: To compare the performance of a Digital \nBreast Tomosynthesis (DBT) Artificial Intelligence (AI) model as a standalone \nreader to that of a large cohort of breast imaging readers, using the Personal \nPerformance in Mammographic Screening (PERFORMS) sc heme. The \nperformance of a subset of readers, assisted by the  DBT AI during image \ninterpretation, will also be reported. \nMethods or Background: 75 challenging combined DBT and Synthetic 2D \nmammography (S2D) screening cases were collated int o a PERFORMS test-\nset. Test-set images were analysed by a prototype s erver allowing batch-\nprocessing of a commercial AI model (Hologic Genius  AI Detection [GAID] \nv2.0). The set was also distributed to 156 readers from 8 UK National Health \nService (NHS) hospitals that use DBT in screening a s part of the PROSPECTS \ntrial, and to 6 readers from 1 US institution that employs DBT in routine \nscreening. The AI performance will be benchmarked a gainst the performance \nof this reader cohort. The US readers will addition ally re-review the test-set \n\n \n \nThursday \nAbstract-based Programme \n \n 81  \nwith AI-markup available for decision support, foll owing a 6-8 week washout \nperiod. Performance with and without AI-support wil l be investigated and \ncompared to the AI as a standalone reader. \nResults or Findings: The AI model achieved an Area Under the Receiver \nOperating Characteristic Curve (AUC) of 0.935, and a sensitivity of 89.5% and \nspecificity of 85.7% at the optimal threshold (=33) . Human readers are \ncurrently undertaking the case review, but their da ta will be reported at the \nconference. \nConclusion: This international, Multiple Reader Multiple Case ( MRMC) study \nenables the comparison of a very large cohort of br east imaging readers to a \nDBT AI model, as well as investigating the affect o f reading DBT with AI-\nsupport. \nLimitations: The test-set is enriched with malignant cases which  may \ninfluence human reader decisions. \nFunding for this study: Funding was acquired from Hologic Inc. \nEthics committee - additional information: This study is classed as a clinical \naudit for quality assurance for improvement of the breast screening \nprogramme. Ethics Reference No: 88-1223. \nAuthor Disclosures:  \nNisha Sharma: Nothing to disclose \nReni Butler: Nothing to disclose \nKeshthra Satchithananda: Nothing to disclose \nMichael Michell: Nothing to disclose  \nGeorge Partridge: Nothing to disclose  \nPeter Phillips: Nothing to disclose \nJohn Lewin: Nothing to disclose \nJonathan James: Nothing to disclose \nYan Chen: Nothing to disclose \n \n \nEvaluation of an AI System for Cancer Detection in Abbreviated Breast \nMRI \n*K. Eppenhof*¹, A. Rodriguez Ruiz¹, W. B. Veldhuis² , C. Van Gils²,  \nA. M. Rosanò³, R. Yang⁴, D. E. Lehrer⁵, L. Çelik⁶, R. Mann¹; ¹Nijmegen/NL, \n²Utrecht/NL, ³Sion/CH, ⁴East Brunswick, NJ/US, ⁵Buenos Aires/AR, \n⁶Istanbul/TR \n(koen.eppenhof@screenpointmed.com) \n \nPurpose or Learning Objective: To investigate the performance of an AI \nsystem for breast cancer detection in abbreviated D CE-MRI. \nMethods or Background: A combination of high-risk screening and diagnostic  \nDCE-MRI exams from five hospital groups and a publi c data set (Duke-Breast-\nCancer-MRI) were acquired. Each MRI exam was proces sed by an AI system, \nwhich takes as input the pre-contrast and a single post-contrast T1 image \n(abbreviated breast MRI), detects suspicious region s, and outputs a \nmalignancy score per breast between 1 and 10. Addit ionally, the AI system \nwas evaluated on an enriched screening dataset from  the DENSE trial. \nResults or Findings: Area under the Receiver Operating Characteristic cu rve \n(AUROC) was computed for classifying exam malignanc y for exams from four \nhospital groups located in Argentina (41 of 780 exa ms containing biopsy-\nproven cancer, AUROC 0.891 (95% CI=0.828-0.944)), S witzerland (98/3499, \n0.863(0.824-0.896)), Turkey (33/164, 0.955(0.898-0. 998)), and the US \n(153/1096, 0.904(0.877-0.929)). The consistency in AUROCs indicates \nrobustness across populations, protocols, and scann ers. Because Duke-\nBreast-Cancer-MRI exams all contain cancer, a breas t-level analysis was done \nwhere breasts without cancer were used as the negat ive class (904/1808 \nbreasts containing cancer). The AUROC (0.965(0.957- 0.972)) is similar to an \nearlier published AI that used two post contrast im ages. For exams that had a \nBIRADS assessment, the agreement between the AI (sc ore >= 9) and the \nradiologist interpretation (BIRADS 1 or 2 vs. 4 or 5) was found to be moderate \n(Cohen kappa=0.502(0.449-0.555)). The performance o n screening-only data \nwas measured in exams from the fifth hospital locat ed in the Netherlands \n(66/2920 exams containing cancer, AUROC 0.812(0.753 , 0.868)), and exams \nfrom the DENSE trial (83/517, AUROC 0.803(0.747-0.8 56)). \nConclusion: A first evaluation of an AI system for abbreviated DCE-MRI \nshows potential for decision support in detecting b reast cancer. \nLimitations: The study has a retrospective design. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nCarla Van Gils: Nothing to disclose \nAlejandro Rodriguez Ruiz: Employee: ScreenPoint Med ical \nKoen Eppenhof: Employee: ScreenPoint Medical \nDaniel Eduardo Lehrer: Nothing to disclose \nWouter B. Veldhuis: Nothing to disclose \nLevent Çelik: Nothing to disclose \nRoger Yang: Nothing to disclose \nAnna Maria Rosanò: Nothing to disclose \nRitse Mann: Advisory Board: ScreenPoint Medical \n \n \nValidating the impact of real-world live use of AI as an additional reader \nin breast cancer screening (BCS) \n*A. Ng*¹, E. Ambrozay², E. Szabó², B. Glocker¹, P. Kecskemethy¹; ¹London/UK, \n²Budapest/HU \n(Annie.ng@deephealth.com) \n \nPurpose or Learning Objective: To validate that the measured impact of \ndeploying AI as an additional reader is a result of  the AI intervention and not \npurely the result of additional reading. \nMethods or Background: Live-use of an AI-system as an additional reader \n(XR) to flag cases for additional review that it su ggested to recall but the \nstandard double reading (DR) decision was “no recal l”, has been demonstrated \nto result in a 0.8/1000 increase in cancer detectio n rate (CDR) and a 0.1% \nincrease in positive predictive value when 6.0% of cases are additionally \nreviewed, compared to DR. To validate that the incr eased effectiveness in \nearly cancer detection of the XR AI-workflow is not  purely from additional \nreading, the maximum CDR increase opportunity due t o third-human-reading a \nrandom 6.0% of cases was simulated, calculated as i nterval cancer rate (ICR) \ntimes the portion of cases to be third-human-read ( 6.0%), times the portion of \nhuman-detectable interval cancers (ICs) i.e. visibl e on priors. A range of ICR of \n0.84-2.11/1000 was used (DOI:10.1038/s41523-017-001 4-x). Studies have \nmeasured that 22% of ICs are human-detectable (DOI: 10.1007/s00330-020-\n07130-y), however, a wider range of 22-100% was use d. Third-human-reading \nwas assumed to have an unrealistic 100% sensitivity  among human-detectable \nICs. \nResults or Findings: For the lower and upper end of assumptions, \nrespectively, the maximum CDR increase opportunity calculated for third-\nhuman-reading is 0.01 and 0.13/1000, which is 98.9%  and 87.2% less than the \nXR AI-workflow, suggesting that the increased CDR i mpact of XR is 6-70 times \nmore effective than third-human-reading without AI.  \nConclusion: Simple simulations show that CDR improvements from third-\nhuman-reading a random set of cases would be margin al, validating that the \nsubstantial CDR increase demonstrated by XR is a di rect effect from using AI \nto flag cases for additional review. \nLimitations: Single AI assessed \nFunding for this study: Kheiron Medical Technologies \nEthics committee - additional information: Not required \nAuthor Disclosures:  \nAnnie Ng: Nothing to disclose \nBen Glocker: Employee: Kheiron Medical Technologies  Ltd \nEva Ambrozay: Nothing to disclose \nPeter Kecskemethy: CEO: Kheiron Medical Technologie s Ltd \nEndre Szabó: Nothing to disclose \n \n \nAdding artificial intelligence (AI) case scoring in  a breast screening \nprogramme to optimize reading workflow and workload : a retrospective \nstudy \n*A. Nitrosi*, R. Vacondio, L. Verzellesi, M. Creola , M. Bertolini, P. Giorgi Rossi, \nV. Iotti, P. Pattacini, C. Campari; Reggio Emilia/I T \n(nitrosi.andrea@ausl.re.it) \n \nPurpose or Learning Objective: The objective of this study was to \nretrospectively evaluate a strategy to optimize rea ding workflow and readers’ \nworkload based on the iCAD Case Malignancy Score (C MS). \nMethods or Background: We analyzed 122,216 2D mammography screening \nreading times (RT) corresponding to 61,108 exams in cluding 244 proven \ntumours, consequentially acquired in Reggio Emilia Breast Screening Program \n(BSP) starting from January 2023 to June 2024 and e laborated by iCAD Inc. \nProFound AI 2D system. ICAD Case Malignancy Scores (CMS) represents the \nrelative confidence that a case is malignant on a s cale of 0% to 100%. A pool \nof radiologists performs blinded double reading plu s arbitration framed in work-\nshift. Packs are assigned to a reader respecting a numerical criterion of \nmaximum readings per work-shift. We analyzed the co rrelations (Spearman) \nbetween the RT of individual readers (normalized on  the personal median) with \nthe CMS and the breast density (D). The analysis wa s repeated considering \nonly the women recalled / not recalled / true posit ive (TP). \nResults or Findings: A positive correlation was demonstrated between CMS  \nand RT (R = 0.76) and slightly between D and RT (R = 0.52), overall and in \nrecalled and non-recalled women separately. Using C MS, packs could be \noptimized based on individual reader characteristic s to maximize the number of \nexams for each reader’s pack with constant recall r ate (and TP): first \nsimulations show up to 14% increase in the number o f exams read over 4 \nhours effective reading period. \nConclusion: This scenario would not undermine the reading scree ning \nworkflow while ensuring resource optimization nor i ntroduce any cognitive bias \ninfluencing the readers since each session would ha ve similar expected recall \nand detection rate. \nLimitations: Cases refer only to Reggio Emilia BSP, limiting thi s study. \nFunding for this study: This study was partially supported by the Italian \nMinistry of Health - Ricerca Corrente \n\n \n \nThursday \nAbstract-based Programme \n \n 82  \nEthics committee - additional information: Compliance with Ethical \nStandards Institutional Review Board approval was n ot required because it is a \nClinical Audit about a technical development. This study was conducted in \naccordance with the routine quality assurance proce dures established by the \nLocal Health Authority for its screening programmes . The Reggio Emilia \nCancer Registry, which routinely collects the scree ning history of each case of \nbreast cancer, has been approved by the Provincial Ethics Committee. \nAuthor Disclosures:  \nLaura Verzellesi: Nothing to disclose \nValentina Iotti: Speaker: Invited speaker \nCinzia Campari: Nothing to disclose \nMarco Bertolini: Nothing to disclose \nPierpaolo Pattacini: Speaker: Invited speaker \nAndrea Nitrosi: Speaker: Invited speaker \nRita Vacondio: Speaker: Invited speaker \nPaolo Giorgi Rossi: Nothing to disclose \nMartina Creola: Nothing to disclose \n \n \nAssessment of an AI-system in indicating breast lat erality for screen-\ndetected and interval cancers in breast screening i n a large-scale \nretrospective study \nA. Ng¹, C. Oberije¹, G. Fox¹, R. Currie², A. Redman ³, A. Leaver³, W. Teh⁴,  \nB. Glocker¹, *P. Kecskemethy*¹; ¹London/UK, ²Exeter /UK, ³Gateshead/UK, \n⁴Harrow/UK \n(peter@kheironmed.com) \n \nPurpose or Learning Objective: Assess the utility of an AI-system in \nindicating breast laterality in breast screening. \nMethods or Background: Employing a commercially available AI-system as \nan independent reader, utilising its case-wise reca ll suggestions, within double \nreading has previously been shown to maintain/impro ve screening \nperformance, while providing substantial workload s avings. This has been \ndemonstrated in a large-scale retrospective clinica l study (306,839 cases from \n236,739 participants between 2017-2021), involving three Hologic sites across \nthe UK’s major genetic clusters (South-East/West/No rth), including more \ndiverse ethnicities in London. To further assess th e AI-system’s utility in \nsupporting follow-up investigations for recalls, it s breast laterality \nrecommendation for screen-detected cancers (SDCs) a nd interval cancers \n(ICs) were compared to pathology information. \nResults or Findings: The study included 2592 SDCs and 379 ICs. The AI-\nsystem correctly recalled 2304 SDCs (88.9% sensitiv ity) and 152 ICs (40.1% \nIC detection rate). Among the correctly recalled SD Cs, the AI-system: A) \nindicated pathology-agreeing laterality in 84.5% (8 3.5% unilateral/1.1% \nbilateral), B) recalled unilateral cases as bilater al in 13.9%, C) recalled one \nside in bilateral cancer cases in 0.4%, and D) indi cated the opposite side not \nassessed in 1.2%. The respective results for ICs we re: A) 58.8% (58.1% \nunilateral/0.7% bilateral), B) 18.4%, C) 2.2%, and D) 20.6%. For category D, it \nis unknown if an early abnormality could be present  as the AI-indicated side \nwas not assessed by biopsy nor by additional diagno stic imaging. The AI-\nsystem provides screening utility in scenarios A-C,  which comprises 98.8% for \nSDCs, 79.4% for ICs, 88.1% for ICs diagnosed within  1 year, and 97.1% for \nfalse negative ICs (FNICs). \nConclusion: The AI-system’s laterality detection demonstrated u tility in almost \nall SDCs/FNICs, and the large majority of ICs, show ing it can support the \nclinical workflow with laterality information for f ollow-up assessments. \nLimitations: Single AI assessed \nFunding for this study: NIHR AI in Health and Care Award \nEthics committee - additional information: UK HRA REC reference: \n21/HRA/4830 \nAuthor Disclosures:  \nCary Oberije: Employee: Kheiron Medical Technologie s Ltd \nAnnie Ng: Employee: Kheiron Medical Technologies Lt d \nWilliam Teh: Nothing to disclose \nBen Glocker: Employee: Kheiron Medical Technologies  Ltd \nAlice Leaver: Nothing to disclose \nGeorgia Fox: Employee: Kheiron Medical Technologies  Ltd \nAlan Redman: Nothing to disclose \nPeter Kecskemethy: CEO: Kheiron Medical Technologie s Ltd \nRachael Currie: Nothing to disclose \n \n \nRepurposed AI-Based Mammography Interpretation in D iverse Clinical \nScenarios \n*H. Ngo*¹, J. Neubauer¹, A. L. Palacios Acedo², M. Windfuhr-Blum¹, E. Kotter¹, \nF. Bamberg¹, J. Weiß¹; ¹Freiburg/DE, ²Marseille/FR \n(helen.ngo@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: This study evaluates the diagnostic \nperformance of an artificial intelligence (AI) tool  originally developed for \nscreening mammography, now repurposed for use in va rious clinical scenarios, \nincluding diagnostic mammograms in 1) asymptomatic women, 2) symptomatic \nwomen and 3) patients with a personal history of br east cancer (PHBC). \nMethods or Background: A total of 601 women with were retrospectively \nincluded and categorized into three subgroups: diag nostic mammograms of 1) \nasymptomatic women (n = 423), 2) symptomatic women (palpable abnormality, \nsuspicious sonography, n=66) and 3) patients with P HBC (n =112). The AI-tool \nprovided continuous scores (1 to 100) for potential  malignancy, with \nhistopathological confirmation and/or follow-up ≥2 years as reference standard. \nResults or Findings: The AI-tool showed high performance across all thre e \ncohorts, with areas under the curve (AUC) for diagn ostic mammograms of 1) \nasymptomatic women: 0.75 (95% CI: 0.51-0.98), 2) sy mptomatic women: 0.92 \n(95% CI: 0.81-1.0), and 3) patients with PHBC: 0.71  (95% CI: 0.52-0.90). \nExcluding women with extremely dense breasts (ACR D ) increased the AUC \nfor diagnostic mammograms of 1) asymptomatic women to 0.79 (95% CI: 0.41-\n1.0), 2) symptomatic women: 0.92 (95% CI: 0.81-1.0) , and 3) patients with \nPHBC: 0.73 (95% CI: 0.51-0.95). Using a threshold o f the highest 10% AI-\nscores to binarize the continuous AI-output resulte d in sensitivity 0.92 and \nspecificity 0.50 for subgroup 1); 0.96 and 0.77 for  2) and 0.81 and 0.67 for 3), \nrespectively. \nConclusion: Repurposed AI-tools can enhance malignancy detectio n across \ndiverse patient groups, especially in less dense br easts. Optimizing thresholds \nfor specified populations, such as asymptomatic and  symptomatic cohorts, may \nfurther improve AI's diagnostic effectiveness. \nLimitations: Varying breast densities, particularly extremely de nse breast, can \npose detection challenges, and the sample size of 6 01 may influence the \ngeneralizability of the findings. \nFunding for this study: Unrestricted research grant from Lunit. \nEthics committee - additional information: Approved by local IRB. \nAuthor Disclosures:  \nHelen Ngo: Research/Grant Support: Lunit \nJakob Neubauer: Research/Grant Support: Lunit \nJakob Weiß: Research/Grant Support: Lunit \nAna Luisa Palacios Acedo: Employee: Lunit Europe \nFabian Bamberg: Research/Grant Support: Lunit \nMarisa Windfuhr-Blum: Research/Grant Support: Lunit  \nElmar Kotter: Research/Grant Support: Lunit \n \n \nPatient perceptions towards the use of artificial i ntelligence (AI) in breast \ncancer imaging \nD. Velazquez-Pimentel, S. Khan, T. Falco, S. Hickma n, S. Dani, *T. Suaris*; \nLondon/UK \n(tamarasuaris@hotmail.com) \n \nPurpose or Learning Objective: The aim of this study is to evaluate patient \nperceptions towards the use of artificial intellige nce (AI) in breast cancer \nimaging \nMethods or Background: Women presenting to a single breast cancer unit in \nEast London were invited to participate in a prospe ctive survey. Baseline \nknowledge and attitude towards technology in daily living and attitude towards \nthe use of AI in mammography screening was measured  using a 4-point Likert \nscale. Demographic data including age, ethnicity, e ducation was collected. \nResults or Findings: 944 responses were analysed. Of these, 90% \n(n=853/944) expressed a preference for combined com puter-physician reading \nwith more women expressing confidence in the accura cy of combined \ncomputer-physician (93%, n=882/944) reading over co mputer reading alone \n(54% n=513/944). Self-reported understanding of tec hnology was associated \nwith a higher level of concern. In patients with li mited understanding 46% \nexpressed concern with regards to the accuracy of c omputer read \nmammograms compared to 38% in patients with expert understanding. Level \nof concern was not significantly associated with ag e, ethnicity or education \nlevel (p > 0.05). Regardless of level of concern, t he majority of respondents \nexpressed a positive opinion on the impact computer  read mammograms can \nhave on improving both efficiency (85%, n=798/944) and pick up rate (84%, \nn=797/944). \nConclusion: Despite confidence in the ability of AI to improve efficiency and \npick up rate there is a strong preference expressed  by patients towards \ncombined computer-physician read mammograms. This s tudy demonstrates \nthat this remains true regardless of age, ethnicity  or level of education. Level of \nconcern is associated with self-reported understand ing of technology; targeted \npatient education programs may support implementati on of AI workflow in \nbreast screening programs. \nLimitations: Survey responses are subject to bias. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Patient Survey - local research \nlead confirmed no formal ethics application necessa ry \nAuthor Disclosures:  \nTamara Suaris: Nothing to disclose  \nThamiris Falco: Nothing to disclose \nSarah Khan: Nothing to disclose \nDiana Velazquez-Pimentel: Nothing to disclose \n\n \n \nThursday \nAbstract-based Programme \n \n 83  \nSarah Hickman: Other: -SEH have research collaborat ions with Vara, Screen-\nPoint, Lunit, Google, Volpara, iCAD, Therapixel, Cu reMetrix, Sunnybrook \nResearch Institute, and Massachusetts Institute of Technology. -SEH is a \nRadiology AI: Trainee Editorial Board member. \nShefali Dani: Nothing to disclose \n \n \nADMEDVOICE – The Pathway to Polish Language Automat ic Structured \nReporting in Breast Ultrasound using Voice Recognit ion and Large \nLanguage Models \n*M. Bobowicz*, D. Szplit, A. Dąbkowska, J. Bogdan, K. Gwozdziewicz,  \nJ. Omernik, B. Graff, A. Czyżewski, K. Narkiewicz; Gdansk/PL \n(maciej.bobowicz@gumed.edu.pl) \n \nPurpose or Learning Objective: Breast ultrasound (BUS) equipped with the \nACR BI-RADS lexicon is a well-described diagnostic procedure with mandatory \nfields and a relatively closed vocabulary. This stu dy aims to generate BUS-\nstructured reports automatically using voice recogn ition and topic modelling in \nPolish. \nMethods or Background: A dataset of 6269 BUS radiology reports from the \nUniversity Clinical Center’s Hospital Information S ystem covering 2013-2023 \nwas obtained. The reports were created by more than  ten experienced breast \nradiologists and multiple residents. They covered v arious clinical scenarios, \nincluding diagnosis, treatment, and follow-up tests  in breast cancer, benign \ndisease, mutation carriers, and studies without pat hology. \nResults or Findings: From 6269 reports, 48721 text fragments were obtain ed, \nrepresenting specific parts of the BUS report used as training data. We \nidentified specific ‘topics’ relating to ‘ontologie s’ in these fragments. Topics \nrepresented parts of the radiologist’s report that could be structured into \nsubsections: 1) reference letter information, 2) ti ssue composition, 3) pathology \ndescriptors (masses and calcifications), 4) associa ted features, 5) axillary and \nintramammary lymph node descriptors, 6) other speci al cases, 7) conclusions, \n8) recommendations, 9) final remarks. For automatic  text recognition \nBERTOPIC was explored. As a next step, we invited 2 5 specialist radiologists, \nresidents, medical students and other HCPs to recor d 3328 separate \nsentences for voice recognition algorithms training . \nConclusion: The presented research, which involved topic modell ing, is a first \nstep towards creating Polish language automatic str uctured BUS reporting \nusing voice recognition and LLMs. The resulting dat abase with voice samples \nat three quality levels will be released soon. It w ill allow AI training to reduce \nthe radiology reporting burden with more natural vo ice commands being \ntransferred to structured reports. \nLimitations: The single-centre design, restriction to the Polish  language, and \nlack of external validation. \nFunding for this study: Funding for the ADMEDVOICE Project was provided \nby the Polish National Centre for Research and Deve lopment; Infostrateg IV \naction; grant number: INFOSTRATEG-IV/003/2022. \nEthics committee - additional information: The study was approved by the \nBioethics Committee for Scientific Research of Medi cal University of Gdansk. \nAuthor Disclosures:  \nAndrzej Czyżewski: Nothing to disclose \nBeata Graff: Nothing to disclose \nMaciej Bobowicz: Nothing to disclose \nKrzysztof Narkiewicz: Nothing to disclose \nKatarzyna Gwozdziewicz: Nothing to disclose \nAnna Dąbkowska: Nothing to disclose \nDariusz Szplit: Nothing to disclose \nJulia Bogdan: Nothing to disclose \nJustyna Omernik: Nothing to disclose \n \n \nContrastive Learning in Breast MRI: MLIP as the Bas e Foundation Model \n*N. Rasoolzadeh*¹, T. Zhang², R. Mann¹; ¹Nijmegen/N L, ²Amsterdam/NL \n(nika.rasoolzadeh@gmail.com) \n \nPurpose or Learning Objective: To explore the potential of utilizing a \ncontrastive language image pretraining approach for  3D breast MRI images. \nMethods or Background: A dataset of 15005 pairs of dynamic contrast-\nenhanced (DCE) and subtraction MRI images with corr esponding radiological \nreports from the Netherlands Cancer Institute were used for training a model to \nfind the most similar image-text pairs by contrasti ng positive pairs (similar) \nagainst negative pairs (dissimilar) samples. Full M RI images and complete \nDutch reports were utilized. The image and text emb eddings were obtained \nusing a 3D ResNet50 architecture and RadioLOGIC as the image and text \nencoders, respectively. Two inference scenarios wer e tested: image retrieval \nby text queries and BI-RADS prediction. The area un der the curve (AUC) was \nused to evaluate the model's performance. The devel oped Multimodal Breast \nMRI Language-Image Pretrained (MLIP) model was firs t used for the zero-shot \nBI-RADS prediction task and was later fine-tuned. \n \n \nResults or Findings: The preliminary results show an AUC of 0.717 (95% C I: \n0.604, 0.824) for BI-RADS 4/5 abnormal MRI images r etrieval, 0.640 (95% CI: \n0.538, 0.740) for dense breast retrieval, and 0.601  (95% CI: 0.505, 0.698) for \nlow background parenchymal enhancement (BPE) retrie val. In the second \ninference, the performance of MLIP was compared to that of a fine-tuned \nmodel. The fine-tuned model demonstrated improved a ccuracy, with a \nreduction in the number of originally benign cases misclassified as malignant. \nConclusion: In this study, a multi-modal breast MRI pretrained model was \ndeveloped. The preliminary results suggest MLIP can  be adjusted to perform \ndiagnostic tasks and radiology report generations, holding the potential to \nserve as a foundation model for breast MRI analysis . \nLimitations: The model needs to be validated on larger datasets and across \nmore downstream tasks. \nFunding for this study: Funding was provided by the ODELIA project (from \nthe European Union’s Horizon Europe research and in novation programme \nunder grant agreement, No 101057091) \nEthics committee - additional information: This study did not require formal \nethics committee approval, as it exclusively used f ully anonymized MRI images \nand reports. No identifiable personal data was coll ected or used in the \nanalysis. All MRI data was anonymized prior to acce ss, ensuring that no \nindividual participants can be identified from the data \nAuthor Disclosures:  \nTianyu Zhang: Nothing to disclose  \nNika Rasoolzadeh: Nothing to disclose  \nRitse Mann: Nothing to disclose \n \n \nGenerating virtual T2w-fat-saturated breast MRI acq uisition using neural-\nnetworks \n*A. Liebert*¹, D. Hadler¹, C. M. Ehring¹, H. Schrei ter¹, F. B. Laun¹, M. Uder¹,  \nE. Wenkel², S. Ohlmeyer¹, S. Bickelhaupt¹; ¹Erlange n/DE, ²Munich/DE \n(andrzej.liebert@uk-erlangen.de) \n \nPurpose or Learning Objective: Multi-parametric breast MRI protocols \ntypically include T2-weighted fat-saturated(T2w-FS)  sequences, which are \nused for tissue characterization. However, their ac quisition can significantly \nincrease scan time. This study aims to evaluate, wh ether a 2D-U-Net neural-\nnetwork can generate virtual T2w-FS images(VirtuT2)  from other acquisitions \nof a routine multiparametric breast MRI protocol. \nMethods or Background: This IRB-approved, retrospective study included \nn=914 breast MRI examinations performed between Jan uary 2017 and June \n2020 at University Hospital Erlangen. The dataset w as divided into \ntraining(n=665), validation(n=74), and test(n=175) sets. 2D-U-Net was trained \non T1w, DWI, and DCE sequences to generate VirtuT2 . Quantitative metrics \nand a qualitative multi-reader assessment by two ra diologists were used to \nevaluate the VirtuT2 images. For qualitative readin gs radiologist were asked to \nidentify, whether an image is original T2w-FS or Vi rtuT2 image, evaluate the \ndiagnostic image quality(DIQ) and wheter they can i dentify presence of edema \naround the mass-lesions. \nResults or Findings: VirtuT2 images demonstrated high structural \nsimilarity(SSIM=0.87) and peak signal-to-noise rati o(PSNR=24.90) compared \nto original T2w-FS images. High level of the freque ncy error norm(HFEN=0.87) \nindicates strong blurring presence in the VirtuT2 i mages, which was also \nconfirmed in qualitative reading. Radiologists corr ectly identified VirtuT2 \nimages with 92.3% and 94.2% accuracy, respectively.  No significant difference \nin DIQ was noted for one reader(p=0.21), while the other reported significantly \nlower DIQ for VirtuT2(p<=0.001). Moderate inter-rea der agreement was \nobserved for edema detection on T2w-FS images( ƙ=0.43), decreasing to fair \non VirtuT2 images(ƙ=0.36). \nConclusion: Neural-networks can technically generate VirtuT2 im ages with \nhigh similarity to real T2w-FS images, using T1w, D WI and DCE acquisitions, \nhowever blurring remains a limitation. Future inves tigations with different \narchitectures and using larger datasets are needed to improve clinical \napplicability. \nLimitations: Limited dataset from a single site was used. Qualit ative reading \nwas performed on just n=52 cases. \nFunding for this study: This project is funded by the Bavarian State Minist ry \nof Science and the Arts in the framework of the bid t Graduate Center for \nPostdocs. L.B. is funded by the DFG Grant No: 51868 9644 \nEthics committee - additional information: The study protocol was approved \nby the ethics committee of the Friedrich-Alexander Universität Erlangen-\nNürnberg. The ethics comitee waived the need for in formed consent. \nAuthor Disclosures:  \nSebastian Bickelhaupt: Research/Grant Support: Siem ens Healthineers \nEvelyn Wenkel: Nothing to disclose \nDominique Hadler: Nothing to disclose \nFrederik Bernd Laun: Nothing to disclose \nMichael Uder: Research/Grant Support: Siemens Healt hineers \nHannes Schreiter: Nothing to disclose \nSabine Ohlmeyer: Nothing to disclose \nAndrzej Liebert: Nothing to disclose \nChris Matthias Ehring: Nothing to disclose \n\n \n \nThursday \nAbstract-based Programme \n \n 84  \n08:00-09:30 Research Stage 4 \nResearch Presentation Session: Neuro \nRPS 711 \nStroke and neurovascular imaging: inside \nand outside of the blood vessels \n \nModerator \nE. Papadaki; Iraklion/GR  \n(fpapada@otenet.gr) \n \n \nAdvancing Neurovascular Imaging with Ultra-High-Res olution Photon-\nCounting Detector CT: Optimization of Reconstructio n Kernel and \nQuantum Iterative Reconstruction \n*A. Toth*, Y. (. Cho, E. Wilson, J. Crow, E. Bass, J. Joyce, M. G. Matheus,  \nS. Tipnis, M. V. Spampinato; Charleston, SC/US \n(adrienntoth706@gmail.com) \n \nPurpose or Learning Objective: Our goal was to identify the optimal \ncombination of dedicated neurovascular reconstructi on kernels and quantum \niterative reconstruction (QIR) levels for ultra-hig h-resolution (UHR) photon-\ncounting detector (PCD)-CT angiography of the head and neck. \nMethods or Background: 18 patients with intracranial aneurysms were \nprospectively included in this study. CT angiograms  were obtained in UHR \nmode using a clinical dual-source PCD-CT scanner. I mages were \nreconstructed with a slice thickness of 0.2 mm, uti lizing six strength levels of a \ndedicated neurovascular kernel (Hv48/Hv56/Hv64/Hv72 /Hv80/Hv89) and four \nlevels of QIR (1-4). We assessed image noise, signa l-to-noise ratio (SNR), \ncontrast-to-noise ratio (CNR), and vessel sharpness  for all reconstructions. \nResults or Findings: With higher kernel sharpness and lower QIR, image \nnoise continuously increased. The best performing r econstructions in terms of \nCNR were Hv48 and Hv72 in combination with QIR-4. V essel sharpness \nimproved with higher kernel levels, reaching a plat eau with the Hv64 and Hv72 \nreconstructions, as observed in the small intracran ial arteries (maximum ΔHU \nvalues of 260.59 and 255.11, respectively). Based o n the results of the \nquantitative analysis, the kernels identified as th e top performers and selected \nfor further evaluation in the qualitative analysis were Hv56, Hv64, and Hv72. \nConclusion: Vessel sharpness increased with higher kernels leve ls, reaching \na plateau at Hv64 and Hv72. Overall, Hv56, Hv64 and  Hv72 were recognized \nas the best performing kernels based on the quantit ative results. In the \nfollowing steps, qualitative image quality evaluati on will be conducted by three \nreaders using a 5-point Likert scale. This evaluati on will focus on the best-\nperforming kernels in combination with the availabl e QIR levels to assess \noverall image quality and diagnostic performance. \nLimitations: The limitations of the study are the relatively sma ll patient cohort \nand the fact that quantitative measurements were pe rformed by a single \nobserver. \nFunding for this study: Funding was provided by Siemens Healthineers \n(research grant). \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board of the Medical Universit y of South Carolina \n(Pro00123327)). \nAuthor Disclosures:  \nEric Bass: Nothing to disclose \nYongjoo (Jennifer) Cho: Nothing to disclose \nJohn Crow: Nothing to disclose \nEvan Wilson: Nothing to disclose \nAdrienn Toth: Nothing to disclose \nMaria G. Matheus: Nothing to disclose \nJennifer Joyce: Nothing to disclose \nSameer Tipnis: Nothing to disclose \nM. Vittoria Spampinato: Research/Grant Support: Sie mens Healthineers \n \n \nPerformance of Dual-Layer Spectral Detector Non-con trast Computed \nTomography in Identifying Early Ischemic Changes in  Acute Ischemic \nStroke Patients \n*Y. Wang*, H. Zhu, J. Wen, S. Ma, S. Yang; Beijing/ CN \n(wangyujie619@126.com) \n \nPurpose or Learning Objective: The study aimed to evaluate the \neffectiveness of non-contrast dual-layer spectral c omputed tomography (DLCT) \nfor detecting early ischemic changes in patients wi th acute ischemic stroke \n(AIS). \n \nMethods or Background: NCCT is a common imaging technique for \nsuspected AIS patients. This study involved 27 AIS patients who underwent \nboth DLCT and MRI within 12 hours of symptom onset.  A retrospective \nanalysis was conducted on the imaging data, focusin g on quantitative \nmeasurements from regions identified as acute infar ction on diffusion-weighted \nMRI. Various parameters, including conventional CT values, virtual \nmonoenergetic (monoE) CT values, and electron densi ty relative to water \n(EDW) were compared between ischemic and normal bra in. Statistical \nanalyses, including the Mann-Whitney U test and ROC  curve analysis, were \nperformed to assess the diagnostic performance of t hese parameters. Z test \nwas performed to compare the ROC curves of differen t parameters. \nResults or Findings: This study analyzed 59 lesions in 27 patients to ev aluate \nthe diagnostic performance of various spectral para meters. The study found \nthat EDW had the highest area under the curve (AUC)  of 0.957, with an \nsensitivity of 95%, specificity of 85%. MonoE at 10 0 keV achieved an AUC of \n0.955, with high sensitivity (95%) and negative pre dictive value (94%). In \ncontrast, monoE at 40 keV showed the lowest perform ance, with an AUC of \n0.701 and sensitivity of 66%. Conventional CT image s had an AUC of 0.887, \ncomparable to monoE at 70 keV (0.910). Statisticall y significant differences \nwere noted between the AUCs of EDW/100 keV and conv entional CT, while no \nsignificant difference was found between EDW and 10 0 keV. \nConclusion: Our findings indicated that EDW and monoE CT images  obtained \nfrom DLCT can improve the detection of AIS compared  to conventional non-\ncontrast CT imaging. \nLimitations: The sample size was small. \nFunding for this study: None \nEthics committee - additional information: This retrospective study was \napproved by the local Ethics Committee of Civil Avi ation General Hospital \n(2024-L-K-122). \nAuthor Disclosures:  \nJing Wen: Nothing to disclose \nHaifeng Zhu: Nothing to disclose \nYujie Wang: Nothing to disclose \nShanrui Ma: Nothing to disclose \nShan Yang: Nothing to disclose \n \n \nDTI-derived Perivascular Space Diffusion Index coul d mirror the \npolarization of AQP4 following cerebral ischemia \n*X. Hao*, J. Tian, Z. Yao; Shanghai/CN \n(haoxiaozhu123@sohu.com) \n \nPurpose or Learning Objective: The purpose of this study was to investigate \nthe dynamic changes of the perivascular space diffu sion index (ALPS) and its \ncorrelation with aquaporin 4 (AQP4) polarization fo llowing cerebral ischemia in \nrats, using advanced diffusion tensor imaging (DTI)  technique. \nMethods or Background: Rats were divided into a normal group (n=5) and an \nischemic group (n=25). The ischemic group underwent  transient middle \ncerebral artery occlusion (tMCAO) and was further s ubdivided into five \nsubgroups (n=5 each) based on the time post-ischemi a (1, 3, 7, 14, and 28 \ndays). Rats underwent MRI scans, including DTI, T2- weighted imaging (T2WI), \nand susceptibility-weighted imaging (SWI). Subseque ntly, immunofluorescence \nstaining for AQP4 and glial fibrillary acidic prote in (GFAP) was performed. The \nALPS index was analyzed based on T2W, SWI, and frac tional anisotropy (FA) \nobtained by DTI post-processing. And regions of int erest were selected on the \nipsilateral periventricular area, the ipsilateral c orpus callosum/cingulate area \nand their mirror areas of the contralateral side. T he AQP4 polarization was \nanalyzed by GFAP/AQP4 in the non-glial scar area ar ound the infarction in \ncortex and striatum. \nResults or Findings: ALPS indexes were markedly reduced in ischemic rats , \nparticularly on the affected side, with a notable d rop on day 1, then rising at \ndays 14 and 28. AQP4 polarization mirrored this tre nd, falling initially, then \nrising significantly by days 14 and 28. The ALPS in dex closely aligns with \nAQP4 index fluctuations. \nConclusion: The DTI-based ALPS index mirrors changes in AQP4 po larization \nafter stroke, dropping sharply in the hyperacute ph ase and recovering in the \nearly chronic phase, which could be a useful biomar ker for glymphatic pathway \nfunction following stroke. \nLimitations: This experiment requires high consistency of the le sions. \nFunding for this study: National Natural Science Foundation of China (No. \n81801660) and the grant of National Natural Science  Foundation of China (No. \n82272061) \nEthics committee - additional information: Institutional Animal Care and \nUse Committee of Fudan University \nAuthor Disclosures:  \nXiaozhu Hao: Nothing to disclose \nZhenwei Yao: Nothing to disclose \nJiaqi Tian: Nothing to disclose \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 85  \nA novel model to quantify blood transit time in cer ebral arteries using \nASL-based 4D magnetic resonance angiography with ex ample clinical \napplication in moyamoya disease \nA. Bhogal¹, S. Uniken Venema¹, *P. T. Deckers*¹, K.  Van De Ven²,  \nM. Versluis², K. Braun¹, B. Van Der Zwan¹, J. Siero ¹; ¹Utrecht/NL, ²Best/NL \n(p.t.deckers-3@umcutrecht.nl) \n \nPurpose or Learning Objective: Angiography is critical for visualizing \ncerebral blood flow in intracranial steno-occlusive  diseases. Current 4D \nmagnetic resonance angiography (MRA) techniques pri marily focus on \nmacrovascular structures, yet few have quantified h emodynamic timing. This \nstudy introduces a novel model to estimate macrovas cular arterial transit time \n(mATT) derived from arterial spin labeling (ASL)-ba sed 4D-MRA. We provide \nexamples of our method that visualize mATT differen ces throughout the brain \nof patients with intracranial steno-occlusive disea se (moyamoya), as well as \nchanges in mATT resulting from the cerebrovascular reactivity (CVR) response \nto acetazolamide (ACZ). \nMethods or Background: The study population consisted of twelve patients \nwith intracranial steno-occlusive disease, with a c linical indication to undergo \nhemodynamic imaging with an ACZ challenge to measur e CVR. CVR is \nmeasured using multi-PLD ASL-MRI, acquired pre- and  post-ACZ, and \nincludes a four-dimensional dynamic MRA sequence us ing an ASL-scheme. \nThe scan indications varied, but mostly involved ce rebrovascular \nhemodynamic evaluation before or after neurosurgica l intervention. \nResults or Findings: We provide examples of our method that visualize mA TT \ndifferences throughout the brain of patients with i ntracranial steno-occlusive \ndisease (moyamoya), as well as changes in mATT in r esponse to an ACZ \ninjection. Furthermore, we present a method that pr ojects sparse arterial \nsignals into a 3D native brain-region atlas space a nd correlates regional mATT \nwith other hemodynamic parameters of interest, such  as tissue transit time and \nCVR. \nConclusion: Our approach offers a non-invasive, quantitative as sessment of \nmacrovascular dynamics, which enhances the understa nding of large-vessel \nand tissue-level hemodynamics and augment monitorin g of treatment \noutcomes in steno-occlusive disease patients. This can directly be used in \nstroke trial stratifications and peri-procedural tr eatment monitoring. \nFurthermore, it sets the stage for more in-depth in vestigations of the \nmacrovascular contribution to brain hemodynamics. \nLimitations: Pilot data in a small subgroup of patients (n=12). \nFunding for this study: This work was supported by the W.M. De Hoop \nFoundation, the Janivo Foundation and Friends of UM C Utrecht & Wilhelmina \nChildren’s Hospital, and an NWO VIDI grant awarded to A.A.B. \n(VI.Vidi.223.085). JCWS is supported by the Brain C enter Young Talent \nFellowship 2019 of the University Medical Center Ut recht, The Netherlands. \nEthics committee - additional information: The Medical Ethics Review \nCommittee of the University Medical Centre Utrecht declared that the Medical \nResearch Involving Human Subjects Act (WMO) did not  apply to the present \nresearch since all study measures were part of rout ine clinical practice. All \npatients or their legal representative (i.e., paren t or guardian) provided written \ninformed consent to use their data. Healthy subject s were acquired under a \nsequence development ethical protocol, which was ap proved by the Medical \nEthics Review Committee of the University Medical C entre Utrecht. Informed \nconsent was given by each healthy subject. \nAuthor Disclosures:  \nKees Braun: Nothing to disclose \nBart Van Der Zwan: Nothing to disclose \nAlex Bhogal: Nothing to disclose \nSimone Uniken Venema: Nothing to disclose \nMaarten Versluis: Other: Works for Philips Healthca re and provided the \nadvanced ASL patch used in this work. No financial support was provided. \nJeroen Siero: Nothing to disclose \nPieter Thomas Deckers: Nothing to disclose \nKim Van De Ven: Other: Works for Philips Healthcare  and provided the \nadvanced ASL patch used in this work. No financial support was provided. \n \n \nThe use and pitfalls of hemodynamic MRI using multi delay arterial  \nspin labelling for intracranial steno-occlusive dis ease in clinical practice: \na single-center experience \n*S. Uniken Venema*, P. Deckers, J. W. Dankbaar, B. Van Der Worp,  \nJ. Hendrikse, B. Van Der Zwan, K. Braun, A. Bhogal,  J. Siero; Utrecht/NL \n(s.m.unikenvenema@umcutrecht.nl) \n \nPurpose or Learning Objective: The primary objective is to describe a \nclinically feasible advanced neuroimaging protocol developed at an academic \nmedical center that uses multi-delay arterial spin labeling (ASL) and blood \noxygen level dependent (BOLD)-MRI with acetazolamid e. This protocol is \ndesigned to assess cerebrovascular reactivity (CVR)  in patients with \nintracranial steno-occlusive disease (e.g. moyamoya ), while avoiding the \nlimitations of PET scans. \n \nMethods or Background: Image acquisition on 3-Tesla MRI involves \nacetazolamide-augmented multi-delay ASL and dynamic  BOLD, in addition to \nstructural sequences. Image processing is done usin g customized MATLAB-\nbased toolboxes. ASL-CVR is calculated by subtracti ng pre-acetazolamide \ncerebral blood flow (CBF) from post-acetazolamide C BF and additional \nhemodynamic maps, such as arterial transit time (AT T), are generated. \nImaging interpretation includes assessment of scan quality and success of the \nhemodynamic challenge. \nResults or Findings: Since 2018, approximately 100 patients were scanned  \nusing this protocol. Multi-delay ASL enables a more  accurate assessment of \nCBF and CVR compared to single-delay ASL in patient s with prolonged ATT \nowing to their stenosis, and enables quantifying AT T simultaneously – a useful \nmarker in itself. While CVR assessment is primarily  done using ASL-derived \nCVR maps, BOLD-CVR provided useful complementary in formation in some \ncases. A typical patient with an intracranial steno sis experiencing ischemic \nsymptoms will demonstrate lower baseline CBF, lower  CVR and prolonged \nATT in the affected hemisphere. \nConclusion: Our multi-delay ASL-based protocol demonstrates cli nical \nfeasibility and utility, allowing detailed cerebral  hemodynamic evaluations of \nindividual patients that is useful for clinical dec ision-making. This work serves \nas a practical guide for clinicians and MRI experts  seeking to implement these \nadvanced imaging methods in their institutions. \nLimitations: Potential pitfalls in imaging acquisition and inter pretation, \nincluding motion artefacts, inadequate labeling, th e effects of anesthesia on \nCVR and the uncertainties of acetazolamide-augmente d BOLD, must be \ncarefully considered. \nFunding for this study: This work was supported by the W.M. De Hoop \nFoundation, the Janivo Foundation and Friends of UM C Utrecht & Wilhelmina \nChildren’s Hospital. \nEthics committee - additional information: NedMec (study number 21-406) \nAuthor Disclosures:  \nKees Braun: Nothing to disclose \nJan Willem Dankbaar: Nothing to disclose \nBart Van Der Zwan: Nothing to disclose \nAlex Bhogal: Nothing to disclose \nPieter Deckers: Nothing to disclose \nBart Van Der Worp: Board Member: Past president at the European Stroke \nOrganization Advisory Board: Liva Nova and Bayer Gr ant Recipient: Dutch \nHeart Foundation, Stryker, and the European Union \nSimone Uniken Venema: Nothing to disclose \nJeroen Siero: Nothing to disclose \nJeroen Hendrikse: Nothing to disclose \n \n \nPredictive Value of Venous Outflow in SAH (PreViOS)  \n*H. Briody*, J. Henry, R. Bruen, P. Mchugh, P. Roha n, M. Javadpour,  \nP. Nicholson; Dublin/IE \n(hayleybriody@rcsi.ie) \n \nPurpose or Learning Objective: Favorable cortical venous outflow (VO) is \nlinked to better outcomes in acute ischemic stroke.  It’s role in aneurysmal \nsubarachnoid hemorrhage (aSAH) remains unclear. Thi s study investigates the \nassociation between VO profiles and functional outc omes in aSAH. \nMethods or Background: Patients with aSAH referred to a tertiary \nneurosurgery center between 2016 and 2023 were incl uded if presentation \ncomputed tomographic angiography (CTA) demonstrated  satisfactory venous \nsystem opacification. VO was assessed using the cor tical vein opacification \nscore (COVES). The primary outcome was poor functio nal outcome (Glasgow \nOutcome Scale [GOS] 1-3) at 90 days. Associations b etween COVES and \noutcomes were assessed using univariable and multiv ariable (adjusted for \nWorld Federation of Neurosurgical Societies [WFNS] grade) binomial \nregression. \nResults or Findings: Of 675 patients with aSAH and available CTA, 204 \n(30%) met inclusion criteria. The median age was 54 .2 years (range 12-85). \n182 (89%) had favorable VO. No significant associat ion was found between \nunfavorable VO and poor 90-day functional outcome ( RR 0.78, 95% CI 0.38-\n1.58, p=0.48), even after adjusting for WFNS grade (RR 1.04, 95% CI 0.38-\n2.83, p=0.94). Unfavorable VO was not associated wi th poor outcome at \ndischarge, need for cerebrospinal fluid diversion, or shunt dependence. \nConclusion: The study failed to demonstrate a link between veno us outflow \n(as measured by COVES) and outcomes in aSAH. This i s an important \nnegative finding. It suggests that, unlike in ische mic stroke, venous outflow \nmight not be a major determinant of outcome in aSAH . Prospective studies are \nneeded to definitively assess the role of VO in aSA H. \nLimitations: This retrospective, single-center study may be subj ect to \nselection bias due to the inclusion criteria requir ing adequate jugular bulb \nopacification on CTA. The impact of delayed cerebra l ischemia, a major \ndeterminant of outcome in aSAH, was not specificall y assessed. \nFunding for this study: None \nEthics committee - additional information: Institutional review board \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 86  \nAuthor Disclosures:  \nPatrick Nicholson: Nothing to disclose \nMohsen Javadpour: Nothing to disclose \nPaul Mchugh: Nothing to disclose \nRichard Bruen: Nothing to disclose \nHayley Briody: Nothing to disclose \nPat Rohan: Nothing to disclose \nJack Henry: Nothing to disclose \n \n \nEnhanced Detection of Cerebral Lesions in Cerebral Amyloid Angiopathy \nUsing 7T MRI: Insights into Cognitive Correlation a nd Clinical \nImplications \n*D. Botta*, A. Cusin, L. Sveikata, K-O. Loevblad, F . T. Kurz; Geneva/CH \n \nPurpose or Learning Objective: The aim of this study is to evaluate the \ndetection of cerebral microbleeds (CMBs) and cortic al microinfarcts using 7 \nTesla (7T) MRI in patients with cerebral amyloid an giopathy (CAA) and to \ncorrelate these findings with cognitive performance  as measured by the \nMontreal Cognitive Assessment (MoCA). \nMethods or Background: 19 patients with probable or possible CAA were \nscanned using both 7T and 3T MRI. Imaging protocols  at 7T included \nsusceptibility-weighted imaging (SWI) for CMB detec tion with an in-plane \nresolution of 0.15x0.15mm, and 3D FLAIR and T1 MP2R AGE sequences for \ncortical microinfarcts detection. Cognitive functio n was assessed using the \nMoCA score. Statistical analyses were conducted to assess correlations \nbetween lesion burden and MoCA scores. \nResults or Findings: 7T MRI detected 379 CMBs compared to 179 at 3T \n(average: 18.37±34.18 CMBs at 7T vs 9.53±15.03 at 3T). Additionally, 7T \nidentified 27 cortical microinfarcts across 5 patie nts, while none were detected \nat 3T. MoCA scores ranged from 9 to 30 with a mean of 23.0±4.82. Weak \ncorrelations were found between the number of CMBs and MoCA scores (7T: r \n=-0.28; 3T: r =-0.24). \nConclusion: 7T MRI at high-resolution is superior to 3T for det ecting CMBs \nand cortical microinfarcts in CAA patients. However , the weak correlation \nbetween lesion burden and cognitive decline suggest s that other factors may \nalso contribute to cognitive impairment in these pa tients. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Written informed consent was \nobtained from all participants, the study was appro ved by the institutional \nreview board of Geneva University Hospitals. \nAuthor Disclosures:  \nProfessor Karl-Olof Loevblad: Nothing to disclose \nLukas Sveikata: Nothing to disclose \nAlexandre Cusin: Nothing to disclose \nFelix T Kurz: Nothing to disclose \nDaniele Botta: Nothing to disclose \n \n \nDiagnostic Performance of Low-Dose Cerebral CTA Ima ges Using \nArtificial Intelligence Iterative Reconstruction fo r Differentiating \nintracranial Aneurysms and Infundibula \nH. Chen¹, S. Xu², G. Zhang², *J. Wang*¹, X. Yin¹; ¹ Baoding/CN, ²Shanghai/CN \n(jianing0218@163.com) \n \nPurpose or Learning Objective: Intracranial aneurysms and infundibula with \nsimilar morphology and anatomical location are diff icult to distinguish using \nlow-dose cerebral CT angiography (CTA). This study evaluated the diagnostic \nperformance of cerebral low-dose CTA with artificia l intelligence iterative \nreconstruction (AIIR) for differentiating intracran ial aneurysms and infundibula. \nMethods or Background: Sixty-four patients (38 male, mean age 62.2 ± 12.5 \nyears) with suspected intracranial aneurysms were p rospectively enrolled. \nEach patient underwent routine-dose (RD) and low-do se (LD) cerebral CTA. \nThe RD protocol used 100kVp, ref. 180mAs, and hybri d iterative reconstruction \n(HIR), whereas the LD protocol used 100kVp, ref. 30 mAs, and AIIR. Two \nradiologists, blinded to scan/reconstruction parame ters, independently \ndetected aneurysms and infundibula on low-dose imag es. The diagnostic \nreports of RD CTA served as references. Diagnostic performance in \ndifferentiating aneurysms and infundibula was asses sed using receiver \noperating characteristic (ROC) analysis, calculatin g sensitivity, specificity, \npositive predictive value (PPV), negative predictiv e value (NPV), accuracy, and \narea under the curve (AUC) with 95% confidence inte rval (CI). \nResults or Findings: A total of 64 lesions were identified, including 44  \naneurysms and 20 infundibula. Reader 1 detected 62 out of 64 lesions (96.9%) \non low-dose images, while Reader 2 detected 61 out of 64 (95.3%). Two \ninfundibula with sizes of 1.5 mm and 2.4 mm were mi ssed by both readers, \nwhereas one aneurysm of 1.8 mm was missed by Reader  2. In differentiating \naneurysms and infundibula, the sensitivity, specifi city, PPV, NPV, diagnostic \naccuracy, and AUC for Reader 1 were 97.72%, 100%, 1 00%, 94.74%, 98.39%, \nand 0.989 (95% CI: 0.963–1.015), while for Reader 2  they were 97.67%, \n100%, 100%, 94.74%, 98.36%, and 0.988 (95% CI: 0.96 2–1.015). \nConclusion: The AIIR shows the potential in reducing the radiat ion dose of the \ncerebral CTA when diagnosing intracranial aneurysms  and infundibula. \nLimitations: Not applicable. \nFunding for this study: the Key Research and Development Program of \nHebei Province (grant number 202330604010017) \nEthics committee - additional information: This study was approved by the \nlocal institutional review board. \nAuthor Disclosures:  \nGuozhi Zhang: Employee: at United Imaging Healthcar e \nShijie Xu: Nothing to disclose \nJianing Wang: Nothing to disclose \nHaoyan Chen: Nothing to disclose \nXiaoping Yin: Nothing to disclose \n \n \nTemporal muscle trophicity as a prognostic factor f or functional recovery \nin non-traumatic intracerebral hemorrhage \n*S. Nataf*¹, O. Curtinot², T. Maghfour², G. Bouloui s², A. Aignatoaie¹,  \nC. Ozsancak¹, M. Pasi², C. Cohen¹; ¹Orleans/FR, ²To urs/FR \n(snataf98@gmail.com) \n \nPurpose or Learning Objective: Intracerebral hemorrhage (ICH) is \nassociated with poor outcome. Identifying patients with higher risk of disability \nis a key feature of optimal care. Recently, tempora l muscle thickness (TMT) \nhas been shown to predict ability after ischemic st roke. We explored the \nrelationship between temporal muscle trophicity and  the functional ability of \nnon-traumatic ICH patients. \nMethods or Background: Patients with acute non-traumatic ICH (2021-2022) \nfrom two university centers were retrospectively in cluded. Imaging ICH \nparameters were retrieved (e.g. volume, location). Temporal trophicity was \nassessed through thickness (TMT), area (TMA) and de nsity (TMD) on baseline \nbrain computed tomography (CT). Good prognosis at 3 -6 months was defined \nas a modified Rankin Scale (mRS)<3. Association bet ween TMT, TMA and \nTMD and 3-6months mRS was evaluated using student T -test. \nResults or Findings: Among 453 ICH patients, 216 with both CT and 3-\n6months mRS were analyzed (49% women, 70±15.9y). Me an hemorrhage \nvolume was 47.3±46mm³, mainly lobar (46%) and locat ed in basal ganglia \n(41%). Mean TMT was 7.32±2.70mm in the good prognos is group, vs \n6.37±2.13mm in the poor prognosis group (p=0.005). Mean TMA was \n408±203mm² in the good prognosis group, vs 343±161mm² in the poor \nprognosis group (p=0.01). Prognosis was not signifi catively associated with \nTMD (p=0.051). 3-6months mRS was associated with he morrhage volume \n(p=0.01). \nConclusion: To our knowledge, this is the first study to analyz e and \ndemonstrate that temporal muscle trophicity serves as a prognostic factor in \nintracerebral hemorrhage (ICH). Temporal muscle thi ckness (TMT) is a simple \nand practical indicator for assessing overall healt h and guiding patients toward \ntargeted rehabilitation. \nLimitations: Since many patients underwent only MRI, further res earch is \nneeded to evaluate the association between temporal  muscle trophicity and \nprognosis using this imaging modality. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was approved by a \nComité de protection des personnes (CPP). \nAuthor Disclosures:  \nOlivier Curtinot: Nothing to disclose \nMarco Pasi: Nothing to disclose \nSimon Nataf: Nothing to disclose \nGrégoire Boulouis: Nothing to disclose \nAndreea Aignatoaie: Nothing to disclose \nCanan Ozsancak: Nothing to disclose  \nClara Cohen: Nothing to disclose \nTasnym Maghfour: Nothing to disclose \n \n \n3D variable flip angle turbo spin echo black-blood MRI for diagnosing \ncerebral venous thrombosis: a systematic review and  meta-analysis \n*A. Akhavi Milani*; Tabriz/IR \n(ali.akhavi.milani@gmail.com) \n \nPurpose or Learning Objective: To evaluate the performance of 3D variable \nflip angle turbo spin echo black-blood MRI (BB-MRI)  in diagnosing cerebral \nvenous thrombosis (CVT). The secondary objectives i ncluded, comparing BB-\nMRI with conventional MRI, MRV, MPRAGE, and SWI in diagnosing CVT, and \nassessing the utility of BB-MRI in estimating throm bus age. \nMethods or Background: This study was registered in PROSPERO [ID: \nblinded]. The PubMed/MEDLINE, Web of Science, Scopu s, and Embase \ndatabases were systematically searched and studies were selected based on \npredefined eligibility criteria. The risk of bias w as assessed using the \nQUADAS-2 tool. Meta-analysis was performed to calcu late pooled sensitivity, \nspecificity, and AUC. \n\n \n \nThursday \nAbstract-based Programme \n \n 87  \nResults or Findings: Nine studies were included in the review. Two of th em \ninsufficiently reported the quantitative data; ther efore, seven studies involving \n176 CVT patients and 217 controls, encompassing 610  thrombosed and 3,279 \nnormal cerebral venous segments were included in th e meta-analysis. The \nstudies demonstrated a high bias risk in the patien t selection and reference \nstandard domains. The pooled sensitivity and specif icity of BB-MRI were 96% \n[95% CI: 92%–98%] and 96% [95% CI: 93%–98%] on a pa tient-based level, \nand 92% [95% CI: 87%–95%] and 98% [95% CI: 92%–99%]  on a venous \nsegment-based level, respectively. The AUC was 0.98  for patient-based data \nand 0.96 for venous segment-based data. For seconda ry objectives, a \nnarrative summary indicated that BB-MRI outperforms  conventional MRI, MRV, \nand MPRAGE. It also outperforms SWI in assessing ce rebral cortical veins. \nMoreover, BB-MRI can prove beneficial in thrombus a ge estimation. \nConclusion: BB-MRI demonstrates significant potential in diagno sing CVT. \nFurther comparative studies are required to specify  its role in clinical decision-\nmaking for CVT. \nLimitations: The small number of the retrieved studies. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This was a systematic review \nstudy for which ethics committee approval was waive d. \nAuthor Disclosures:  \nAli Akhavi Milani: Nothing to disclose \n \n \nThe probability of cerebral amyloid angiopathy acco rding to the \nSimplified Edinburgh CT criteria in a large, unsele cted lobar intracerebral \nhemorrhage population \n*A. Hillal*¹, T. Ullberg², J. Wassélius¹; ¹Lund/SE,  ²Malmö/SE \n(amirhilal13@gmail.com) \n \nPurpose or Learning Objective: Early identification of the underlying cause of \nintracerebral hemorrhage (ICH) is important for tre atment and prognosis. This \nstudy aims to investigate the association of hemato ma volume and other \nclinical parameters on the distribution of Cerebral  Amyloid Angiopathy (CAA) \nprobability according to the simplified Edinburgh C T criteria in a large, \nunselected intracerebral hemorrhage (ICH) populatio n. \nMethods or Background: Patients with spontaneous ICH residing in Skane \ncounty registered with clinical data in the Swedish  Stroke Register 2016–2020 \nwere included. Radiological parameters were evaluat ed using baseline non-\ncontrast CT (NCCT) for categorization according to the simplified Edinburgh \nCT criteria by the presence of subarachnoid hemorrh age (SAH) and fingerlike-\nprojections (FLP). Multivariable logistic regressio n analysis was used to \ndetermine factors associated with an increased (int ermediate/high) CAA \nprobability. \nResults or Findings: Of 666 patients with lobar ICH, 190 (29%) had high CAA \nprobability, 92 (14%) intermediate, and 384 (58%) l ow CAA probability. \nPatients with increased CAA probability more often presented with decreased \nlevel of consciousness, larger hematoma volumes, an d had higher 90-day \nmortality. Female sex, age, and increasing baseline  hematoma volume (Odds \nRatio up to 30) were associated with increased odds  of having an increased \nCAA probability. \nConclusion: We identified a strong association between baseline  hematoma \nvolume and an increased probability of CAA in lobar  ICH patients on NCCT, \nindicating that large hematoma volumes per se may c ontribute to the \noccurrence of FLP and SAH, and act as a confounder for the Simplified \nEdinburgh CT Criteria. Validation against MRI is wa rranted. \nLimitations: The lack of MRI studies to allow for the correlatio n between CT \nimaging characteristics and the gold imaging standa rd MRI Boston criteria \nFunding for this study: ALF grants to Teresa Ullberg and Johan Wasselius, \nthe Crafoord Foundation to JW, VINNOVA to Johan Was selius, and by SUS \nStiftelser & Fonder to Johan Wasselius. \nEthics committee - additional information: This study was approved, and \nindividual informed consent was waived by the Swedi sh Ethical Review \nAuthority (reference number 2020-06800). \nAuthor Disclosures:  \nTeresa Ullberg: Nothing to disclose \nJohan Wassélius: Nothing to disclose \nAmir Hillal: Nothing to disclose \n \n \nEvaluation of a FLAIR Hyperintensity Algorithm for the prediction of DWI-\nFLAIR Mismatch in Acute Ischemic Stroke \n*C. M. Offersen*, J. Johansen, A. H. Brandt, T. C. Truelsen, A. Pai, S. Darkner, \nM. B. Bachmann Nielsen, J. F. Carlsen; Copenhagen/D K \n \nPurpose or Learning Objective: Moderate inter-rater variability of Diffusion-\nWeighted Imaging (DWI) – Fluid-Attenuated Inversion  Recovery (FLAIR) \nmismatch in wake-up stroke raises concerns about th e potential exclusion of \neligible patients for treatment with thrombolysis. A novel FLAIR algorithm has \nshown promising potential to perform this mismatch assessment but has only \nbeen evaluated on a small dataset. In the present s tudy, we aimed to evaluate \nan updated version of the FLAIR algorithm for predi cting the DWI-FLAIR \nmismatch in a large cohort of wake-up stroke patien ts. \nMethods or Background: We conducted a single-centre, retrospective study. \nA consecutive cohort of patients suspected of wake- up stroke, who underwent \nMRI between 2019 and 2021 was included. Two radiolo gists and one resident, \nblinded to clinical data, manually assessed DWI-FLA IR mismatch according to \nthe current clinically used binary categorisation. Cohens Kappa was calculated \nfor the inter-rater agreement. The FLAIR algorithm depends on a DWI \nsegmentation. We used a commercial DWI segmentation  model and then \ntested the ability of the FLAIR algorithm on the id entified ischemic lesions to \npredict manual mismatch. This was analysed with log istic regression test. \nResults or Findings: The DWI model identified 495 patients with possible  \nischemic lesions. Manual radiological assessments f ound 365 of those patients \nto have actual ischemic lesions. Inter-rater agreem ent for binary DWI-FLAIR \nmismatch assessment was moderate (κ = 0.461 ± 0.028SD). We found a high \naccuracy (82.3 ± 3.3SD). Low sensitivity (59.1 ± 9.1SD). Specificity (91.5 ± \n4.2SD) and AUC (0.845 ± 0.03SD) were high in the au tomatic mismatch \nassessment. \nConclusion: The FLAIR algorithm predicted DWI-FLAIR mismatch st atus with \na high AUC, which suggests the algorithm could prov ide a more standardized \ndecision on mismatch, and reduce the inter-rater va riability through an \nobjective assessment to assist the radiologist. \nLimitations: This was a single-centre, retrospective study. \nFunding for this study: Innovation Fund Denmark \nEthics committee - additional information: Danish National Center for Ethics \nAuthor Disclosures:  \nThomas Clement Truelsen: Nothing to disclose \nCecilie Mørck Offersen: Grant Recipient: Received f unding from IFD for part of \nthe work focused on automated assessment in wake up  stroke. This project \nwas in collaboration with Cerebriu A/S. \nMichael B Bachmann Nielsen: Nothing to disclose \nAkshay Pai: Founder: Founder and CTO at Cerebriu A/ S, a company that \nmakes automated diagnostic software for radiology. \nJacob Johansen: Other: Earlier employment as an ind ustrial PhD student at \nCerebriu A/S \nAndreas Hjelm Brandt: Nothing to disclose \nSune Darkner: Nothing to disclose \nJonathan Frederik Carlsen: Nothing to disclose \n \n \n10:00-11:00 Research Stage 1 \nResearch Presentation Session: Head and \nNeck \nRPS 808 \nImaging the skull base and face \n \nModerator \nA. Bernaerts; Antwerp/BE  \n(anja.bernaerts@zas.be) \n \n \nDual-layer spectral detector CT for differentiating  middle ear \ncholesteatoma and chronic suppurative otitis media \n*S. Zhou*¹, L. Mei¹, H. Liu¹, X. M. Liu², J. Li¹; ¹ Changsha/CN, ²Guangzhou/CN \n(zsy_amory1003@163.com) \n \nPurpose or Learning Objective: To compare the diagnostic performance of \ndual-layer spectral detector CT (DLCT) and high-res olution CT (HRCT) in \ndifferentiating middle ear cholesteatoma and chroni c suppurative otitis media. \nMethods or Background: This prospective, institutional review board-\napproved study included sixty-six patients who were  preliminary diagnosed as \ncholesteatoma or otitis media, and received DLCT sc anning before surgery. \nThirty-three patients were finally diagnosed choles teatoma based on \nintraoperative or pathological findings. Two blinde d readers (Reader 1: one \nradiologist; Reader 2: one otologist) provided diag noses and diagnostic \nconfidence scores using a five-point scale, based o n HRCT images and DLCT \nmulti-parameter images, including virtual mono-ener getic image at 40keV (VMI \n40keV) and effective atomic number (Zeff). Diagnost ic accuracy of HRCT and \nDLCT maps were compared using McNemar’s test. Inter observer agreement \nwas evaluated by Kappa statistic. \nResults or Findings: HRCT and DLCT identified a total of 27/33 and 31/33  \ncholesteatomas by Reader 1, 19/33 and 30/33 cholest eatomas by Reader 2 \nrespectively. The sensitivity, specificity, PPV, NP V, accuracy of HRCT and \nDECT by Reader 1 were 81.8, 75.8, 77.1, 68.2, 78.8%  and 93.9, 69.7, 75.6, \n92.0, 87.5%, by Reader 2 were 57.6, 90.9, 86.3, 68. 2, 74.2% and 90.9, 75.8, \n\n \n \nThursday \nAbstract-based Programme \n \n 88  \n78.9, 89.3, 83.3%, respectively. Compared to HRCT, the diagnostic sensitivity \nof DLCT increased for both readers, with a statisti cally significantly \nimprovement in Reader 2 (p＜0.05). Diagnostic confidence scores of DLCT \nversus HRCT by Reader 1 and Reader 2 both improved significantly \n(4.85±0.36 vs 3.77±1.30, 4.44±0.73 vs 3.45±0.98, respectively, all p＜0.05). \nInterobserver reproducibility was higher for diagno ses made with DLCT maps \n(k =0.717) than for that made with HRCT images (k =  0.495). \nConclusion: Spectral CT improved diagnostic performance and int erobserver \nreproducibility of determination of cholesteatoma v ersus otitis media. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: lRB of Xiangya Hospital \nAuthor Disclosures:  \nShuangyuan Zhou: Nothing to disclose \nXiao Min Liu: Nothing to disclose \nJuan Li: Nothing to disclose \nLingyun Mei: Nothing to disclose \nHeng Liu: Nothing to disclose \n \n \nRedefining Sinonasal Cancer response assessment to induction \nchemotherapy with tumor volumetry, results from two  prospective \nmulticentric trials: SINTART-1 and SINTART-2 \nP. Rondi¹, *E. Massoni*¹, A. Borghesi¹, P. Bossi², C. Resteghini², D. Farina¹,  \nM. Ravanelli¹; ¹Brescia/IT, ²Milan/IT \n(e.massoni@unibs.it) \n \nPurpose or Learning Objective: Aim of this study is to identify the best \nresponse criteria in patients with sinonasal cancer  undergoing induction \nchemotherapy (IC). \nMethods or Background: Patients enrolled in SINTART-1 and SINTART-2 \nwere included in this study. Unidimensional diamete rs (antero-posterior, AP; \nlatero-lateral, LL; cranio-caudal, CC), maximum axi al area (Amax) and volume \n(V) were performed on MRI by two radiologists. RECI ST 1.1 assessment was \nincluded as a categorical variable. Variables were evaluated at baseline, after \n1st chemotherapy cycle and at best response. Intero bserver repeatability was \nanalyzed. Stepwise univariable and multivariable Co x proportional-hazards \nregression models were used to correlate variables with Disease-Free Survival \n(DFS) and Overall Survival (OS). \nResults or Findings: 60 patients were included in this study. Interobser ver \ncorrelation at baseline and after 1st IC cycle was excellent for V (0.916 and \n0.928 respectively), CC diameter (0.96 and 0,863) a nd AP diameter (0.846 and \n0.796); good for LL diameter and moderate for Amax.  RECIST 1.1 criteria after \n1st IC cycle and at best response were not correlat ed with OS and DFS. \nVolume variation after 1st IC cycle and at best res ponse was the variable most \ncorrelated with OS (p<0.0001and p=0.002) and DFS (p <0.0001and p=0.005). \nAt the multivariable analysis V variation after 1st  IC cycle and at best response \nresulted to be the only variable statistically asso ciated with OS (p<0.001 and \np=0.0019 respectively) and DFS (p=0.0004and p=0.004 respectively). \nConclusion: Volume variation should be preferred to RECIST 1.1 and other \nmeasurements as objective radiological response in sinonasal cancer. \nLimitations: The main limitation of this study is that despite c onsidering two \nprospective studies the size of the cohort is small  and this could reduce the \nstatistical power. \nFunding for this study: This study has received no funding. \nEthics committee - additional information: The ethics committee of \nparticipating centers have approved this study \nAuthor Disclosures:  \nElena Massoni: Nothing to disclose \nAndrea Borghesi: Nothing to disclose \nPaolo Bossi: Nothing to disclose \nDavide Farina: Nothing to disclose \nPaolo Rondi: Nothing to disclose \nMarco Ravanelli: Nothing to disclose \nCarlo Resteghini: Nothing to disclose \n \n \nDiagnostic accuracy of MRI for orbital and intracra nial invasion of \nsinonasal malignancies: a systematic review and met a-analysis \n*U. B. Abdullaeva*¹, B. Pape², J. Hirvonen³; ¹Tashk ent/UZ, ²Turku/FI, \n³Tampere/FI \n(umidasamira2@gmail.com) \n \nPurpose or Learning Objective: To review the diagnostic accuracy of MRI in \ndetecting orbital and intracranial invasion of sino nasal malignancies using \nhistopathological or surgical evidence as the refer ence standard. \n \n \n \n \nMethods or Background: The systematic review protocol was pre-registered \nin the Prospective Register of Systematic Reviews ( PROSPERO) under \nregistration number CRD42024492090. A systematic se arch of the studies in \nEnglish was conducted in PubMed and Embase, limited  to articles published \nsince 1990. We included studies that used preoperat ive MRI to detect \nintracranial and orbital invasion of sinonasal mali gnancies, using histological or \nsurgical confirmation as the reference standard, an d reported patient numbers \nin each class required for assessing diagnostic acc uracy. The outcome \nmeasures were sensitivity, specificity, positive pr edictive value (PPV), and \nnegative predictive value (NPV). Heterogeneity was assessed with the Higgins \ninconsistency test (I2). \nResults or Findings: Seven original articles with 546 subjects were incl uded \nin the review, six of these in the meta-analysis. P ooled overall accuracy for \norbital invasion was higher at 0.88 (95% CI, 0.75-0 .94) than for intracranial \ninvasion - 0.80 (95% CI, 0.76-0.83). Meta-analytic estimates and their 95% \nconfidence intervals were as follows for intracrani al/orbital invasion: sensitivity \n0.77 (0.69-0.83)/ 0.71 (0.40-0.90); specificity 0.7 9 (0.74-0.83)/0.91 (0.78-0.97); \nPPV 0.76 (0.64-0.85)/0.78 (0.61-0.88); and NPV 0.82  (0.72-0.89)/0.90 (0.63-\n0.98). Significant heterogeneity was observed in th e Higgins inconsistency test \n(I2) for orbital invasion (84%, 83%, and 93% for se nsitivity, specificity, and \nNPV, respectively). \nConclusion: MRI yielded moderate to high diagnostic accuracy fo r intracranial \nand orbital invasion, but there are limitations lea ding to false diagnoses. Loss \nof the hypointense zone on MRI predicts dural invas ion. Infiltration of the \nextraconal fat beyond the periorbita is an MRI feat ure of orbital invasion. \nLimitations: Limitations include a small number of predominantly  retrospective \nstudies, some with a small subset of patients. \nFunding for this study: Funding was provided by the Sigrid Jusélius \nFoundation, grant number 240053. \nEthics committee - additional information: Since this is a systematic review, \nInstitutional Review Board approval was not necessa ry. \nAuthor Disclosures:  \nBernd Pape: Nothing to disclose \nJussi Hirvonen: Nothing to disclose \nUmida Bafoevna Abdullaeva: Nothing to disclose \n \n \nAdvanced MRI Techniques for Evaluation of Sinonasal  Masses: Exploring \nthe Additive Utility \n*M. Saini*, S. Manchanda, A. S. Bhalla, D. Kandasam y; New Delhi/IN \n \nPurpose or Learning Objective: Sinonasal area is affected by a wide \nspectrum of benign and malignant tumours presenting  with nonspecific \nsymptoms and differentiation solely based on conven tional magnetic \nresonance imaging has only limited specificity \nTo evaluate the role of newer imaging techniques li ke intravoxel incoherent \nmotion (IVIM), diffusion kurtosis imaging (DKI) and  dynamic contrast enhanced \nMRI (DCE-MRI) in differentiating benign and maligna nt sinonasal masses \nMethods or Background: A prospective study was performed on 30 patients \nwith sinonasal masses (18 malignant and 12 benign) who underwent routine \nMRI, DWI, IVIM and DCE MRI. Apparent diffusion coef ficient (ADC) from \ndiffusion weighted imaging, true diffusion coeffici ent (Dt), Pseudodiffusion \ncoefficient (Dx), perfusion fraction (f) from IVIM,  apparent kurtosis coefficient \n(Kapp) and apparent diffusion coefficient (Dapp) fr om DKI, semiquantitative \nand quantitative perfusion parameters from DCEMRI w ere measured and \ncompared between two groups. \nResults or Findings: The mean ADC, Dt and Dapp values were significantly  \nlower in malignant sinonasal lesions than in benign  sinonasal lesions with p \nvalues of 0.000, 0.015 and 0.030 respectively. The mean Kapp value was \nhigher in malignant lesions than in benign lesions (p value of 0.001). There \nwas no significant difference Dx, f and in semiquan titative and quantitative \nperfusion parameters. \nConclusion: The mean ADC derived from the DWI, Dt derived from IVIM, and \nDapp & Kapp derived from the DKI can be used as a n on-invasive method to \ndifferentiate benign and malignant sinonasal masses . Among these, ADC is \nthe best parameter to differentiate, however there is no incremental role of DKI \nand IVIM over conventional DWI. The perfusion param eters showed no \nsignificant difference \nLimitations: The sample size was small and a heterogeneous group  of \npathologies were included in the final analysis. In  addition, the b values were \nset arbitrarily in the IVIM-DKI sequence. \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: Study was preapproved by the \nInstitute Review Board (IRB) [Ref No: IECPG-487/25. 08.2021]. \nAuthor Disclosures:  \nManish Saini: Nothing to disclose \nAshu Seith Bhalla: Nothing to disclose \nSmita Manchanda: Nothing to disclose \nDevasenathipathy Kandasamy: Nothing to disclose \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 89  \nPresence of bone exposure to the aerodigestive trac t as an important \nimaging feature in patients with skull base osteomy elitis following prior \nirradiation for nasopharyngeal carcinoma \n*H. S. Leung*, K. K. F. Tsoi, Q-Y. H. Ai, A. D. Kin g; Hong Kong/HK \n \nPurpose or Learning Objective: Osteoradionecrosis is one the long-term \ncomplications associated with radiotherapy for naso pharyngeal carcinoma \n(NPC), and may result in the significant complicati ons particularly skull base \nosteomyelitis (SBOM). A limited number of clinical case series have reported \non SBOM and their predictors, imaging features and associations with other \nradiation-induced complications remain poorly under stood. This study is to \nevaluate the imaging factors associated with SBOM i n patients with previous \nirradiation for NPC. \nMethods or Background: This is a retrospective matched case-control study,  \nof patients with clinically proven SBOM and compute r tomography (CT) scan \nperformed at diagnosis selected as cases, while con trols were selected within \nNPC post-RT patients and without SBOM, matched by i nitial staging and time \nfrom initial treatment. CT studies were reviewed fo r the presence of bone \nexposure to aerodigestive tract, bony sclerosis, bo ne loss and dehiscence and \nabscess formation; baseline demographics and clinic al outcomes were \nanalyzed by logistic regression and survival analys es. \nResults or Findings: A total of 31 SBOM cases and 31 controls were \nanalysed. Presence of bone exposure to aerodigestiv e tract is the only \nindependent factor associated with SBOM (p<0.001 by  McNemar’s test), while \nthe degree of bone loss also shows borderline signi ficance (p=0.052 by \nWilcoxon sign rank test). Bone exposure to upper ae rodigestive tract remains \nsignificant upon regression controlling for bone lo ss and staging of initial NPC. \nSBOM patients had worse survival with a higher inci dence of other RT-related \ncomplications including carotid occlusion, blowout or RT-induced malignancy. \nConclusion: The presence of bone exposure to upper aerodigestiv e tract is an \nindependent factor associated with SBOM, which coul d be helpful in early \nidentification and treatment to avoid complications  of SBOM which adversely \naffects survival. \nLimitations: Recall bias from retrospective study. \nFunding for this study: Nil \nEthics committee - additional information: This study has been approved by \nJoint CUHK-NTEC Clinical Research Ethics Committee;  The Chinese \nUniversity of Hong Kong (Reference number: CREC 202 4.306) \nAuthor Disclosures:  \nKelvin Kam Fai Tsoi: Nothing to disclose \nAnn Dorothy King: Nothing to disclose \nHo Sang Leung: Nothing to disclose \nQi-Yong Hemis Ai: Nothing to disclose \n \n \nCan HU analysis by used for jaw lesions differentia tion? \n*C. Nadler*, Y. Pakanaev-Levi, H. Rushinek, N. Yavn ai, Y. Zadik, I. Zeevi; \nJerusalem/IL \n(Nadler@hadassah.org.il) \n \nPurpose or Learning Objective: Bone lesions demonstrated on Computed \nTomography (CT) images may be differentiated by sev eral methods including \nHounsfield Unit (HU) analysis. However, the routine  use of this method in pre-\noperative assessment of jaw lesions remains underex plored. We aimed to \ndistinguish using HU analysis 3 types of jaw lesion s. \nMethods or Background: We retrospectively included pre-operative \nMultidetector CT (MDCT) scans of patients with unil ocular hypodense jaw \nlesions, with histologically proven, non-inflame ei ther odontogenic keratocyst \nOKC, central giant cell granuloma CGCG or unicystic  ameloblastoma UA. \nDemographic data and anonymized DICOM files were re trieved. Two \nobservers, blinded to the lesions’ diagnosis, measu red three HU values for \neach lesion, in axial slices, on Philips IntelliSpa ce Portal software. Statistical \nanalysis included intra and inter-observer reliabil ity and validity evaluations as \nwell as comparisons of mean HU values between the d ifferent lesions and \nbetween the same lesion in different jaws \nResults or Findings: Our cohort included 30 jaw lesions (17 OKCs, 8 CGCG s \nand 5 UAs). Mean HU values for OKCs, CGCGs and UAs were 27.99±13.8, \n70.68±46.3 and 31.38±7.4, respectively. Statistically significant difference was \nfound between mean HU values of OKC, CGCG and UA (P =0.035). Following \nadditional pooled analysis, mean HU values of CGCG was statistically higher \nthan OKC and UA (<0.001). No statistically differen ce was found between HU \nvalues of different lesions in different jaws. \nConclusion: Pre-operative differentiation using HU analysis may  be used to \ndiagnose CGCG from OKC and UA. Future multi-center studies with additional \ntypes of lesions are needed to substantiate our res ults \nLimitations: Our limitation included: small sample size, as is a  result of \nstringent inclusion criteria and the transition of referrals from MDCT to CBCT \nand the fact that all cases were from a single medi cal center. \nFunding for this study: None \n \n \nEthics committee - additional information: The study protocol was approved \nby the Institutional Review Board \nAuthor Disclosures:  \nYehuda Zadik: Nothing to disclose \nYehuda Pakanaev-Levi: Nothing to disclose \nItai Zeevi: Nothing to disclose \nNirit Yavnai: Nothing to disclose \nChen Nadler: Nothing to disclose \nHeli Rushinek: Nothing to disclose \n \n \nMultiparametric magnetic resonance imaging in deter mining disease \nactivity of thyroid-associated ophthalmopathy: Adde d value from \ndynamic contrast-enhanced and diffusion-weighted im aging \n*X-Y. Pu*, H. Hu, J. Zhou, L. Jinling, X-Q. Xu, F-Y . Wu; Nanjing/CN \n(pxyxxzy@163.com) \n \nPurpose or Learning Objective: To evaluate the performance of dynamic \ncontrast-enhanced MRI (DCE-MRI) and diffusion-weigh ted imaging (DWI) in \ndetermining disease activity of thyroid-associated ophthalmopathy (TAO) and \nto establish their additional value for staging TAO  compared to conventional \nT2-weighted imaging (T2WI). \nMethods or Background: Seventy-two patients with TAO (48 active, 96 eyes; \n24 inactive, 48 eyes) who underwent DCE, DWI and T2 WI with fat suppression \nwere prospectively enrolled. Simplified histogram p arameters (mean, max, min) \nof DCE-MRI derived parameters (Ktrans, Kep, Ve), ap parent diffusion \ncoefficient (ADC) and signal intensity ratio (SIR) at extraocular muscles were \ncalculated for each orbit and compared between the active and inactive \ngroups. Multivariate analyses were used to identify  independent predictors. \nReceiver operating characteristic curves analyses a nd DeLong tests were \nperformed to evaluate and compare the performances of the identified \nsignificant imaging parameters and their combinatio ns. \nResults or Findings: Active TAO patients showed significantly higher mea n \nand maximum Ve, higher minimum, mean and maximum AD C, higher \nminimum, mean and maximum SIR than inactive patient s (P < 0.05). The mean \nVe, mean ADC and mean SIR were found to be independ ent predictors for \nactive TAO (all P < 0.05). Combination of mean Ve, mean ADC and mean SIR \noutperformed mean SIR alone in staging TAO (AUC, 0. 839 vs 0.769, P = \n0.016). \nConclusion: DCE-MRI and DWI could help to determine the disease  activity of \nTAO. DCE-MRI-derived Ve and DWI-derived ADC values could provide added \nvalue to conventional T2WI-derived SIR in staging T AO \nLimitations: First, the sample size was relatively small. More r esearch should \nbe done to increase the number of datasets to verif y our findings. Second, only \nmultiparametric MRI-derived simplified histogram me trics were analysed. \nFuture studies integrating radiomics and machine le arning have the potential to \nfurther improve staging performance. \nFunding for this study: Jiangsu Province Hospital (the First Affiliated Hos pital \nwith Nanjing Medical University) Clinical Capacity Enhancement Project \n(JSPH-MC-2021-8 to Xiao-Quan Xu) , Jiangsu Province  Capability \nImprovement Project through Science, Technology and  Education \n(JSDW202243 to Fei-Yun Wu) and National Natural Sci ence Foundation of \nChina (NSFC) (81801659 to Hao Hu), \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board of the First Affiliated Hospital of Nanjing Medical \nUniversity (IRB No.2021-SRFA-024) \nAuthor Disclosures:  \nFei-Yun Wu: Nothing to disclose \nJiang Zhou: Nothing to disclose \nLu Jinling: Nothing to disclose \nXiao-Quan Xu: Nothing to disclose \nHao Hu: Nothing to disclose \nXiong-Ying Pu: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 90  \n10:00-11:00 Research Stage 2 \nResearch Presentation Session: \nMusculoskeletal \nRPS 810 \nImaging in metabolic and inflammatory \narthropathies \n \nModerator \nT. Diekhoff; Berlin/DE  \n(torsten.diekhoff@charite.de) \nAuthor Disclosures:  \nTorsten Diekhoff: Advisory Board: Eli Lilly, AbbVie , UCB; Speaker: Novartis, \nMSD, UCB, Janssen, Eli Lilly, Canon MS, Berlinflame , Bracco \n \n \nEvaluation of contrast-enhanced ultrasound for rheu matoid arthritis \nactivity in patients who do not respond to second-l ine biologic therapy \ncompared with superb microvascular imaging: first r esults \n*S. Lavalle*¹, A. Montana², Y. Dal Bosco², F. Aiell o³, R. Foti², G. Privitera²,  \nR. Foti², P. Romeo²; ¹Milan/IT, ²Catania/IT, ³Enna/ IT \n(slavalle6@gmail.com) \n \nPurpose or Learning Objective: Detection of synovitis is essential for \nassessing rheumatoid arthritis (RA) activity and ch anging the therapy. This \nstudy aim to evaluate the level of agreement and co rrelation between DAS 28 \n(Disease activity score 28) and contrast-enhanced u ltrasound (CEUS) and \nSuperb microvascular imaging (SMI) in the classific ation of disease severity \nindex in patients with RA who did not respond to se cond-line biologic therapy. \nMethods or Background: SMI and CEUS were applied to 25 patients with \nactive RA not respond to second-line biologic thera py. We evaluate the \nradiocarpal joint of both wrists. Differences in po sitive synovial vascularity (SV) \nand its semi-quantitative scale were observed, and the correlations of SMI and \nCEUS results with DAS-28. \nResults or Findings: The results indicate that CEUS method shows high-\nmoderate agreement with DAS 28 clinical method (Kap pa = 0.406), 95% CI \n(0.1916, 0.5854), p = 0.00584, while SMI has weaker  agreement (Kappa = \n0.121) 95% CI (0.0098, 0.2466), p = 0.0846 (Kappa =  0.121). The correlation \nbetween CEUS and SMI is very strong (ρ = 0.828), CI 95% (0.6911, 0.9044), \nsuggesting that the two radiological methods tend t o produce very similar \nclassifications, although CEUS is more in line with  the gold standard. The \nWilcoxon signed-rank test showed significant differ ences between DAS 28 and \neach of the two radiological methods, with CEUS app earing closer to the \nclinical method \nConclusion: Use of CEUS to detect vessels in the synovium and v isualization \nof local SV is the method that most correlates with  disease severity in relation \nto DAS 28 in patients with synovial arthritis who d o not respond to second-line \nbiologic therapy compared with SMI. \nLimitations: Small sample size and the need for larger multicent er studies to \nconfirm our findings. \nFunding for this study: No funding \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nPlacido Romeo: Nothing to disclose \nSalvatore Lavalle: Nothing to disclose \nRoberta Foti: Nothing to disclose \nFabio Aiello: Nothing to disclose \nRosario Foti: Nothing to disclose \nAngelo Montana: Nothing to disclose \nGiambattista Privitera: Nothing to disclose \nYlenia Dal Bosco: Nothing to disclose \n \n \nRole of imaging in inflammatory hand arthritis with  diagnostic ambiguity: \nhow complementary MRI findings in clinically establ ished DIP arthritis \nmay facilitate the specific diagnosis \n*Y. Yaraşir*, G. Ayan, H. Avci, L. Kılıç, Ü. Aydingöz, U. Ka lyoncu, A. E. Yildiz; \nAnkara/TR \n(yasinyarasirmd@gmail.com) \n \nPurpose or Learning Objective: Ascertaining whether synovium or \nsynovioentheseal complex (SEC) is predominantly inv olved and pattern of any \ndegeneration would help radiologists in distinguish ing inflammatory hand \narthritis (IHA). We aimed to characterize the role of MRI in reaching specific \ndiagnosis in IHA. \nMethods or Background: Patients aged ≥18 years with suspicious IHA in at \nleast one joint (but not treated other than with NS AIDs) were consecutively \nenrolled in this prospective study. 3T-MRI with a f ine-tuned protocol was \nutilized, whereby differential diagnoses were made according to the \npredominant involvement of synovium or SEC, and/or specific degenerative \nfindings. Physical examination, laboratory and imag ing findings, treatment \nresponse, and already-established rheumatological c lassification criteria were \nused to reach final diagnosis. \nResults or Findings: Of 80 patients initially enrolled, 57 (42 females; mean \nage, 54 [range, 28-79]) constituted the final group  with eventual clinical \ndiagnoses of 11 psoriatic arthritis (PsA), 14 rheum atoid arthritis (RA), 11 \nerosive osteoarthritis and/or calcium pyrophosphate  dihydrate deposition \ndisease, 21 arthritis with distal interphalangeal j oint involvement (ADIPI) not \notherwise classified into any group. MRI revealed n o difference between PsA \nand ADIPI groups, except for nail-bed enthesitis. C omparison between PsA \nand RA disclosed that enthesitis (p=0.033) and peri articular soft tissue edema \n(p=0.042) were more frequent in PsA. When ADIPI and  PsA groups were \ncombined, enthesitis and periarticular soft tissue edema were more common \nthan in other groups (p<0.001). Those with enthesit is were 24 times more likely \nto be in the PsA+ADIPI group than those without ent hesitis (95% CI: 2.6–63.3). \nAccurate classification rate of the model was 83.7% , and area under the curve \n(AUC) value was 0.81. \nConclusion: SEC inflammation and periarticular edema on MRI are  strong \npredictors of PsA, especially in patients with DIP arthritis who don’t meet \nrheumatological classification criteria. \nLimitations: Small sample size \nFunding for this study: Funding was provided by Hacettepe University \nScientific Research Projects Coordination Unit \nEthics committee - additional information: Our study was approved by \nHacettepe University Clinical Studies Ethics Commit tee (2021/23-20) \nAuthor Disclosures:  \nUmut Kalyoncu: Nothing to disclose \nGizem Ayan: Nothing to disclose \nLevent Kılıç: Nothing to disclose \nAdalet Elcin Yildiz: Nothing to disclose \nYasin Yaraşır: Nothing to disclose \nÜstün Aydingöz: Nothing to disclose \nHanife Avci: Nothing to disclose \n \n \nHemosiderin Quantification in Hemophilic Arthropath y of the Knee using \nQuantitative Magnetic Resonance Imaging \n*S. Sedaghat*¹, P. Leutz-Schmidt¹, J. Park², E. Fu² , H. Jang²; ¹Heidelberg/DE, \n²Davis/US \n(samsedaghat1@gmail.com) \n \nPurpose or Learning Objective: This study aims to establish quantitative \nmagnetic resonance imaging (qMRI) as a precise, non invasive tool for \nevaluating hemosiderin deposition in hemophilic art hropathy (HA) of the knee. \nMethods or Background: This prospective study included nine ex-vivo knee \nsynovial tissues from HA patients and the same tiss ues from healthy controls. \nAll tissues underwent standardized qMRI protocols u sing quantitative \nsusceptibility mapping (QSM), based on ultrashort e cho time MRI, which was \noptimized to detect and quantify hemosiderin deposi ts. Also, standard MRI \nsequences were employed. The HA tissues were proces sed histologically \nusing Perl’s Prussian Blue (PPB) staining to identi fy iron contents. Several \nregions of interest were drawn in each tissue. Usin g specialized algorithms, \nvoxel-wise magnetic susceptibility was calculated t o assess iron deposition \nwithin the knee tissues objectively. \nResults or Findings: qMRI demonstrated high sensitivity in detecting and  \nquantifying hemosiderin deposition, whereas convent ional imaging showed no \nabnormalities. The estimated susceptibility values (ESVs) showed significant \ndifferences between HA and control samples. HA tiss ues presented a mean \nESV of 0.48 ± 1.08 ppm and control tissues of 0.13 ± 0.12 ppm (p<0.05). A \nsignificant linear correlation was found between th e iron level quantified by \nhistology and the ESV estimated by QSM (R = 0.908, p < 0.01). There was a \nsignificant difference in the susceptibility in hig h load (HL) tissues compared to \nlow load (LL) tissues (ESV = 5.57 ± 1.23 ppm for HL vs. 0.57 ± 0.85 ppm for \nLL, p<0.001). \nConclusion: This study establishes qMRI, particularly QSM, as a  noninvasive \nand highly sensitive technique for quantifying hemo siderin in HA of the knee. \nBy providing an objective measure of hemosiderin de position, qMRI offers \npotential as a tool for early diagnosis and disease  monitoring in patients with \nhemophilic arthropathy. \nLimitations: Main limitation: ex-vivo study design. \nFunding for this study: This study was funded by the National Institutes of  \nHealth (NIH R01AR078877) and the Deutsche Forschung sgemeinschaft (DFG \nSE 3272/1-1) \nEthics committee - additional information: The Institutional Review Board \n(IRB) of the University of California San Diego app roved the study. \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 91  \nAuthor Disclosures:  \nJinil Park: Nothing to disclose \nSam Sedaghat: Nothing to disclose \nHyungseok Jang: Nothing to disclose \nEddie Fu: Nothing to disclose \nPatricia Leutz-Schmidt: Nothing to disclose \n \n \nDiabetes-related foot disease: the added value of z te \nM. Di Diego, D. Perla, A. Infante, A. M. Costantini , M. L. Angeli, C. Gullì,  \n*G. Ferrara*; Rome/IT \n(giuseppe.ferrara07@icatt.it) \n \nPurpose or Learning Objective: The primary endpoint is to evaluate the \nadditional diagnostic information obtained from ZTE  sequences added to the \nstandard MRI protocol (particularly compared to T1- weighted images) for \nimproved assessment of bone structures in diabetes- related foot disease. \nSecondary endpoints include assessing accuracy comp ared to CT in the \nevaluation of bone alterations and investigating in terobserver agreement \nbetween three musculoskeletal radiologists with dif ferent expertise. \nMethods or Background: This retrospective single-center study analyzed 32 \nMRIs with ZTE sequences from 31 patients (22 males,  10 females; age range: \n49-87 years) from March 2024 to September 2024. Inc lusion criteria included \npatients >18 years old with a confirmed diagnosis o f diabetes mellitus. In 11 \ncases, comparison between ZTE sequences and CT bone  imaging was \npossible. \nResults or Findings: ZTE sequences compared to standard MRI protocol \n(particularly to T1-weighted images) were superior in identifying soft tissue air \n(43% vs 31%), bone pneumatosis (19% vs 9%), bone er osions (77% vs 67%), \nbone exposure (20% vs 15%), bone sclerosis (49% vs 41%), periosteal \nreaction (19% vs 17%), and bone fragments (39% vs 2 5%). Compared to CT, \nZTE demonstrated high sensitivity (75-100%) and spe cificity (92-100%) for all \nmusculoskeletal alterations analyzed. Interobserver  agreement between \nmusculoskeletal radiologists was excellent (k-range  0.82). \nConclusion: ZTE sequences provided additional musculoskeletal i nformation \ncompared to T1-weighted MRI sequences, particularly  for the morphological \nevaluation of bones affected by diabetes-related fo ot disease. This is crucial \nwhen hypointensity on T1-weighted images reduces th e ability to visualize \nbone structures and their alterations. Moreover, th e high sensitivity and \nspecificity values compared to CT suggest that ZTE is a valid alternative. The \ninterobserver agreement for the qualitative evaluat ion of ZTE sequences was \nexcellent, indicating the ease of interpretation. \nLimitations: Small sample size. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nAmato Infante: Nothing to disclose \nConsolato Gullì: Nothing to disclose \nMaria Luigia Angeli: Nothing to disclose \nDaniele Perla: Nothing to disclose \nGiuseppe Ferrara: Nothing to disclose \nMario Di Diego: Nothing to disclose \nAlessandro Maria Costantini: Nothing to disclose \n \n \nThe assessment of rheumatoid arthritis and other ar thropathies through \npower Doppler and superb microvascular imaging: is there any \ndifference? \n*P. Del Nido Recio*, J. D. Aquerreta, A. Paternain Nuin,  \nM. R. López De La Torre Carretero, M. Jiménez Vázqu ez, C. Mbongo, \nC. Urtasun Iriarte, D. A. Zambrano, M. B. Barrio Pi queras; Pamplona/ES \n(pdelnidor@unav.es) \n \nPurpose or Learning Objective: This study aims to analyze whether the \nupgrade Doppler activity, when comparing SMI and PD , is significantly different \nin Rheumatoid Arthritis (RA), compared to other art hropathies. \nMethods or Background: Between May 2023 and April 2024, we \nprospectively analyzed a cohort of 57 joints of 21 different patients. Most of \nthem were previously diagnosed with RA and other ar thropathies, such as \nOsteoarthritis, Psoriatic Arthritis, Gout, etc. PD and SMI imaging were obtained \nin all joints and the individual grades for Doppler  Activity were registered for \neach joint with active synovitis, raging from 0 to 3, according to the EULAR - \nOMERACT US Score. Mann-Whitney U test was applied t o calculate means in \nindependent samples. Two-tailed p-values of <0.05 w ere considered \nstatistically significant. \n \n \n \n \n \n \nResults or Findings: 21 joints of 8 patients with RA and 36 joints of 13  \npatients with other arthropathies were studied. Met acarpophalangeal joints \nwere the most frequently analyzed in the RA group ( 11), and interphalangeal \njoints were the most frequently analyzed in the oth er group (18). When \ncomparing Doppler activity, the mean PD was signifi cantly higher in the RA \ngroup compared to the other group (1.14 vs. 0.75, p  = 0.043). When upgrading \nthe Doppler activity with SMI, we did not find stat istically significant differences \n(2.52 vs 2.69, p = 1.04). Nevertheless, when compar ing the mean of the \ndifference between SMI and PD, it was significantly  lower in the RA group \n(1.38 vs. 1.94, p = 0.016). \nConclusion: Our results indicate that a higher upgrade between PD and SMI \nexams is more likely seen in arthropathies such as oligoarthritis, Psoriatic \nArthritis or Gout, rather than in RA. \nLimitations: The number of patients. Interobserver variability. The degree of \narthritis and type of treatment. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study received institutional \nreview board approval and written informed consent was obtained from all \nparticipants \nAuthor Disclosures:  \nPablo Del Nido Recio: Nothing to disclose \nCarmen Mbongo: Nothing to disclose \nCesar Urtasun Iriarte: Nothing to disclose \nJesús Dámaso Aquerreta: Nothing to disclose \nMiguel Barrio Barrio Piqueras: Nothing to disclose \nManuel Rafael López De La Torre Carretero: Nothing to disclose \nDaniel Alfonso Zambrano: Nothing to disclose \nAlberto Paternain Nuin: Nothing to disclose \nMarcos Jiménez Vázquez: Nothing to disclose \n \n \nPrevalence of diffuse idiopathic skeletal hyperosto sis (DISH) according \nto recently established imaging criteria and coexis tent pelvic \nenthesophytes \n*V. Yaman*, A. E. Yıldız, B. Fırlatan, H. Avcı, O. Karadağ, U. Kalyoncu,  \nÜ. Aydingöz; Ankara/TR \n(drvedatyaman@gmail.com) \n \nPurpose or Learning Objective: Recent (2019) criteria that supplement \nclassic DISH criteria were developed to diagnose th is condition at an earlier \nstage. The aims of the present study were to invest igate the prevalence of the \nearly-stage (ES-DISH) as well as classic (late-stag e) (LS-DISH) disease and to \ndetermine pelvic enthesophyte status in patients wi th ES-DISH and LS-DISH. \nMethods or Background: 636 consecutive patients aged ≥18 years who \nunderwent thorax CT during October 2023 in a tertia ry medical center were \nretrospectively evaluated. CTs were scored accordin g to 2019 DISH criteria by \ntwo independent observers. Pelvic (including parasa croiliac) enthesophytes in \npatients who also had a simultaneous abdomen CT wer e examined by a \nmusculoskeletal radiologist blinded to patients’ DI SH status as well as clinical \ninformation, yielding a personal “pelvic enthesophy te load score” (PELS). \nMalignancy and metabolic syndrome data from patient s were also analyzed. \nResults or Findings: Prevalences of ES-DISH and LS-DISH were 16.7% and \n15.4%, respectively. Intra- and interobserver agree ment were “almost perfect” \n(ICC = 0.88; 95% CI, 0.77–0.99) and “substantial” ( ICC = 0.73; 95% CI, 0.68–\n0.78), respectively. Mean ages (and age range) of “ no DISH”, ES-DISH, and \nLS-DISH subgroups were 55.1 (18–86), 65.5 (46–90), and 69.2 (48–96), \nrespectively. DISH prevalence was similar in patien ts without and with cancer. \nWhen age was factored in, frequency of metabolic sy ndrome in DISH patients \ndid not differ significantly compared to those with out DISH. Mean±SD PELS \nwere 6.2±3.5, 8.9±3.4 and 10.1±3.2 in “no DISH”, ES-DISH and LS-DISH \ngroups. Logistic regression analysis yielded an ove rall accuracy of 68% for \nPELS in predicting DISH status. \nConclusion: This study demonstrated that prevalence of DISH dou bled when \nES-DISH criteria were used. Pelvic enthesophytes ar e more profusely seen in \nES- or LS-DISH than in patients without DISH. \nLimitations: Retrospective study \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was approved by \nHacettepe University Clinical Studies Ethics Commit tee (SBA 24/166). \nAuthor Disclosures:  \nUmut Kalyoncu: Nothing to disclose \nHanife Avcı: Nothing to disclose \nBüşra Fırlatan: Nothing to disclose \nÜstün Aydingöz: Nothing to disclose \nOmer Karadağ: Nothing to disclose \nVedat Yaman: Nothing to disclose \nAdalet Elçin Yıldız: Nothing to disclose \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 92  \n10:00-11:00 Research Stage 3 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 805 \nArtificial intelligence in neuroimaging \n \nModerator \nM. M. Serra; Avignon/FR  \n(mechyserra@gmail.com) \nAuthor Disclosures:  \nMaria Mercedes Serra: Employee: BC Platforms; Resea rch Grant/Support: \nSanofi; Share Holder: BC Platforms \n \n \nImpact of data quality variations caused by dose an d image \nreconstruction on AI assessment of intracranial ane urysms \nL. Gölz, A. Laudani, U. Genske, M. Scheel, G. Bohne r, H-C. Bauknecht, \nS. Mutze, *P. Jahnke*; Berlin/DE \n(paul.jahnke@charite.de) \n \nPurpose or Learning Objective: To assess the performance of a commercial \nAI algorithm in detecting intracranial aneurysms wh en scan data quality \nvariations occur due to changes in dose and image r econstruction. \nMethods or Background: Consistency testing of AI performance was \nperformed using a realistic head CT phantom designe d for AI evaluation. The \nphantom simulated a patient with three intracranial  aneurysms located in the \nanterior communicating artery (ACoA), middle cerebr al artery (MCA), and \nbasilar artery (BA). The phantom was repeatedly exa mined at 21 dose levels \n(0.47 to 20.09 mGy) using iterative reconstruction and filtered back projection. \nAneurysm labeling by an FDA-approved and CE-marked AI solution was \nanalyzed. In addition, five neuroradiologists rated  aneurysm visiblity in all \nexaminations. \nResults or Findings: AI detection rates varied by aneurysm type, with \ndetection rates of 74.6% for the ACoA, 92.9% for th e MCA, and 2.4% for the \nBA aneurysm across all examinations. The AI respons e was inconsistent at \ndoses below 8 mGy with iterative reconstruction and  at doses below 7 mGy \nand above 14 mGy with filtered back projection. In contrast, readers \nconsistently reported 100% visibility for all aneur ysms at doses above 2 mGy \nregardless of image reconstruction. \nConclusion: AI approved for managing intracranial aneurysms sho ws \nperformance issues due to variations in data qualit y and requires different data \nquality standards than neuroradiologists. \nLimitations: This prospective study was limited to a single AI a pplication, a \nsingle scanner system, and three intracranial aneur ysms. \nFunding for this study: This work has not received any funding. \nEthics committee - additional information: Ethics committee of the Charité \nAuthor Disclosures:  \nHans-Christian Bauknecht: Nothing to disclose \nPaul Jahnke: Shareholder: PhantomX GmbH Employee: P hantomX GmbH \nUlrich Genske: Nothing to disclose \nGeorg Bohner: Nothing to disclose \nLeonie Gölz: Nothing to disclose \nAngelo Laudani: Nothing to disclose \nMichael Scheel: Shareholder: PhantomX GmbH \nSven Mutze: Nothing to disclose \n \n \nImproving diagnostic precision: a deep learning sys tem for differentiating \nmultiple sclerosis from small vessel disease using standard non-\nenhanced brain MRI scans \nK. Firouznia, *M. Arab Ahmadi*, H. Hashemi, M. Boro omand-Saboor,  \nR. Ghavami Modegh, M. Akhlaghpasand, H. Dashti, M. Gity,  \nM. Mohammadzadeh; Tehran/IR \n(mehran_arabahmadi@yahoo.com) \n \nPurpose or Learning Objective: The diagnosis of Multiple Sclerosis (MS) \nprimarily depends on clinical evaluation, bolstered  by magnetic resonance \nimaging (MRI) interpreted by skilled radiologists.H owever, the typical imaging \ncharacteristics of MS can resemble those of other c entral nervous system \ndisorders.One such condition is Cerebral SVD, which  can complicate the \nradiologist's ability to make a diagnosis.This diff erential diagnosis can be \nparticularly challenging in the early stages of the  disease.The objective of this \nstudy is to create and assess a Computer-Aided Diag nosis (CAD) system \nutilizing brain MRI images to differentiate between  MS and SVD. \n \nMethods or Background: Brain MRI scans were obtained from a 3 Tesla \nscanner for patients diagnosed with MS during acute  attacks and silent \nphases, alongside individuals diagnosed with SVD ba sed on cardiovascular \nrisk factors. MRI sequences included FLAIR,T1, and T2. An expert \nneuroradiologist identified white matter lesions, w hich were segmented using \nartificial intelligence software. The dataset was d ivided into 80% for training, \n10% for validation, and 10% for testing. A neurorad iologist then evaluated the \nAI results against established clinical and imaging  criteria. \nResults or Findings: The study included 80 MS patients with 265 lesions \ncompared to 67 SVD patients with 218 lesions. The A I tool achieved a \nsensitivity of 78.57% and specificity of 93.33% (P- value < 0.05). It also \ndemonstrated a positive predictive value (PPV) of 9 1.67%, a negative \npredictive value (NPV) of 82.35%, balanced accuracy  of 85.95%, and an area \nunder the curve (AUC) of 78.71. \nConclusion: The findings suggest that artificial intelligence c an effectively \ndifferentiate MRI images of MS from those of SVD us ing routine \nsequences.Implementing AI in distinguishing between  MS and SVD lesions \ncould enhance diagnostic accuracy and improve patie nt management in \nclinical practice. \nLimitations: Sample volume and one center study listed as some o f the \nlimitations. \nFunding for this study: None \nEthics committee - additional information: This study was approved by an \ninstitutional ethics committee. \nAuthor Disclosures:  \nMasoumeh Gity: Nothing to disclose \nKavous Firouznia: Nothing to disclose \nMohammadhosein Akhlaghpasand: Nothing to disclose \nHassan Hashemi: Nothing to disclose \nMelika Boroomand-Saboor: Nothing to disclose \nMaryam Mohammadzadeh: Nothing to disclose \nRassa Ghavami Modegh: Nothing to disclose \nHamed Dashti: Nothing to disclose \nMehran Arab Ahmadi: Nothing to disclose \n \n \nPreviously proposed radiomics features for ruptured  intracranial \naneurysm classification: Overview, auto-segmentatio n, and external \nvalidation \n*D. Zhu*, Y. Yang; Wenzhou/CN \n(zhudongqin@wmu.edu.cn) \n \nPurpose or Learning Objective: To automatically segment and extract \nradiomics features of intracranial aneurysms (IAs),  validate existing radiomics \npredictors for ruptured IAs, and construct machine learning (ML) and deep \nlearning (DL) models for classifying ruptured IAs. \nMethods or Background: In this retrospective study, we used data from the \nMIRACLE Cohort, registered with the Chinese Clinica l Trial Registry \n(ChiCTR2400084601). IAs were segmented automaticall y using the DGIS \nmethod. We systematically reviewed studies reportin g radiomics predictors for \nruptured IAs and externally validated those predict ors. We developed five ML \nand DL models for classifying ruptured IAs, employi ng the SHapley Additive \nexPlanations (SHAP) method to enhance model interpr etability. \nResults or Findings: The study included 632 patients with 668 aneurysms,  \ndivided into training (n=593) and external testing (n=75) datasets. The DGIS \nmethod achieved great segmentation accuracy with Di ce coefficients of 0.98 \nand 0.75 in the source and target domains, respecti vely. When comparing \nradiomics features derived from manual and automati c segmentations, the \noriginal_shape_VoxelVolume, MeshVolume, and Surface Area showed the \nhighest stability (with all ICC of >0.9). Upon exte rnal validation of radiomics \npredictors from 12 studies, the AUCs ranged from 0. 59 to 0.71 in the training \ndataset and 0.48 to 0.65 in the external testing da taset. The original_shape_ \nElongation feature emerged as the most frequently u tilized predictor. The \nGradient Boosting and DRE models performed well in classifying ruptured IAs, \nwith AUCs reaching 0.995 and 0.95 in the training d ataset, and 0.85 and 0.80 \nin the external testing dataset, respectively. \nConclusion: This study presents a comprehensive workflow for au tomatic IAs \nrupture risk analysis and an overview of existing r adiomics predictors. After \nexternal validation, certain original shape feature s demonstrated significant \nstability, utility, and predictive power. The ML an d DL models offer a promising \ntool for risk stratification of IAs. \nLimitations: Not applicable. \nFunding for this study: This study was supported by the Wenzhou Major \nProgram of Science and Technology Innovation (Grant  No. ZY2020012) and \nKey Laboratory of Novel Nuclide Technologies on Pre cision Diagnosis and \nTreatment & clinical Transformation of Wenzhou City  (Grant No. \n2023HZSY0012). \nEthics committee - additional information: Ethics Committee of the First \nAffiliated Hospital of Wenzhou Medical University \nAuthor Disclosures:  \nYunjun Yang: Nothing to disclose \nDongqin Zhu: Nothing to disclose \n\n \n \nThursday \nAbstract-based Programme \n \n 93  \nDiagnostic Performance of Neural Network Algorithms  in Skull Fractures \nDetection in CT Scans: A Systematic Review and Meta -Analysis \n*R. Hajibeygi*¹, G. Sharifi¹, M. Fathi¹, A. Bahrami ², R. Eshraghi²,  \nI. Dixe De Oliveira Santo³, A. Mirjafari⁴, J. Chan⁴, L. Tu³; ¹Tehran/IR, \n²Kashan/IR, ³New Haven, CT/US, ⁴Los Angeles, CA/US \n(Ramtin.beygi75@gmail.com) \n \nPurpose or Learning Objective: The potential intricacy of skull fractures as \nwell as the complexity of underlying anatomy poses diagnostic hurdles for \nradiologists evaluating CT scans. The necessity for  automated diagnostic tools \nhas been brought to light by the shortage of radiol ogists and the growing \ndemand for rapid and accurate fracture diagnosis. C onvolutional Neural \nNetworks (CNNs) are a potential new class of medica l imaging technologies \nthat use deep learning (DL) to improve diagnosis ac curacy. The objective of \nthis systematic review and meta-analysis is to asse ss how well CNN models \ndiagnose skull fractures on CT images. \nMethods or Background: PubMed, Scopus, and Web of Science were \nsearched for studies before February 2024 that used  CNN models to detect \nskull fractures on CT scans. Meta-analyses were con ducted for area under the \nreceiver operating characteristic curve (AUC), sens itivity, specificity, and \naccuracy. Egger's and Begg's tests were used to ass ess publication bias. \nResults or Findings: Meta-analysis was performed for 11 studies with 207 98 \npatients. Pooled average AUC for implementing pre-t raining for transfer \nlearning in CNN models within their training model’ s architecture was 0.96 ± \n0.02. The pooled averages of the studies' sensitivi ty and specificity were 1.0 \nand 0.93, respectively. The accuracy was obtained 0 .92 ± 0.04. Studies \nshowed heterogeneity, which was explained by differ ences in model \ntopologies, training models, and validation techniq ues. There was no \nsignificant publication bias detected. \nConclusion: CNN models perform well in identifying skull fractu res on CT \nscans. The results suggest that CNNs have the poten tial to improve diagnostic \naccuracy in the imaging of acute skull trauma. To f urther enhance these \nmodels' practical applicability, future studies cou ld concentrate on the utility of \nDL models in prospective clinical trials. \nLimitations: One of the limitations is lack of homogeneity in CT  image quality \nacross studies. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nReza Eshraghi: Nothing to disclose \nMobina Fathi: Nothing to disclose \nJanine Chan: Nothing to disclose \nGuive Sharifi: Nothing to disclose \nAshkan Bahrami: Nothing to disclose \nIrene Dixe De Oliveira Santo: Nothing to disclose \nArshia Mirjafari: Nothing to disclose \nRamtin Hajibeygi: Nothing to disclose  \nLong Tu: Nothing to disclose \n \n \nAI as a second reader in post-traumatic head CT at Oktoberfest 2024: A \nprospective performance monitoring study \n*M. B. Steinberger*, M. Bock, A. S. Duque, B. F. Ho ppe, J. P. Rudolph,  \nY. Dikhtyar, P. Reidler, W. Flatz, D. Hinzmann, V. Bogner-Flatz, J. Ricke,  \nC. C. Cyran; Munich/DE \n(maria.steinberger@med.uni-muenchen.de) \n \nPurpose or Learning Objective: To prospectively assess the impact of an AI \nalgorithm on radiologists’ diagnostic confidence in  detecting intracranial \nhaemorrhage (ICH) in post-traumatic head CT at Okto berfest 2024. \nMethods or Background: A mobile CT scanner (Somatom go.Top, Siemens \nHealthineers) was operated on-site for triaging pat ients with mild to moderate \ntraumatic head injuries. This prospective study inc luded n=219 patients who \nunderwent head CT. Instant AI analysis was provided  via auto-routing to a fully \nPACS-integrated, GDPR-compliant clinical AI platfor m. Initially, one of 15 \nboard-certified radiologists, alternating in shifts , read the head CT unassisted, \nrating ICH likelihood on a 5-point Likert scale (-2 , “very low”; 2, “very high”). \nAfter submitting this evaluation, algorithm results  were made available for \nreassessment of ICH likelihood. Performance monitor ing of the AI tool was \nimplemented in PACS (Visage Imaging) via Fast Healt hcare Interoperability \nResources (FHIR) pop-up forms. \nResults or Findings: AI support was utilised in 66% (146/222 scans) of t he \nreadings, varying significantly between readers (43 %-100%). At a probability \nthreshold of 0.1, the AI tool correctly identified 6 out of 7 ICH, 139 true \nnegatives, no false positives (sens=0.857, spec=1.0 00, acc=0.993, ppv=1.000, \nnpv=0.993). AI assistance increased radiologists' c onfidence in ruling out ICH \nin 19 cases (-1 to -2) and confirming ICH in two ca ses (1 to 2). In two \nborderline cases, AI aided in excluding ICH (0 to - 1). Overall, diagnostic \nconfidence was significantly higher with AI support  (p<0.001). \nConclusion: AI assistance significantly improved diagnostic con fidence of \nradiologists reading trauma head CT at Oktoberfest 2024, serving as a virtual \nsecond reader in this emergency setting. PACS-integ rated FHIR forms set a \nframework for seamless monitoring of AI performance  and its impact on \ndiagnostic workflow. \nLimitations: In 40 cases, no AI analysis was performed due to in correct \nspecifications or failed auto-routing. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number UID 24-0813. Written informed consent was \nobtained from all participants and the study was re gistered in the German \nClinical Trial Register (DRKS00034969). \nAuthor Disclosures:  \nJan Philipp Rudolph: Nothing to disclose \nBoj Friedrich Hoppe: Nothing to disclose \nYevgeniy Dikhtyar: Nothing to disclose \nClemens C. Cyran: Nothing to disclose \nWilhelm Flatz: Nothing to disclose \nDominik Hinzmann: Nothing to disclose \nMatthias Bock: Nothing to disclose \nPaul Reidler: Nothing to disclose \nMaria Barbara Steinberger: Nothing to disclose  \nVeronika Bogner-Flatz: Nothing to disclose \nAnna Sophie Duque: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nImpact of Defacing Procedures on Brain Age Gap Esti mation \n*V. L. Ivan*¹, J. Caspers¹, M. Vach¹, D. M. Hedderi ch², D. Weiß¹, C. Rubbert¹; \n¹Düsseldorf/DE, ²Munich/DE \n \nPurpose or Learning Objective: Removal of facial features from MRI brain \nscans (“Defacing”) is mandatory from data privacy p erspective. We \ninvestigated the impact of defacing on Brain Age Ga p Estimation (BrainAGE), \nan imaging biomarker used in various research areas  such as atypical aging. \nMethods or Background: A total of 364 Alzheimer’s disease (AD) patients \nand 717 cognitively normal (CN) participants were a nalyzed including \nunaccelerated (AD:n=290; CN:n=386) and accelerated 3DT1 imaging \n(AD:n=203;; CN:n=500). BrainAGE was computed after defacing using either \nafni_refacer, fsl_deface, mri_deface, mri_reface, P yDeface, or spm_deface \nand without defacing. For BrainAGE, gray matter fea tures were extracted using \nCAT12 for SPM12. BrainAGE was calculated as predict ed age minus \nchronological age. A subset of participants (AD:n=7 4, CN:n=84) had within-\nsession repeat imaging available and were processed  without defacing, \nserving as a benchmark for BrainAGE differences. Me an absolute error (MAE), \nand mean squared error (MSE) were calculated. Outli ers due to defacing were \nidentified using Grubbs’s tests. \nResults or Findings: Benchmark analysis found MAE of 1.15 and MSE of \n2.25 for BrainAGE differences between initial and r epeat scans without \ndefacing in CN, and an MAE of 1.43 and MSE of 3.29 for AD. Among defacing \nmethods, PyDeface exhibited the best performance wi th an overall MAE of \n0.33 and MSE of 0.27, showing a mean BrainAGE diffe rence of 0.08±0.52. \nPyDeface also had the fewest outliers (n=99) based on the benchmark criteria. \nGrubbs’s test identified 23 outliers after PyDeface , with 11 found after \nmri_reface and 20 after spm_deface. \nConclusion: Defacing can be employed for data privacy without s ignificantly \naffecting the reliability of BrainAGE as an imaging  biomarker. PyDeface is \nrecommended. \nLimitations: BrainAGE may be affected by defacing, however, in m ost \napproaches this influence is lesser than the variab ility observed in BrainAGE in \nrepeat non-defaced imaging. \nFunding for this study: No \nEthics committee - additional information: No \nAuthor Disclosures:  \nMarius Vach: Nothing to disclose \nJulian Caspers: Nothing to disclose \nChristian Rubbert: Nothing to disclose \nDaniel Weiß: Nothing to disclose \nVivien Lorena Ivan: Nothing to disclose \nDennis M Hedderich: Nothing to disclose \n \n \nIdentification of depression subtypes in Parkinson' s disease patients via \nstructural MRI whole-brain radiomics: an unsupervis ed machine learning \nstudy \n*Z. Shu*; Zhejiang, Hangzhou/CN \n(cooljuty@hotmail.com) \n \nPurpose or Learning Objective: Unsupervised machine learning methods \nbased on whole-brain radiomic analysis were used to  identify subtypes of \ndepression that occur during the progression of Par kinson's disease (PD). \nMethods or Background: Data from 272 PD patients in the PPMI database \nwere used, among which 81 experienced depression in  Parkinson's Disease \n(DPD) during a 5-year follow-up period. Quantitativ e radiomic features were \nextracted from the whole-brain magnetic resonance s tructural images of each \n\n \n \nThursday \nAbstract-based Programme \n \n 94  \npatient, and principal component analysis (PCA) was  used for feature \ndimensionality reduction. All of the cases were cla ssified into different subtypes \nby unsupervised cluster analysis (UCA). The high-ri sk subtypes were selected \nthrough comparative analysis. The high-risk subtype  data were divided into \ntraining subgroups and testing subgroups at a 7:3 r atio. On the basis of the \nclinical characteristics of the training subgroups,  multiple logistic regression \nanalysis was performed to confirm the risk factors for DPD subtypes. The DPD \nsubtypes were subsequently identified on the basis of the risk factors. A \nprediction model was constructed via decision trees , and the diagnostic \naccuracy of the model was evaluated via receiver op erating characteristic \n(ROC) curves. \nResults or Findings: Logistic regression analysis based on high-risk sub type \ngroups revealed that rem, updrs1_score, updrs2_scor e, and ptau were \nindependent predictors of DPD. The prediction model  based on high-risk \nsubgroups had AUC values of 0.853 and 0.81 in the t raining and testing \nsubgroups, sensitivities of 0.765 and 0.786, and sp ecificities of 0.771 and \n0.815, respectively. The AUC, sensitivity, and spec ificity in the non-high-risk \nsubgroup were 0.859, 0.654, and 0.852, respectively . \nConclusion: An UCA based on MRI structural imaging features can  identify \nhigh-risk subtypes of DPD, and the constructed mode l can also predict the \nprogression of DPD well. \nLimitations: This study was designed as a retrospective analysis . \nFunding for this study: The work was supported by the Natural Science \nFoundation of Zhejiang Province of China (LGF22H090 021) \nEthics committee - additional information: The case data used in this study \ncame from the Parkinson's Progression Markers Initi ative (PPMI) \n(http://www.PPMI-info.org) database, and data colle ction was approved by \ninstitutional review board; For ethical review info rmation on the data, please \nrefer to the website. \nAuthor Disclosures:  \nZhenyu Shu: Nothing to disclose \n \n \nA Machine-Learning Model Based on US Radiomics to C lassify Benign \nand Malignant Thyroid Nodules \nA. Guerrisi¹, *V. Dolcetti*¹, L. Miseo¹, A. Valenti ¹, F. Elia¹, G. Del Gaudio¹,  \nF. Raponi², E. David², V. Cantisani¹; ¹Rome/IT, ²Ca tania/IT \n(vincenzodolcetti@gmail.com) \n \nPurpose or Learning Objective: The aim of this work was to develop a \nmachine learning model based on thyroid ultrasound images in order to \nclassify nodules into benign and malignant classes.  Ultrasound and fine needle \nbiopsy are the most reliable diagnostic methods to date, but they have some \nlimitations. Radiomics and machine learning could b e useful to improve \ndiagnosis while reducing invasive procedures. \nMethods or Background: Ultrasound images from 142 subjects were \ncollected: 40 patients belonged to \"malignant\" and 102 to \"benign\" class, \naccording to histological diagnosis (fine-needle as piration). Those images were \nused to train, cross-validate and internal test thr ee different machine learning \nmodels, using the “Trace for Research” software. A robust radiomic approach \nwas applied, and the models (random forests, SVM an d k-NN classifiers) were \nevaluated. Finally, the best model was externally t ested on an additional cohort \nof 21 patients. \nResults or Findings: The best model (ensemble of random forest) showed \nROC-AUC (%) of 85 (majority vote), 83.7** (mean) [8 0.2-87.2], accuracy (%) of \n83, 81.2** [77.1-85.2], sensitivity (%) of 70, 67.5 ** [64.3-70.7], specificity (%) of \n88, 86.5** [82-91], PPV (%) of 70, 66.5** [57.9-75. 1], and NPV (%) of 88, \n87.1** [85.5-88.8] (*p<0.05, **p<0.005) in the inte rnal test cohort. This model \nwas then externally tested, achieving an Accuracy o f 90.5%, a sensitivity of \n100%, a specificity of 86.7%, a PPV of 75% and an N PV of 100%. \nConclusion: The best model could successfully identify all the malignant \nnodes and the consistent majority of benign in exte rnal testing cohort. Further \ninvestigations could be conducted by testing the mo del with images of nodules \nfrom different centers. \nLimitations: Additional external tests should be performed, with  images from \ndifferent ultrasound machines and different healthc are centers to increase \nvariability of target population. \nFunding for this study: This research was supported by Italian Ministry of \nHealth \nEthics committee - additional information: This study was performed in line \nwith the principles 417 of the Declaration of Helsi nki. Approval was granted by \nthe Ethics Committee of IRCCS 418 IFO-Fondazione GB  Bietti (Date: \n23/01/2023, N: 1820/23) \n \n \n \n \n \n \n \n \n \nAuthor Disclosures:  \nAntonino Guerrisi: Nothing to disclose \nGiovanni Del Gaudio: Nothing to disclose \nAlessandro Valenti: Nothing to disclose \nFulvia Elia: Nothing to disclose \nVincenzo Dolcetti: Nothing to disclose \nFlavia Raponi: Nothing to disclose \nLudovica Miseo: Nothing to disclose \nVito Cantisani: Nothing to disclose \nEmanuele David: Nothing to disclose \n \n \n10:00-11:00 Research Stage 4 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 801 \nAdvances in liver imaging \n \nModerator \nJ.-I. Choi; Seoul/KR  \n(dumkycji@gmail.com) \nAuthor Disclosures:  \nJoon-Il Choi: Grant Recipient: Guerbet Korea, Sieme ns Healthineers, \nSamsong Medison, Bracco Korea; Speaker: Bayer Heath care \n \n \nMultiparametric spectral imaging for characterizati on of small \nhypoattenuating liver lesions \n*N. Abou Zeid*, C. Nelles, Z. Gurbanova, N. Große H okamp, T. Persigehl,  \nS. Lennartz; Cologne/DE \n \nPurpose or Learning Objective: To investigate the diagnostic utility of \nspectral reconstructions for determining cystic nat ure of small, hypoattenuating \nliver lesions. \nMethods or Background: Patients with portal venous phase dual-layer dual-\nenergy CT (dlDECT) who were diagnosed with hypoatte nuating liver lesions \nsmaller than one centimeter that were verified as c ysts in corresponding MRI \nexaminations were retrospectively included. ROI-bas ed measurements were \nconducted by two raters in conventional images (CI) , virtual unenhanced \nimages (VUE) and iodine images. CT-based determinat ion of cystic nature of \nthe lesions was conducted using a HU attenuation th reshold of less than 20 \nHU or an iodine concentration threshold of less tha n 0.5 mg/dl, the latter of \nwhich has been reported as the scanner-specific low er limit of iodine detection. \nAccuracy for determining cystic nature was compared  between CI, VUE and \niodine images. \nResults or Findings: 77 patients with 287 small liver cysts were include d. \nMean attenuation in CI for small liver cysts was 20 .8 ± 24.4 HU, and 11.1 ± \n16.8 HU in VUE images. Mean iodine concentration wa s 0.46 ± 0.57 mg/dl. \nUsing the 20 HU threshold in CI resulted in an accu racy of 60.3 % (173/287), \nwhereas the corresponding accuracy using the same t hreshold in VUE images \nwas 76.7 % (220/287). Accuracy solely based on the iodine threshold was 57.5 \n% (165/287). Combining the VUE and iodine threshold  resulted in an accuracy \nof 92.7 % (266/287) for determining cystic nature o f the lesions in dlDECT. \nConclusion: Combining quantitative VUE and iodine measurements using \nestablished thresholds facilitated correctly diagno sing 92.7% of small \nhypoattenuating liver lesions as cysts, compared to  an accuracy of 60.3 % \nwhen using HU measurements in conventional images. This approach may \nhelp reducing correlative imaging and thereby accel erating staging of cancer \npatients. \nLimitations: Retrospective, mono-centric study \nFunding for this study: None to report \nEthics committee - additional information: IRB waiver due to retrospective \nnature of the study. \nAuthor Disclosures:  \nSimon Lennartz: Speaker: Amboss GmbH Author: Amboss  GmbH \nNils Große Hokamp: Research/Grant Support: Philips Speaker: Philips \nSpeaker: Amboss GmbH \nChristian Nelles: Nothing to disclose \nNour Abou Zeid: Nothing to disclose \nZuleykha Gurbanova: Nothing to disclose \nThorsten Persigehl: Nothing to disclose \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 95  \nCould MRI-radiomics predict Liver Metastasis Recurr ence and Overall \nSurvival after surgery or ablation treatment? \n*T. Russo*, A. Belardo, A. Della Corte, D. Santange lo, F. Calabrese, \nM. M. Vincenzi, M. Mori, C. Fiorino, F. De Cobelli;  Milan/IT \n(t.russo.1994@gmail.com) \n \nPurpose or Learning Objective: To investigate the potentials of MRI \nradiomics to predict recurrence (R), hepatic recurr ence (IR), and overall \nsurvival (OS) in a cohort of patients with colorect al liver metastases (CLM) who \nunderwent microwave ablation (MWA) performed alone or in combination with \nsurgical resection. \nMethods or Background: 121 CLM patients with pre-operative Gadoxetic \nacid-MRI treated at our Institute between October 2 015 and December 2022 \nwere analyzed. One observer manually segmented the largest CLM on T2 \nscans. The abdominal aorta at the level of second l umbar vertebrae was used \nfor the z-score normalization of the lesion. Cox mu ltivariate analysis was run to \nestablish a few-features radiomic model (RAD-T2), t o predict recurrences and \ndeath. A bootstrap-based methodology for robust fea ture selection, including \nredundancy filtering, was optimized to select the b est combination of two, \nthree, four features. Correction of the models for optimism was then performed \nby internal bootstrap-based validation. \nResults or Findings: For R, IR and OS the median follow-ups were \nrespectively 12, 13 and 23 months; the number of ev ents were 80, 68 and 34. \nAfter corrections for optimism, the resulting best RAD-T2 models were based \non the combination of 2-3 features; they were able to predict R with C-\nindex=0.65 (p=0.0002), IR (C-index=0.64, p=0.0029) and OS (C-index=0.71, \np=0.0046). As an example, based on the best cut-off  value of the RAD-T2 \nindex, OS at 2 year was 58% and 88% when the cohort  was stratified \naccordingly. \nConclusion: T2-MRI-based radiomic evaluation of CLMs is feasibl e and \npotentially useful for outcome prediction. \nLimitations: The limited number of patients and the retrospectiv e nature of the \nstudy. \nFunding for this study: The limited number of patients and the retrospectiv e \nnature of the study. \nEthics committee - additional information: All procedures were carried out \nin accordance with the Declaration of Helsinki (196 4) and its later \namendments. \nAuthor Disclosures:  \nMonica Maria Vincenzi: Nothing to disclose \nAngelo Della Corte: Nothing to disclose \nClaudio Fiorino: Nothing to disclose \nDomenico Santangelo: Nothing to disclose \nAlfonso Belardo: Nothing to disclose \nMartina Mori: Nothing to disclose \nTommaso Russo: Nothing to disclose \nFrancesca Calabrese: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \n \n \nIntraindividual Comparison of Half-dose Gadopicleno l and Standard Dose \nof Gadobenate Dimeglumine for Abdominal MRI \n*A. Del Gaudio*¹, K. Kalisz¹, D. Kruse¹, F. Ria¹, D . De Santis², L. Lofino¹,  \nD. Marin¹; ¹Durham, NC/US, ²Rome/IT \n(antonella.delgaudio@uniroma1.it) \n \nPurpose or Learning Objective: To intraindividually compare image quality \nand lesion conspicuity of abdominal MRI using gadop iclenol at 0.05 mmol/kg \nand gadobenate dimeglumine (Gd-BOPTA) at 0.1 mmol/k g. \nMethods or Background: From September 2023 to March 2024, consecutive \npatients who had undergone two clinically indicated  abdominal MRIs within 12 \nmonths using gadopiclenol and Gd-BOPTA on the same scanner were \nretrospectively enrolled. One independent radiologi st manually measured the \nsignal intensity of abdominal organs, arterial and venous vessels, and \nabdominal lesions (liver, pancreas, and kidneys) on  unenhanced, late arterial, \nvenous, and equilibrium phases. SNR, CNR, and magni tude of contrast \nenhancement (ΔE) were calculated for all organs and vessels on ea ch \ncontrast-enhanced phase. Percentage enhancement (%E ) was calculated for \nall lesions on contrast-enhanced phases. Subjective  image quality was \nassessed using a 5-point Likert scale, including: o rgans' and vessels' \nenhancement, liver-to-vessels contrast, and overall  image quality. Lesion \ncharacteristics were also evaluated. Statistical an alysis employed paired t- and \nWilcoxon tests. \nResults or Findings: One hundred subjects (64 years ± 14; 55 men) and 21  \nabdominal lesions were included. Compared to Gd-BOP TA, gadopiclenol \nyielded significantly higher CNR and SNR for pancre as, porta, and kidney in \nthe late arterial phase (p ≤ .040). No significant differences in CNR and SNR \nwere observed between gadopiclenol and Gd-BOPTA acr oss all organs in the \nportal venous and equilibrium phases. Gadopiclenol showed significantly \nhigher pancreatic ΔE in all contras-enhanced phases (p ≤ .049) compared to \nGd-BOPTA. The %E of abdominal lesions was comparabl e between \ngadopiclenol and Gd-BOPTA for all contrast-enhanced  phases (p ≥ .100). No \nsignificant differences were observed in readers’ p erception of image quality \nand lesions’ characteristics. \nConclusion: Gadopiclenol at 0.05 mmol/kg yields similar image q uality and \nimproved pancreatic enhancement compared to Gd-BOPT A at 0.1 mmol/kg. \nLimitations: Retrospective study design \nFunding for this study: None \nEthics committee - additional information: Written informed consent was \nwaived and Institutional Review Board approval was obtained \nAuthor Disclosures:  \nDanielle Kruse: Nothing to disclose \nLudovica Lofino: Nothing to disclose \nDomenico De Santis: Nothing to disclose \nDaniele Marin: Nothing to disclose \nAntonella Del Gaudio: Other: Bracco research fellow ship \nFrancesco Ria: Nothing to disclose \nKevin Kalisz: Nothing to disclose \n \n \nPreclinical profile of a new macrocyclic MRI liver agent \n*J. Lohrke*, T. Brumby, S. Herbert, T. Frenzel, G. Jost, M. Berger, H. Pietsch; \nBerlin/DE \n(jessica.lohrke@bayer.com) \n \nPurpose or Learning Objective: The established liver-specific gadolinium-\nbased magnetic resonance imaging (MRI) contrast age nts, gadoxetate \ndisodium and gadobenate dimeglumine are based on li near, DTPA like ligands. \nIn the present study a new early preclinical macroc yclic liver-specific candidate \nwill be presented. \nMethods or Background: The MRI efficiency (r1-relaxivity) of the candidate  \nBAY 3393081 was determined at 37°C 1.41 T in human plasma. The kinetic \ninertness of the complex stability was investigated  using an established zinc \ntransmetallation assay. The in vitro liver cell upt ake was assessed in rat \nhepatocytes and human transfected organic anion tra nsporter protein 1B1 or \n1B3 embryonic kidney cells. Pharmacokinetic paramet ers were evaluated in \nrodent (rats) and non-rodent (dogs) species by anal yzing the gadolinium (Gd) \nconcentrations in plasma over time. The in vivo liv er elimination was examined \nin bile-duct-cannulated rats and the bile was analy zed using inductively \ncoupled plasma mass spectroscopy. Contrast-enhanced  liver MRI was \nperformed in mice, pigs and a VX2 tumor model in ra bbits. \nResults or Findings: The relaxivity of BAY3393081 was determined with \n8.7±0.3 L/(mmol·s) in human plasma at 1.41 T. The k inetic inertness of the \ncomplex stability was comparable to marketed macroc yclic GBCAs. The in vitro \ncell uptake results revealed that BAY3393081 is spe cifically taken up by rat \nand human OATPs. BAY 3393081 showed a high plasma c learance in rat and \ndog. In rats ~80% of the injected dose were elimina ted in an unchanged form \nvia the bile. Strong signal enhancement of liver pa renchyma was demonstrated \nin rats, rabbits and pigs. \nConclusion: The preclinical candidate BAY3393081 showed the hig h kinetic \ninertness of macrocyclic GBCAs and a strong liver p arenchyma enhancement \nin the MRI. \nLimitations: Limited transferability of preclinical data to huma n liver elimination \ndue to significant interspecies variability in hepa tobiliary transporter \nexpression. \nFunding for this study: Funding for this preclinical study was provided by \nBayer AG. \nEthics committee - additional information: All animal experiments were \napproved by LaGeSo. \nAuthor Disclosures:  \nSimon Herbert: Employee: Bayer AG \nMarkus Berger: Employee: Bayer AG \nGregor Jost: Employee: Bayer AG \nThomas Frenzel: Employee: Bayer AG \nJessica Lohrke: Employee: Bayer AG \nThomas Brumby: Employee: Bayer AG \nHubertus Pietsch: Employee: Bayer AG \n \n \nNon-invasive diagnosis of chronic liver disease and  portal hypertension \nusing intravoxel incoherent motion imaging and magn etic resonance \nelastography \n*D. Catucci*, S. U. Von Däniken, V. Obmann, A. Berz igotti, L. Ebner,  \nJ. T. Heverhagen, A. Christe, P. Vermathen, A. T. H uber; Bern/CH \n \nPurpose or Learning Objective: This study aimed to analyse the \nperformance of intravoxel incoherent motion (IVIM) imaging parameters \n(tissue-diffusivity D, perfusion-fraction PF and ps eudo-diffusion-coefficient D*) \nand liver stiffness (LS) measured by magnetic reson ance elastography (MRE) \nto screen for chronic liver disease (CLD) and clini cally significant portal \nhypertension (CSPH) on liver MRI examinations. \nMethods or Background: This prospective study included 103 patients \nwithout CLD (noCLD-group, n=103) and 82 patients wi th biopsy-proven CLD \nwho underwent liver MRI examinations including MRE and IVIM imaging \n\n \n \nThursday \nAbstract-based Programme \n \n 96  \nbetween 03/2016 and 11/2023. Patients with CLD were  subdivided based on \ntheir liver fibrosis degree: early CLD (F0-F1; eCLD -group, n=21), intermediate \nCLD (F2; iCLD-group, n=19), advanced CLD (F3-F4; aC LD-group, n=20) and \naCLD with CSPH according to the BAVENO VII consensu s (aCLDPH-group; \nn=22). IVIM imaging parameters (D, PF and D*) of th e liver as well as LS were \nmeasured in all patients. Statistical analysis incl uded the Kruskal-Wallis test for \ngroup comparison and receiver operating characteris tic (ROC) curve analysis \nwith multiple logistic regression analysis for grou p differentiation. \nResults or Findings: D, PF, D* and LS differed significantly between all  \ngroups (p<0.001). The overall best parameter for de tecting CLD and CSPH \nwas LS with a cut-off value of >3.2 kPa (sensitivit y 67%/specificity 96%) \nrespectively >3.8 kPa (sensitivity 95%/specificity 88%), both with p<0.001. For \nCLD-detection, a combination of D <99 x 10-5 mm2/se c and LS >3.2 kPa \nincreased the AUC from 0.89 with LS to 0.95. For CS PH-detection, a \ncombination of D* <434 x 10-5 mm2/sec and LS >3.8 k Pa increased the \nspecificity (90%) with a slightly lower sensitivity  (91%, p<0.001). \nConclusion: IVIM imaging parameters (D, PF and D*) as well as L S measured \nby MRE allow non-invasive screening for CLD and CSP H. \nLimitations: This was a single center study and should be extern ally validated. \nFunding for this study: This study received funding by the Swiss National \nScience Foundation (SNF), grant number 188591. \nEthics committee - additional information: This study was approved by the \ncantonal ethics committee of Bern (Kantonale Ethikk ommission Bern). \nAuthor Disclosures:  \nSandro Urs Von Däniken: Nothing to disclose \nJohannes T. Heverhagen: Nothing to disclose \nDamiano Catucci: Nothing to disclose \nVerena Obmann: Nothing to disclose \nPeter Vermathen: Nothing to disclose \nLukas Ebner: Nothing to disclose \nAndreas Christe: Nothing to disclose \nAdrian Thomas Huber: Nothing to disclose \nAnnalisa Berzigotti: Nothing to disclose \n \n \nEvaluation of Artificial Intelligence supported Thi rd Harmonic B-mode in \ngallbladder ultrasound \n*P. Spiesecke*, T. Fischer; Berlin/DE \n \nPurpose or Learning Objective: Superharmonic imaging is a useful B-mode \nultrasound technology increasing spatial resolution . Currently, there is a novel \ntechnology available which produces B-mode images b y a combination of \ndifferent harmonics up to third harmonic using an A rtificial Intelligence (AI) \ndriven algorithm. The aim of the present study is t he first evaluation of this \ntechnology. \nMethods or Background: For this prospective study, overall 52 healthy test  \npersons and patients were included. Standard and no vel B-mode images of the \ngallbladder were captured in each subject – in each  case by recording several \ncombinations of additional parameters such as Dynam ic Range and Speckle \nReduction. For this purpose, a premium ultrasound s ystem was used (Canon \nAplio i800 including third-harmonic imaging). The i mages were manually \nsegmented manually and were subjected to computer-a ided analysis to \nanalyze artifacts and edge sharpness. Additionally,  Radiologists in different \nstages of training and subspecialisation rated the images by means of different \nparameters on a Likert scale. \nResults or Findings: N = 26 data sets each with and without gallbladder \npathology were included. In gallbladder B-mode ultr asound, ratings of the Third \nHarmonic Imaging-derived images showed a significan t reduction of artifacts in \nthe gallbladder lumen as well as a higher sharpness  of interfaces. Subjective \nanalysis revealed higher image quality in Third Har monic Imaging compared to \nstandard B-mode. \nConclusion: Our results suggest, that the AI-driven Third Harmo nic Imaging \ncan be a useful tool to increase the sharpness of i nterfaces and reduce the \nartifacts in gallbladder B-mode ultrasound. \nLimitations: The present study is a single center study which fi rst evaluates \nthis novel AI-driven Third Harmonics Imaging B-mode  ultrasound technology. \nFurther studies are necessary to evaluate this tech nology more detailed. \nFunding for this study: None. \nEthics committee - additional information: Local ethics committee. \nAuthor Disclosures:  \nPaul Spiesecke: Nothing to disclose \nThomas Fischer: Nothing to disclose \n \n \n \n12:30-13:30 Research Stage 1 \nResearch Presentation Session: Physics \nin Medical Imaging \nRPS 913 \nStriving for lower radiation dose and better \nimage quality \n \nModerator \nD. Kostova-Lefterova; Sofia/BG  \n(dessi.zvkl@gmail.com) \n \n \nResults from a decade (2012-2021) of periodical pat ient dose surveys for \nCT in Belgium \n*A. S. L. Dedulle*, T. Vanaudenhove, K. Van Slambro uck, A. Fremout; \nBrussels/BE \n(andedulle@gmail.com) \n \nPurpose or Learning Objective: In Belgium, diagnostic reference levels \n(DRLs) are established based on periodical patient dose surveys carried out by \nthe regulatory body. This study evaluates trends in  doses from CT scans, using \ndata from these surveys. \nMethods or Background: From 2012 to 2021, 10 periodical patient dose \nsurveys were conducted in Belgium, covering CT equi pment nationwide, as \nparticipation in the surveys is mandatory. Anonymou s patient dose data were \ncollected for 10 types of CT examinations (abdomen,  chest angiography, \ncoronary angiography, colon, cervical spine, lumbar  spine, skull, sinus, thorax, \nthorax-abdomen). For each type of examination, dose  data (CTDIvol, DLP) \nwere registered for minimal 30 adult patients per C T device. The typical dose \nvalue was calculated (median) for each type of exam ination and each CT \ndevice. DRLs for each type of examination along wit h other statistical \nparameters were derived from the distribution of th ese values. \nResults or Findings: The participation rate exceeded 85% across all 10 \nsurveys. Between 2012 and 2021, the 75th percentile  of the typical DLP-values \nfor complete examinations showed a decrease between  22% and 63%, \ndepending on the type of examination. Additionally,  the data spread narrowed \nbetween 8% and 65%, and the 95th percentiles decrea sed between 16% and \n62%. The DRL for CTDIvol per acquisition reduced be tween 31% and 71%, \nwhile the 95th percentile of the typical CTDIvol-va lues decreased between \n38% and 71%. For most examination types, the larges t decrease in DRL was \nobtained during the first five periodical surveys. \nConclusion: Over the 10-year period, patient doses from CT scan s in Belgium \nsubstantially decreased. This is reflected in both lower DRLs and a reduced \nspread in dose data. This decrease is likely the re sult of improved protocols \nand the introduction of advanced CT technology. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received. \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nAn Saskia Luc Dedulle: Nothing to disclose \nThibault Vanaudenhove: Nothing to disclose \nKatrien Van Slambrouck: Nothing to disclose \nAn Fremout: Nothing to disclose \n \n \nAccounting for imaging dose in Hodgkin’s lymphoma p atients \nundergoing PBS proton therapy and photon VMAT: a SI NFONIA study \n*M. Azizi*¹, M. Romero-Expósito², I. Múñoz², A. Gka vonatsiou², O. Norrlid²,  \nC. Goldkuhl³, D. Molin², I. Toma-Dasu¹, A. Dasu²; ¹ Stockholm/SE, \n²Uppsala/SE, ³Gothenburg/SE \n(mona.azizi@fysik.su.se) \n \nPurpose or Learning Objective: This project aimed to fill a knowledge gap on \nthe magnitude of secondary doses including the out- of-field and the imaging \ndoses contributing to the risks from photon and pro ton radiotherapy. \nMethods or Background: A framework was developed for determining and \nintegrating the imaging and therapy doses for indiv idual determination of total \norgan doses. VirtualDose software [1] was used for dose determinations from \nindividual CT scans, while Monte Carlo simulations were used for CBCT dose \ndeterminations. Synthetic whole body CTs from the i ndividual planning CTs \nwere generated using IS2aR-software [2]. Neutron do ses in proton \nradiotherapy were calculated using MCNP. Out-of-fie ld doses in photon \ntherapy patients were determined with Periphocal3D [3]. \n \n\n \n \nThursday \nAbstract-based Programme \n \n 97  \nResults or Findings: To our knowledge, this was the first systematic \nassessment of total dose administered to patients t hroughout the course of the \nradiotherapy, encompassing clinically relevant freq uency of use of the imaging \nprocedures. The numbers of kV-CBCT ranged from 3 to  17 and 3 to 11 CTs, \nrespectively in photon and proton plans. Imaging do ses contribute 60-570 mSv \nfor photon and 6-200 mSv for proton treatments over  the entire treatment \ncourse for organs close to the target. Distant orga ns like the stomach, bladder, \nand liver showed a 13.5% increase in imaging dose r elative to the photon \ntreatment dose, while PBS indicates a 400% increase  (though with lower \nabsolute doses), indicating its greater relative im pact. \nConclusion: Radiation burden in high precision radiotherapy dep ending on the \nimaging protocols will have to be taken into accoun t in epidemiological studies \non the incidence of second cancers in future patien t cohorts. References: [1] A. \nDing et al., Phys Med Biol; 2015. [2] I. S. Muñoz-H ernández et al., Phys. \nMedica, 2023. [3] B. Sánchez-Nieto, et al., Front. Oncol., 2022. \nLimitations: No limitation was identified. \nFunding for this study: This project has received funding from Euratom’s \nresearch and innovation programme 2019-20 under gra nt agreement no. \n945196. \nEthics committee - additional information: The study is retrospective. \nAuthor Disclosures:  \nMona Azizi: Nothing to disclose \nAlexandru Dasu: Nothing to disclose \nMaite Romero-Expósito: Nothing to disclose \nOla Norrlid: Nothing to disclose \nAngeliki Gkavonatsiou: Nothing to disclose \nIuliana Toma-Dasu: Nothing to disclose \nChristina Goldkuhl: Nothing to disclose \nDaniel Molin: Nothing to disclose \nIsidora Múñoz: Nothing to disclose \n \n \nFetal radiation dose from iodine-125 seeds in pregn ant breast cancer \npatients \n*J. Pluim*¹, J. Van De Kamer², E. Heeling², I. Ploe g², D. Hulsen¹;  \n¹'s-Hertogenbosch/NL, ²Amsterdam/NL \n \nPurpose or Learning Objective: The treatment of breast cancer during \npregnancy (PrBC) requires careful consideration of consequences for both \nmaternal and fetal health. In non-pregnant patients , the use of radioactive \niodine-125 (125I)-seeds is standard practice for lo calising non-palpable breast \ntumors before breast-conserving surgery. However, t he use of 125I-seeds in \npregnant patients has been avoided due to concerns about fetal radiation \nexposure. \nMethods or Background: This study developed a mathematical model to \nestimate the fetal absorbed dose based on several f actors: the radioactivity of \nthe 125I-seed, the duration of implantation, and th e distance between the 125I-\nseed and fetus as a function of maternal anatomy, g estational age, and fetal \ndevelopment. Three scenarios, representing a range of maternal and fetal \nanatomy, were evaluated, including a worst-case sce nario from a radiation \nsafety perspective. \nResults or Findings: The results show that the fetal absorbed dose varie s \nacross the three scenarios, with ranges of 0–1.6 mG y, 0.0–1.0 mGy, and 0.0–\n0.4 mGy, depending on when the 125I-seed was implan ted and when it was \nremoved. These dose ranges are similar to conventio nal diagnostic x-ray \nscans. The maximum calculated absorbed dose (1.6 mG y) is unlikely to be \nreached in practice and is well below the 100 mGy t hreshold associated with \npossible fetal malformations. The associated cancer  risk increase (0.016%) is \nminimal. \nConclusion: The use of 125I-seeds as localisation method of bre ast tumors in \npregnant patients results in low fetal radiation do ses and should not be avoided \ndue to dose concerns. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nEva Heeling: Nothing to disclose \nIris Ploeg: Nothing to disclose \nDennis Hulsen: Nothing to disclose \nJip Pluim: Nothing to disclose \nJeroen Van De Kamer: Nothing to disclose \n \n \nImplementing novel optimization strategies in x-ray  interventional \ncardiology imaging for paediatric examinations with  a simulation \nframework \nR. Massera, *N. W. Marshall*, H. Bosmans; Leuven/BE  \n(nicholas.marshall@uzleuven.be) \n \nPurpose or Learning Objective: To apply novel optimization strategies in the \nsearch for optimal x-ray technique factors in paedi atric interventional \ncardiology examinations. \nMethods or Background: A simulation framework previously developed for \nadult interventional radiology examinations was ada pted to use paediatric \nphantoms. The optimization framework implemented th e Monte Carlo (MC) \ncode PENELOPE(2018)/penEasy(2020) for dose and imag e quality (IQ) \ncalculations, combined with a ray-tracing routine t o calculate the attenuation \nthrough the patient and table. A figure of merit (F OM) defined as \nSDNRw²(u)/Dose was used. SDNRw(u) is a signal-diffe rence-to-noise ratio \nweighted for the impact of geometric blurring from the focal spot and from \nobject motion. This was evaluated for the task of d etecting a 0.36 mm diameter \niron guidewire. Dosimetric quantities comprised inc ident air kerma (AK) at the \nreference point, used to approximate skin dose, and  the effective dose (Deff), \nused to estimate stochastic risk. To calculate Deff  and SDNRw(u), ICRP \nfemale paediatric phantoms of 1- and 5-year-old wer e used. The tube voltage, \nspectral copper filtration and x-ray focus that yie lded the highest FOM value for \na particular dose quantity were found, taking into consideration x-ray tube \nloading limitations. \nResults or Findings: For the 1-year-old and 5-year-old phantoms, optimal  \nFOM values were achieved at respectively 65kV/0.7 m m Cu/micro focus and \n64kV/0.5 mm Cu/small focus, when AK was the cost fu nction. Using effective \ndose as the cost function gave optimal factors of 5 9kV/0.5 mm Cu/micro focus \nand 60kV/0.3 mm Cu, for the 1-year-old and 5-year-o ld cases, respectively. \nConclusion: The framework was successfully adapted to work with  paediatric \nphantoms. Optimization based on effective dose sele cted lower tube voltages \nand copper spectral filtration thicknesses compared  to a typical optimization \nusing incident air kerma. \nLimitations: A limited number of phantoms were used in the simul ations. \nFunding for this study: This study is the result of a research agreement wi th \nSiemens Healthineers. \nEthics committee - additional information: Na \nAuthor Disclosures:  \nNicholas William Marshall: Nothing to disclose \nRodrigo Massera: Nothing to disclose \nHilde Bosmans: Other: Research agreement with Sieme ns Healthineers \n \n \nImage Quality in lung cancer screening LDCT: compar ing the NELSON \ntrial to current conventional and photon-counting t horacic CT \n*K. Torfs*¹, D. Petrov¹, L. D'Hondt², M. Lefere³, K . Bacher², A. Snoeckx⁴,  \nW. De Wever¹, H. Bosmans¹; ¹Leuven/BE, ²Gent/BE, ³B onheiden/BE, \n⁴Zandhoven/BE \n(kwinten.torfs@uzleuven.be) \n \nPurpose or Learning Objective: Current guidelines for lung-cancer-screening \n(LCS) with low-dose chest CT (LDCT) are focused on dose, without specifying \nimage quality (IQ) targets. This study compares noi se and resolution in patient \nscans between the NELSON LCS trial, an ultra-low-do se (ULDCT) LCS study \nand current clinical standard-dose (SDCT) and LDCT on both energy-\nintegrating (EIDCT) and photon-counting CT (PCCT). \nMethods or Background: IQ was measured in 54 patient scans (24-26cm \nwater-equivalent-diameter, sharp, 1mm slice-thickne ss reconstructions) of 6 \nprotocols: LDCT-NELSON (Siemens Sensation 16), SDCT -EIDCT, LDCT-\nEIDCT and ULDCT-EIDCT (Siemens SOMATOM Force) and L DCT-PCCT and \nSDCT-PCCT (Siemens Naeotom Alpha). Noise was comput ed per scan by \naveraging global-noise-levels (GNL) for soft tissue  (0-170HU) from 50 \nequidistant slices. Resolution was quantified using  AUC of the digital \nmodulation-transfer-function (MTF) measured from th e patient skin-air-\ninterface. Protocol averages were presented as: [kV p|reconstruction \nkernel|CTDIvol(mGy)|GNL-soft(HU)|MTF-AUC(mm-1)]. To  assess \nstandardized-condition-protocols, patient-specific influence of pixel-size and \ndose was removed by predicting GNL at 1.6mGy CTDIvo l and measuring \npresampled-MTF. \nResults or Findings: The results can be summarized as follows: LDCT-\nNELSON [120|B50|1.6±0.2mGy|155±8HU|0.52±0.06mm-1] ULDCT-EIDCT \n[Sn100|Br64-IR3|0.16mGy|151±7HU|0.49±0.04mm-1] SDCT-EIDCT \n[120|Br54|5.8±1.6mGy|70±4HU|0.49±0.06mm-1] SDCT-PCCT \n[120|Bl56|4.8±0.6mGy|127±9HU|0.86±0.05mm-1] LDCT-PCCT [Sn100|Bl56-\nIR1|1.08±0.15mGy|129±3HU|0.71±0.06mm-1] Compared to LDCT-NELSON \nscans, noise was significantly lower (p<0.001) in S DCT-EIDCT, SDCT-PCCT \nand LDCT-PCCT and the AUC-MTF significantly sharper  (p<0.001) in SDCT-\nPCCT and LDCT-PCCT. ULDCT had similar noise and res olution properties as \nLDCT-NELSON, at a mean dose of only 0.16mGy versus 1.6mGy. However, \nfor standardized-conditions, LDCT-PCCT, SDCT-EIDCT and ULDCT-EIDCT \nprotocols were inherently less noisy (p<0.01) than NELSON, with SDCT-\nPCCT, LDCT-EIDCT and LDCT-PCCT being significantly sharper (p<0.001). \nConclusion: We have proposed a method to compare IQ of successf ul \nhistorical LCS scans to current state-of-the-art ca ndidates with a dose – image \nquality evaluation from patient CT scans. Taking th e NELSON setting as \nminimal reference, there are several candidate (ult ra)LDCT protocols, with the \nLDCT on PCCT outperforming. \nLimitations: Limited no. cases \nFunding for this study: This work was performed with a grant from Kom op \nTegen Kanker (G0B1922N), a Flemish NGO active in th e fight against cancer \n\n \n \nThursday \nAbstract-based Programme \n \n 98  \nEthics committee - additional information: Study approved under internal \nreference number S68527 \nAuthor Disclosures:  \nMathieu Lefere: Nothing to disclose \nKlaus Bacher: Nothing to disclose \nKwinten Torfs: Nothing to disclose \nLouise D'Hondt: Nothing to disclose \nAnnemiek Snoeckx: Nothing to disclose \nHilde Bosmans: Nothing to disclose \nDimitar Petrov: Nothing to disclose \nWalter De Wever: Nothing to disclose \n \n \nCombining rapid kVp-switching and photon-counting d etectors for high-\nresolution spectral CT imaging at ultra-low doses \n*O. Sandvold*¹, R. Proksa¹, A. Perkins², P. Noël¹; ¹Philadelphia, PA/US, \n²Cleveland, OH/US \n \nPurpose or Learning Objective: This work presents a CT acquisition \nparadigm utilizing sparse spectral imaging to deliv er both high spatial \nresolution and spectral imaging, specifically desig ned for pediatric imaging. \nMethods or Background: Combining spectral imaging with high spatial \nresolution at ultra-low doses is challenging with c urrent technology. In our \nmethod, most of the scan is captured using single l ow tube voltage with the \ndetector operating in non-spectral, high-resolution  mode by combining x-ray \nphotons across energy bins (excluding electronic no ise). During sparse \nintervals, the system switches to rapid kVp mode, l everaging the detector's \nspectral capabilities. Data is continuously acquire d and combined to generate \nboth high-resolution and spectral images. A Monte C arlo simulation \ndemonstrated this pediatric imaging protocol, using  70 kVp with intermittent \n110 kVp pulses for spectral data. The detector pixe l size was set to 0.5x0.5 \nmm², with an additional sampling protocol investiga ted using 1x1 mm² pixels. \nThe simulated phantom represented a 150 mm pediatri c patient. Spectral SNR \nin monoenergetic images was estimated using the Cra mér-Rao Lower Bound, \nand the area under the monoenergetic curve (AUMC) w as calculated as the \ntotal SNR over 35–120 keV. \nResults or Findings: The sparse spectral protocol improved AUMC spectral  \nSNR by 220% compared to a constant 100 kVp photon-c ounting scan using \nthe same dose level and pixel size. Binning pixels to measure 1x1 mm², the \nsparse spectral performance was 475% the 100 kVp re ference scan AUMC. At \n62% of the 100 kVp dose, the sparse protocol AUMC w as 170% greater than \nreference AUMC. \nConclusion: Pediatric spectral CT faces three main challenges: achieving high \nspatial resolution, obtaining low-noise spectral da ta, and minimizing radiation \ndose. Our proposed acquisition method combines mult iple technologies to \naddress these challenges. Future clinical translati on promises improved \npediatric care with minimal radiation exposure. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nOlivia Sandvold: Nothing to disclose \nPeter Noël: Nothing to disclose \nRoland Proksa: Nothing to disclose \nAmy Perkins: Employee: Philips Healthcare \n \n \nA Machine Learning-based method for predicting norm alized glandular \ndose coefficients and associated uncertainty in dig ital mammography \nand digital breast tomosynthesis \n*A. Sarno*¹, R. Massera², G. Paternò³, P. Cardarell i³, N. W. Marshall²,  \nH. Bosmans², K. Bliznakova⁴; ¹Milan/IT, ²Leuven/BE, ³Ferrara/IT, ⁴Varna/BG \n(antonio.sarno@unimi.it) \n \nPurpose or Learning Objective: To investigate the use of a machine learning \nalgorithm and patient-derived digital breast phanto ms for predicting normalized \nglandular dose (DgN) coefficients and factors that influence the DgN \nuncertainty in digital mammography (DM) and digital  breast tomosynthesis \n(DBT). \nMethods or Background: Monte Carlo dosimetry calculations were performed \nfor a set of 126 anatomically realistic digital bre ast phantoms to establish the \nground truth DgN. The DgN was then predicted using a linear regression with \nan Automatic Relevance Determination Regression alg orithm from 5 \nanatomical breast features: compressed breast thick ness, glandular fraction, \ntotal glandular volume, center of mass and standard  deviation of the glandular \ntissue distribution in the cranio-caudal direction.  An algorithm for data \nimputation was explored to account for the cases wh ere the latter two features \nare not available. The regression algorithm was val idated using 5-fold Cross \nValidation. \n \n \nResults or Findings: With the use of all 5 selected anatomical features,  \naverage difference between predicted DgN and the gr ound truth was 1%, with \n50% of cases differing from the ground truth by les s than 3%; estimated \nuncertainty on the DgN values was 9%. Uncertainty o n DgN coefficients \nincreased to 17% when the features related to the g landular distribution were \nexcluded; however, this had only a minor impact on the prediction accuracy. \nThe data imputation algorithm reduced the uncertain ty on the predicted values, \nbut could not match the prediction performance obta ined by using all the \navailable anatomical features. \nConclusion: The proposed methodology predicts the normalized gl andular \ndose in DM and DBT with an error of 1%, on average,  and with an estimated \nuncertainty of only 9%. 50% of the predicted DgN co efficients differed by less \nthan 3% from the ground truth. \nLimitations: Limited to single DM/DBT geometry \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nNicholas William Marshall: Nothing to disclose \nGianfranco Paternò: Nothing to disclose \nKristina Bliznakova: Nothing to disclose \nRodrigo Massera: Nothing to disclose \nHilde Bosmans: Nothing to disclose \nAntonio Sarno: Nothing to disclose \nPaolo Cardarelli: Nothing to disclose \n \n \n12:30-13:30 Research Stage 2 \nResearch Presentation Session: Paediatric \nRPS 912 \nInsights into foetal imaging \n \nModerator \nM. Rebollo Polo; Barcelona/ES  \n(monica.rebollo@sjd.es) \n \n \nValidation of fetal brain 3D slice-to-volume regist ration (SVR) in detecting \nthe cause of antenatal ventriculomegaly confirmed b y neonatal scan \n*W. H. E. Hamed*, G. Kendall, L. Dyet, L. Srinivasa n, D. Peebles, A. David,  \nM. Sokolska, K. P. Baruteau; London/UK \n(weaam.hamed@nhs.net) \n \nPurpose or Learning Objective: Validates 3DSVR for detecting anatomical \nand structural pathologies in fetal MRI and assesse s quality improvement in a \ncohort of antenatal ventriculomegaly confirmed by n eonatal MRI. \nMethods or Background: Detecting subtle anatomical abnormalities in fetal \nbrain MRI is challenging due to motion artefacts an d the limited spatial \nresolution of 2D slices. Recently, slice-to-volume reconstruction (SVR) \nsoftware has been developed to realign multiple 2D stacks into a high-\nresolution 3D volume (3DSVR), enabling better visua lisation through \nmultiplanar reconstruction. However, clinical valid ation of 3D-SVR is limited \ndue to lack of ground truth data. A retrospective c ross-sectional study was \nconducted on pregnancies with ventriculomegaly. Inc lusion criteria included \nfetal and neonatal MRI performed with standard prot ocols and 3DSVR. The \nmedian gestational age at fetal MRI was 28weeks (ra nge 21-33w), and at \nneonatal MRI, 1week (range 1d-4w). Ventriculomegaly  causes were assessed \non fetal 2DT2w-HASTE and 3DSVR and confirmed with 2 DT2w-TSE on \nneonatal scans. Eleven brain structures were scored  on a 3-point visibility \nscale, and image quality was rated based on signal- to-noise ratio (SNR) and \nmotion artefacts. Statistical analysis was performe d using the Wilcoxon signed-\nrank test. \nResults or Findings: Of 20 subjects, eight had aqueduct stenosis identif ied on \nneonatal MRI. This was confirmed in 3/8 on 2D and 8 /8 on 3DSVR. Fetal \n3DSVR improved visibility scores in six of eleven s tructures, with significant \ndifferences in PLIC (0.65vs1.85, p<0.001), Sylvian aqueduct (1.05vs1.9, \np<0.001), olfactory bulbs (0.9vs1.7, p<0.01), and g rey-white matter contrast \n(0.9vs1.9, p<0.01). SNR improved in 35% of scans, a nd motion artefacts were \nreduced in 25%. \nConclusion: 3DSVR provides improved diagnoses of aqueduct steno sis, as \nevidenced by comparison with neonatal ground-truth scans. This is achieved \nby improving the visibility and overall quality of fetal brain MRI. Future work will \nvalidate 3DSVR in other pathologies. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The study is retrospective. \n \n\n \n \nThursday \nAbstract-based Programme \n \n 99  \nAuthor Disclosures:  \nMagdalena Sokolska: Nothing to disclose \nDonald Peebles: Nothing to disclose \nGiles Kendall: Nothing to disclose \nKelly Pegoretti Baruteau: Nothing to disclose \nLeigh Dyet: Nothing to disclose \nWeaam Hamed Elsayed Hamed: Nothing to disclose \nLatha Srinivasan: Nothing to disclose \nAnna David: Nothing to disclose \n \n \nQuantification of pathological fetal brain developm ent through \nsegmentation: a novel generative AI approach for sy nthetic pathological \ndata generation \n*M. Kaandorp*¹, H. Asma-Ull², H. G. Kim², D. Agbele se¹, K. Payette¹,  \nA. Jakab¹; ¹Zurich/CH, ²Seoul/KR \n(Misha.Kaandorp@kispi.uzh.ch) \n \nPurpose or Learning Objective: Fetal MRI is increasingly used for the \nquantification of the developing brain through segm entation of various \nanatomical structures. One of the challenges is tha t current segmentation \nalgorithms perform poorly in cases with ventriculom egaly due to limited high-\nquality annotations. Additionally, privacy concerns  often restrict data sharing. \nWe aimed to overcome these challenges through the g eneration of realistic \nsynthetic pathological MRIs from manipulated health y label images using \ngenerative AI. \nMethods or Background: We trained a stable diffusion model for 3D brain \nMRI synthesis (Med-DDPM) using 727 fetal and preter m neonatal MRI-label \nimage pairs. We generated 47 synthetic ventriculome galy label images (S47-\nventriculomegaly) from 33 healthy fetal MRIs (R33-h ealthy) by dilating \nventricular labels. Combining label images from R33 -healthy and S47-\nventriculomegaly, we generated 80 synthetic MRIs (S 80-ventriculomegaly) \nusing Med-DDPM and visually assessed their quality.  To evaluate \nsegmentation performance, we trained two nnUNet mod els on S80-\nventriculomegaly or R33-healthy MRI-label pairs and  measured performance \nusing Dice score on 40 test cases from FeTA2021 Cha llenge and 26 spina \nbifida severe ventriculomegaly cases (SPINABIFIDA).  Performance was also \nassessed for 80 healthy and pathological MRI-label pairs (R80) and their \nsynthetic equivalent (S80). \nResults or Findings: Med-DDPM generated diverse, high-quality synthetic \nfetal MRIs. In segmentation tasks, S80-ventriculome galy outperformed R33-\nhealthy, achieving higher Dice scores in FeTA2021 ( 0.773 vs. 0.751) and \nSPINABIFIDA (0.759 vs. 0.722), especially for ventr icles (0.780 vs. 0.672). S80 \nalso surpassed R80 in SPINABIFIDA (0.822 vs. 0.815) , with comparable \nperformance in FeTA2021 (0.766 vs. 0.773). \nConclusion: This study demonstrates that generating realistic p athological \nfetal MRIs by manipulating labels from normally dev eloping subjects can \nenhance data augmentation and data anonymization in  prenatal imaging. This \nis an important step towards addressing privacy con cerns while improving \nsegmentation performance. \nLimitations: Our method could be expanded to include more pathol ogies, \nincreasing clinical applicability. \nFunding for this study: Funding was provided by SNF grant: IZKSZ3_218590 \nEthics committee - additional information: The study has been approved by \nthe ethics committee of Zurich Children's Hospital,  decision number: 2022-\n01157) \nAuthor Disclosures:  \nHyun Gi Kim: Nothing to disclose \nAndras Jakab: Nothing to disclose  \nHosna Asma-Ull: Nothing to disclose  \nKelly Payette: Nothing to disclose \nDamola Agbelese: Nothing to disclose \nMisha Kaandorp: Nothing to disclose \n \n \nFeasability of fetal cardiac MRI in the prenatal ev aluation of congenital \nheart defects in comparison to US \n*G. Biechele*¹, B. Stos², D. Laux², S. Stöcklein¹, D. Grevent², L. J. Salomon²; \n¹Munich/DE, ²Paris/FR \n(gloria.biechele@med.uni-muenchen.de) \n \nPurpose or Learning Objective: Congenital heart defects (CHD) are common \nsevere birth defects, resulting in a clinical need for precise prenatal \nassessment. Magnetic Resonance Imaging (MRI) of the  fetal heart recently \nbecame feasible with the advent of novel gating tec hniques , but its feasibility \nhas not yet been systematically evaluated in pregna ncies with fetal CHD. \n \n \n \n \n \nMethods or Background: This study yet evaluated 42 singleton pregnancies, \nconsisting of 25 fetuses (60%) with diagnosis of CH D and 17 healthy control \nfetuses (40%). Diagnosis of CHD was refined by a ca rdiopediatrician. All MRI \nimages were reviewed by a single blinded operator. Image quality of cardiac \nsequences was rated from 0 (not usable) to 5 (high image quality in multiple \norientations). Cardiac anatomy was evaluated and co rrelation between MRI \nfindings and US diagnosis was assessed. \nResults or Findings: Fetal cardiac MRI was feasible in 40 patients (imag e \nquality score >0) due to extensive fetal movement i n n=2 cases. Among, in \nn=13 (32.5%), n=21 (52.5%) and in n=6 (15.0%) cases , image quality was \nrated high (image quality score 5), medium (image q uality score 3-4) and poor \n(image quality score 1-2), respectively. Among the 25 fetuses with CHD, fetal \ncardiac MRI was capable to fully confirm US-diagnos is in n=19 cases (76.0%) \nand to partly confirm US-diagnosis in n=6 (24.0%) c ases, mostly due to \nincomplete recording of structures of interest. The re was a substantial \nagreement between MRI and US (κ-value 0.70 with an accuracy of 0.85). \nConclusion: Our preliminary study suggests a similar diagnostic  performance \nof MRI to US. \nLimitations: With improvements in sequences, gating techniques a nd operator \nexperience, MRI examination might soon become an es sential part of the \nprenatal management of CHD, not only to identify as sociated abnormalities but \nalso to reinforce the assessment of the heart itsel f. \nFunding for this study: DFG Walter-Benjamin-Stipendium (BI 2563/1-1, G.B.) \nEthics committee - additional information: Local ethic commitee number \nNCT 04142606 \nAuthor Disclosures:  \nLaurent J. Salomon: Nothing to disclose \nSophia Stöcklein: Nothing to disclose \nDavid Grevent: Nothing to disclose \nBertrand Stos: Nothing to disclose \nDaniela Laux: Nothing to disclose \nGloria Biechele: Equipment Support Recipient: Compa ny Northh Medical for \nthe lend of a cardiac gating device during the peri od of a fellowship abroad. \n \n \nThoracic findings at early gestation fetal post-mor tem micro-CT \n*I. C. Simcock*¹, A. Lamouroux², S. C. Shelmerdine¹ , C. Hutchinson¹,  \nN. Sebire¹, O. Arthurs¹; ¹London/UK, ²Nimes/FR \n(ian.simcock@gosh.nhs.uk) \n \nPurpose or Learning Objective: To identify the range and frequency of \nthoracic diagnoses at less-invasive autopsy, follow ing a post-mortem micro-CT \ninvestigation for early gestation pregnancy loss. \nMethods or Background: Micro-CT provides high-resolution imaging for early  \ngestation fetuses (<300g), typically following misc arriage or termination of \npregnancy, allowing parents a less invasive autopsy  option. We retrospectively \nanalysed micro-CT diagnoses made on an unselected p opulation of over 1190 \nearly gestation fetuses between 2017 and 2024 at ou r tertiary referral \ninstitution. \nResults or Findings: Thoracic abnormalities on micro-CT were identified in \n(147/1190; 12.3%), comprising 148 cardiac and 79 no n-cardiac individual \nabnormalities. The commonest cardiac abnormalities were septal defects \n(50/147; 34.0%), aortic abnormalities e.g. coarctat ion (23/147; 15.6%), \ncongenital heart disease particularly tetralogy of Fallot (16/147; 10.9%) and \nventricular abnormalities e.g. hypoplastic left hea rt (11/147; 7.5%). Most \ncommon non-cardiac chest abnormalities were abnorma l fluid accumulation \n(41/147; 27.9%), which included pleural effusion (3 4/147; 23.1%), hydrops \n(4/147; 2.7%), and pericardial effusion (3/147; 2%) . A thoracic wall defect was \nobserved in (18/147; 12.7%) of cases. \nConclusion: A range of thoracic abnormalities were made by micr o-CT in our \ncohort, commonly from cardiac causes. This informat ion is useful for parents \nregarding the likelihood of congenital abnormalitie s in subsequent pregnancies \nand provides an alternative to conventional invasiv e autopsy. \nLimitations: Single centre data from a large specialist centre. Not all parents \nconsented to invasive autopsy, so some diagnoses co uld not be histologically \nconfirmed. \nFunding for this study: ICS was funded by a National Institute for Health \nResearch (NIHR) Clinical Doctoral Research Fellowsh ip (ICA-CDRF-2017-03-\n53), Development and Skills Enhancement Award (NIHR 302390) and a \nResearch for Patient Benefit Award (NIHR206174), OJ A was funded by a \nNIHR Career Development Fellowship (NIHR-CDF-2017-1 0-037) and SCS is \nsupported by a NIHR Advanced Fellowship Award (NIHR -301322), and the \nwork is funded by the Great Ormond Street Hospital Children’s Charity. AL was \nfunded by two mobility funding from Nimes and Montp ellier university hospital \nAll research at Great Ormond Street Hospital NHS Fo undation Trust and UCL \nGreat Ormond Street Institute of Child Health is ma de possible by the NIHR \nGreat Ormond Street Hospital Biomedical Research Ce ntre. The views \nexpressed are those of the author(s) and not necess arily those of the NHS, the \nNIHR or the Department of Health & Social Care. \n \n\n \n \nThursday \nAbstract-based Programme \n \n 100  \nEthics committee - additional information: We obtained ethical approval \nfrom the UK NHS Health Research Authority (HRA) Eth ics Committee (IRAS \nID: 131395). \nAuthor Disclosures:  \nIan C. Simcock: Nothing to disclose \nNeil Sebire: Nothing to disclose \nAudrey Lamouroux: Nothing to disclose \nCiaran Hutchinson: Nothing to disclose \nSusan Cheng Shelmerdine: Nothing to disclose \nOwen Arthurs: Nothing to disclose \n \n \nOptimizing Outcomes in PPROM: Ultrasound-Based Feta l Lung Maturity \nAnalysis \n*A. Verma*, A. Malik; New Delhi/IN \n(vermaanimesh53@gmail.com) \n \nPurpose or Learning Objective: To evaluate fetal lung maturity using \nmultiparametric ultrasound in pregnant women with p reterm premature rupture \nof membranes (PPROM), and to determine the accuracy  of ultrasound \nparameters in predicting neonatal respiratory distr ess in this population. This \nstudy highlights the role of multiparametric ultras ound in predicting neonatal \nrespiratory distress. \nMethods or Background: We conducted an 18 month observational \nprospective cohort study on 81 women with singleton  pregnancies under 37 \nweeks gestation, complicated by PPROM with fetuses between the 10th and \n90th weight percentiles. Fetal biometric measuremen ts included biparietal \ndiameter (BPD), head circumference (HC), abdominal circumference (AC), and \nfemur length (FL). We also assessed distal femoral and proximal tibial \nepiphyseal ossification centers, placental maturity  (Grannum system), and fetal \nmain pulmonary artery (MPA) Doppler flow measuring resistive index (RI), \npulsatility index (PI), acceleration time, and ejec tion time. These parameters \nwere correlated with neonatal respiratory distress to predict fetal lung maturity \nin PPROM cases. \nResults or Findings: BPD demonstrated highest diagnostic accuracy \n(81.48%) for predicting neonatal respiratory distre ss syndrome (RDS). The \nappearance of proximal tibial epiphyses showed rema rkable sensitivity \n(92.31%) in predicting RDS. Among the Doppler param eters, fetal MPA \nresistive index (RI) showed the highest diagnostic accuracy (88.89%). Notably, \ncombined fetal main pulmonary artery indices yielde d the highest diagnostic \naccuracy (95.06%), followed by combined fetal biome try (91.36%). \nConclusion: Fetal lung maturity, a key determinant for neonatal  respiratory \ndistress can be assessed by grayscale ultrasound an d Doppler parameters. \nMultiparametric ultrasonographic assessment is a pr omising tool in prediction \nof neonatal RDS in women with PPROM. \nLimitations: The study's generalizability is constrained by its limited sample \nsize. Additionally, assessment of the fetal main pu lmonary artery via \nultrasonography is subject to inter-observer variab ility and may be affected by \ninherent artifacts in ultrasound and Doppler imagin g techniques. \nFunding for this study: The cost of this study was covered by VMMC and \nSafdarjung Hospital, which provided essential resou rces and infrastructure to \nensure its completion. \nEthics committee - additional information: This study was approved by \nInstitutional Ethics Committee, VMMC and Safdarjung  Hospital, New Delhi \n(06/2022/CC-265) \nAuthor Disclosures:  \nAmita Malik: Nothing to disclose \nAnimesh Verma: Nothing to disclose \n \n \nPrenatal Prediction of Fetal Lung Maturity Using 3D  Lung Volume, Lung-\nto-Liver Intensity Ratio Tissue Histogram and Pulmo nary Artery Doppler \nIndices \nA. Omar, A. Mohamed Tharwat, M. Aboelnasr, H. Abo-A li Hamza,  \n*S. A. Hassanein*, W. Gaber Eldamaty; Shebin El Kom /EG \n(shaimaahamid@hotmail.com) \n \nPurpose or Learning Objective: Fetal lung maturity assessment is the most \ncritical factor for identifying the optimal deliver y time. A non-invasive \nsonographic technique is necessary to evaluate feta l lung development. we \naimed to predict maturity of fetal lung utilizing 3 D lung volume ultrasound, lung \nto liver intensity ratio, and pulmonary artery dopp ler indices measurement \nMethods or Background: A prospective observational study was conducted \non 200 pregnant females with gestational age of 32 to 40 weeks age \nunderwent 3D ultrasound (3DUS) for determination th e fetal lung volume (FLV) \nand fetal lung-to-liver intensity ratio (FLLIR) (tissue histogram) with doppler \nexamination of the main pulmonary artery (MPA) for the following parameters; \nacceleration-time to ejection-time ratio (At/Et), p ulsatility index (PI) and \nresistive index (RI) during a period of one week fr om delivery and comparing \nthe results to the neonatal outcome. \n \nResults or Findings: Of 200 fetuses investigated; 113 cases (56.5%) were  \nfound to have respiratory distress syndrome. The MP A RI and PI were \nsignificantly greater in fetuses with respiratory d istress syndrome comparing \nwith those without (2.6± 0.3 and 0.9±0.05 vs. 1.9±0.3 and 0.8±0.2, respectively \nwith p-value < 0.001 for both). MPA At/Et was signi ficantly lesser for fetuses \nwith RDS than fetuses without RDS (0.2±0.1 vs. 0.3±0.1 respectively, p-value \nless than 0.001). FLLIR was significantly lesser in  RDS +ve group comparing \nwith RDS -ve group (0.9±0.2 versus 1.3±0.3 respectively, p-value < 0.001) and \nFLV was significantly smaller in fetuses with respi ratory distress syndrome \ncomparing with those with no (31.5±2.5 vs. 38.1±2.8; p-value < 0.001). \nConclusion: The utilization of main pulmonary artery Doppler in dices, together \nwith mean fetal lung volume and FLLIR for assessing  fetal lung is a quick non-\ninvasive accurate technique for estimation of neona tal lung maturity and \nrespiratory distress syndrome \nLimitations: A uni-center study \nFunding for this study: No funding was present. \nEthics committee - additional information: Menoufia faculty of medicine \nresearch ethics committee under code no: (4/2022OBS G35 \nAuthor Disclosures:  \nAmal Omar: Nothing to disclose \nMohamed Aboelnasr: Nothing to disclose \nAhmed Mohamed Tharwat: Nothing to disclose \nShaimaa Abdelhamid Hassanein: Nothing to disclose \nWael Gaber Eldamaty: Nothing to disclose \nHaytham Abo-Ali Hamza: Nothing to disclose \n \n \nIntrauterine blood transfusion causes dose- and tim e-dependent signal \nalterations in the liver and the spleen on fetal ma gnetic resonance \nimaging \nM. Schwarz, V. Schmidbauer, *N. M. Nowak*, P. Kiena st, D. Bettelheim,  \nJ. Binder, T. Reiberger, D. Prayer, G. Kasprian; Vi enna/AT \n \nPurpose or Learning Objective: This study aimed to investigate the effects of \nIUTs on MRI findings in the fetal liver and spleen.  \nMethods or Background: Intrauterine transfusions (IUTs) are a life-saving \ntreatment for fetal anemia. However, with each tran sfusion iron bypasses \nuptake regulation through the placenta and accumula tes in fetal organs. Unlike \nother imaging modalities, fetal magnetic resonance imaging (MRI) is capable of \nnon-invasively assessing fetal liver disease and/or  organ iron overload. This \nstudy aimed to investigate the effects of IUTs on M RI findings in the fetal liver \nand spleen. For this retrospective study, we includ ed eight fetuses undergoing \nIUT and prenatal MRI from 2014 to 2023. The fetuses  were gestational age-\nmatched with a cohort that received fetal MRI for o ther indications, but no \nIUTs. Signal intensity (SI) and volumetric analyses  of the liver and the spleen \nwere performed. \nResults or Findings: Fetuses receiving transfusions had significantly la rger \nvolumes of both liver (p=0.003) and spleen (p=0.029 ). T1 SI inversely \ncorrelated with the number of IUTs (Pearson’s r=-0. 43, p = 0.099). This effect \nregressed over time (r=0.69, p=0.057). T2 SI did no t correlate significantly with \ntransfusion frequency but showed a strong positive correlation with the number \nof days between IUT and MRI (r=0.91, p=0.002). For splenic SI measures, \nsimilar effects were observed regarding T1 SI reduc tion per received \ntransfusion (r=-0.36, p=0.167) and recovery of T2 S I after IUT (r=0.88, \np=0.004). \nConclusion: This is the first study to report the effects of IU Ts on MRI data of \nfetal livers and spleens. We observed considerable dose- and time-dependent \nSI alterations of the liver and spleen following IU T. Furthermore, fetal \nhepatosplenomegaly can be expected following IUT. \nLimitations: Two fetuses have had parvovirus B19 infection, whic h may cause \nhepato-/splenomegaly, but may also mandate the need  for IUT. \nFunding for this study: Medical University of Vienna \nEthics committee - additional information: In concordance with the \nprinciples of the Declaration of Helsinki, approved  by the ethics committee of \nthe Medical University of Vienna. \nAuthor Disclosures:  \nVictor Schmidbauer: Nothing to disclose \nNikolaus Michael Nowak: Nothing to disclose \nJulia Binder: Nothing to disclose \nMichael Schwarz: Nothing to disclose \nDieter Bettelheim: Nothing to disclose \nDaniela Prayer: Nothing to disclose \nThomas Reiberger: Nothing to disclose \nPatric Kienast: Nothing to disclose \nGregor Kasprian: Nothing to disclose \n \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 101  \nImproving image quality and decreasing SAR with hig h dielectric \nconstant pad in 3.0T fetal MRI \n*Z. Zhu*, C. Yan, Z. Lin, B. Zhang; Nanjing/CN \n(161230042@smail.nju.edu.cn) \n \nPurpose or Learning Objective: To assess the potential of high dielectric \nconstant (HDC) pad in increasing image quality and decreasing specific \nabsorption rate (SAR) in 3.0T fetal MRI. \nMethods or Background: This prospective single-center observational study \nincluded 168 pregnant participants taking 3.0T feta l MRI scanning with and \nwithout HDC pad between 1 May 2021 and 31 November 2023. Quantitative \nImage-quality analysis included signal-to-noise rat io (SNR) and contrast-to-\nnoise ratio (CNR). Qualitative analysis was perform ed by three radiologists with \nfour-point scale to evaluate overall image quality,  dielectric artifact and \ndiagnostic confidence. Whole-body total SAR was als o compared. Correlation \nbetween image quality variable changes and particip ant clinical characteristics \nwas evaluated using spearman correlation. \nResults or Findings: 128 participants (mean gestational age 30.25±3.53 \nweeks, range 22-37 weeks) undertook balanced steady  state free precession \n(bSSFP) sequence and 40 participants (mean gestatio nal age 30.38±3.50 \nweeks, range 23-37 weeks) undertook single-shot fas t spin-echo (SSFSE) \nsequence. With HDC pad, SNR and CNR was significant ly higher (41.45% \nincrease in SNR, 54.05% increase in CNR on bSSFP, p <0.001; 258.76% \nincrease in SNR, 459.55% increase in CNR on SSFSE, p<0.001). Overall \nqualitative image quality, dielectric artifact and diagnostic confidence improved \nsignificantly (p<0.001). Adding HDC pad significant ly reduced Whole-body total \nSAR (32.60% on bSSFP, p<0.001; 15.40% on SSFSE, p=0 .005). There was \nno significant correlation between image quality va riable changes and \nparticipant clinical characteristics (p>0.05). \nConclusion: In a clinical setting, adding HDC pad can increase overall \nquantitative and qualitative image quality while re ducing dielectric artifact and \nSAR. \nLimitations: Due to technical limitations, we could not conduct further analysis \nto compare the radiofrequency power deposited on ut erus, whole fetal body \nand fetal brain separately. We only conducted exper iments on magnetic \nresonance machines produced by United Imaging and d idn’t include machines \nfrom other manufactures. \nFunding for this study: National Science and Technology Innovation 2030 \n(2022ZD0211800) \nEthics committee - additional information: Ethic committe of Nanjing Drum \nTower Hospital \nAuthor Disclosures:  \nBing Zhang: Nothing to disclose \nZhengyang Zhu: Nothing to disclose \nZengping Lin: Nothing to disclose \nChenchen Yan: Nothing to disclose \n \n \n12:30-13:30 Research Stage 3 \nResearch Presentation Session: Hybrid, \nMolecular and Translational Imaging \nRPS 906 \nExploring the frontiers in hybrid and \nmolecular imaging \n \nModerator \nM. Naik; London/UK  \n(m.naik@nhs.net) \n \n \nLow-dose Fibronectin-targeted Gd-based contrast age nt Enables Early \nand Accurate Assessment of Chemotherapy Response in  Pancreatic \nCancer \n*W. Zhang*, X. Liang, Y. Du, J. Tian, N. Hong; Beij ing/CN \n \nPurpose or Learning Objective: Albumin-bound paclitaxel and gemcitabine \n(AG) chemotherapy is a mainstay in treatment of pan creatic ductal \nadenocarcinoma (PDAC), unfortunately not all patien ts respond to this \ntreatment. Clinical imaging techniques cannot preci sely evaluate and predict \nthe response to AG therapies over several weeks. A strong fibrotic reaction is a \nhallmark of drug-resistance while depletion of fibr osis is a positive response to \nAG. Extradomain-B fibronectin (EDB-FN) is an import ant element of fibrosis in \nPDAC. Here, we prepared EDB-FN targeted Gd-based co ntrast agent (EDB-\nGd) to perform molecular MRI for early, noninvasive  and quantitative \nassessment of treatment response in PDAC. \nMethods or Background: BxPC-DR was pre-treated with AG to establish \nacquired drug resistance. Subcutaneous and orthotop ic models with BxPC-DR \nor BxPC were established. Mice were intravenously i njected with EDB-Gd or \nGd-DOTA. The ratio of T1 value reduction (T1d%) wer e compared \nquantitatively. For chemotherapy monitoring, MRI wa s performed before and \nafter AG treatments. Histological analyses were use d for validation. \nResults or Findings: Molecular MRI with EDB-FN could specifically detect  \nand quantify fibrogenesis in PDAC xenografts at a l ow dose 0.05mmol/kg, \nwhich is half of clinical dosage of Gd. The optimal  imaging time point was \n30min after injection of EDB-Gd. In addition, the t argeted probe generated \nmore robust contrast-enhanced and longer retention compared to traditional \nGd-DOTA. For chemotherapy montoring, T1d% were sign ificantly increased \n2.5-fold in drug-resistance xenografts group in fib rotic tumor areas compared \nto AG-sensitive group (p < 0.05). Comparing the T1d % before and 5 days after \nAG predicted treatment response. \nConclusion: This study indicates EDB-FN-targeted molecular MRI possesses \nclinical applications in accurate assessment and pr ediction of AG \nchemotherapy. \nLimitations: Correlations between the observed fibrotic changes on MRI with \npathologic markers of treatment response and even s urvival and prognosis are \nneeded in the future. \nFunding for this study: This study was funded by the Beijing Natural Scienc e \nFoundation (Grant No. 7244524, 7212207) and Nationa l Natural Science \nFoundation of China (Grant Nos. 62027901, 82272111,  92159303, 82071896, \n81871422, 81871514, and 81227901). \nEthics committee - additional information: All experimental studies were \napproved by the Ethics Committee of the Peking Univ ersity People's Hospital \n(2024PHE048). \nAuthor Disclosures:  \nJie Tian: Nothing to disclose \nYang Du: Nothing to disclose \nNan Hong: Nothing to disclose \nWenjia Zhang: Nothing to disclose \nXiaolong Liang: Nothing to disclose \n \n \nInvestigating the Relationship of Endothelin Recept or Expression and \nTumor Hypoxia by Optoacoustic Tomography \n*A. Helfen*, M. Mallik, M. Stölting, E. Hoffmann, C . Höltke; Münster/DE \n(anne.helfen@ukmuenster.de) \n \nPurpose or Learning Objective: A significant prognostic factor of tumor \nmalignancy is the formation of a hypoxic supportive  tumor microenvironment \n(TME). The endothelin (ET)M signaling network is li nked to tumor hypoxia \nthrough stabilized hypoxia-inducible factor 1 in a feedback loop. Here, we \nexamined the interrelation of both cellular signali ng systems by multispectral \noptoacoustic tomography (MSOT). \nMethods or Background: Murine syngeneic 4T1 breast tumors were \nexamined in vivo using MSOT depicting deoxygenated (Hb) and oxygenated \n(HbO2) hemoglobin content to detect hypoxic regions  over one week. An \nexogenous fluorescent endothelin-A receptor (ETAR) probe served for \nevaluating ETAR expression status. Therapeutic inte rventions (anti-angiogenic \nand macrophage depletion) were evaluated concerning  Hb/HbO2 ratio and \nETAR expression changes. Treatment response to Beva cizumab, Clodronate \nand Sorafenib was detected over one week. \nResults or Findings: MSOT was capable of delineating and quantifying \nhypoxia within tumor lesions. 4T1 tumors were highl y hypoxic compared to \nhealthy tissue, represented by oxygen saturation (s O2) of 0.33 vs. 0.79, \nrespectively. Baseline data of the ETAR probe showe d an initial rise in signal \nintensity from day 0 to day 8, corresponding to tum or growth. Therapeutic \ninterventions showed that the ETAR signal intensity  could be significantly \nreversed, while all applied therapies did not lead to significant tumor growth \nreduction. However, Sorafenib and Bevacizumab led t o a significant increase \nin sO2 values. MSOT data were supported by subseque nt \nimmunohistochemistry. \nConclusion: Tumor hypoxia within syngeneic murine breast cancer  can be \nvisualized non-invasively by MSOT evaluating hemogl obin (oxygenated and \ndeoxygenated ratios) as well as ETAR expression. Fu rthermore, MSOT was \nable to depict therapeutic effects already in the e arly course of treatment \nrepresenting a potential imaging biomarker. \nLimitations: Limited penetration depth of MSOT is significantly improved \ncompared to optical imaging, but with regard to tra nslation still mainly suitable \nfor superficial tissues. \nFunding for this study: Financial support from the German Research \nFoundation (DFG, SF656 A04), from the medical facul ty of the University of \nMünster (IMF: I-HÖ111709) and from the Joachim Herz  Stiftung is gratefully \nacknowledged. \nEthics committee - additional information: All animal experiments described \nin this study were approved by the responsible auth orities (“Landesamt für \nNatur, Umwelt und Verbraucherschutz NRW”, Germany, Protocol No. 84-\n02.04.2017.A011). \n \n\n \n \nThursday \nAbstract-based Programme \n \n 102  \nAuthor Disclosures:  \nMiriam Stölting: Nothing to disclose \nCarsten Höltke: Nothing to disclose \nAnne Helfen: Nothing to disclose \nEmily Hoffmann: Nothing to disclose \nMoushami Mallik: Nothing to disclose \n \n \nTheranostic Innovation: SPCCT-Triggered Photodynami c X-ray Therapy \n(XPDT) in a Murine Breast Cancer Model \n*P. Akl*, A. Carret, A. Houmeau, A. Gautheron, I. G oddard, J-B. Langlois,  \nB. Montcel, F. Lerouge, P. C. Douek; Lyon/FR \n(piaakl@gmail.com) \n \nPurpose or Learning Objective: Photodynamic therapy (PDT) uses light to \nactivate photosensitizers (PS) but is limited by sh allow penetration. X-ray PDT \n(XPDT) overcomes this by using X-rays for deeper ac tivation. Spectral Photon-\nCounting Computed Tomography (SPCCT) enhances imagi ng with K-Edge \ncapabilities. This study aims to optimize XPDT in v itro and in vivo using \nSPCCT and gadolinium nanoparticles (GdNp) as contra st and therapeutic \nagents. \nMethods or Background: GdNp doped with terbium were synthesized, \nfunctionalized for biocompatibily with different co atings (PEG and Silica) and \nthe photosensitizer, rose bengal (RB) and irradiate d at different concentrations \n(0.02-1 M) in vitro using SPCCT with different X-ra y dose parameters (80-140 \nKVp and 10-300 mAs). Luminescence emission was reco rded using optical \nfiber immersed in the GdNp solution. 63 athymic nud e mice with MDA-MB-231 \ncell-derived breast cancer xenograft models were ir radiated using \nSPCCT(120kVp and 300mAs) 60 axial acquisitions/ 1 s ec, meandose 30mGy, \n1800mGy in total ,24 hours after intra tumoral inje ction. Safety, biodistribution, \nand therapeutic effects were assessed through mice examination every 3 days \nfor tumor volume to assess tumor growth delay .Biod istribution was monitored \nusing SPCCT imaging. \nResults or Findings: Fluorescence emission increased with X-ray dose and  \nconcentration of GdNp. SPCCT imaging revealed speci fic distribution of GdNp, \nwith luminescence emission proportional to the admi nistered dose in vitro and \nin vivo. The silica + RB nanoparticle group showed a slight improvement in \nefficacy compared to controls, with delayed tumor g rowth. Biodistribution \nstudies indicated relatively different patterns of intratumoral and peritumoral \nlocalization, and heterogenous distribution. \nConclusion: This technology shows potential for customizing and  enhancing \nthe efficacy of XPDT, modulate luminescence intensi ty by adjusting X-ray dose \nparameters and nanoparticle concentration. Silica +  RB showed a slight \nimprovement in efficacy compared to controls. \nLimitations: Though further studies are required to confirm sign ificance. \nFunding for this study: EU Horizon 2020 grant agreement 899549 \nEthics committee - additional information: ethical comittee IRB \nAPAFIS#44558. \nAuthor Disclosures:  \nIsabelle Goddard: Nothing to disclose \nJean-Baptiste Langlois: Nothing to disclose \nAngele Houmeau: Nothing to disclose \nFrederic Lerouge: Nothing to disclose \nPhilippe Charles Douek: Nothing to disclose \nBruno Montcel: Nothing to disclose \nAlison Carret: Nothing to disclose \nArthur Gautheron: Nothing to disclose \nPia Akl: Nothing to disclose \n \n \n[18F]FDG-PET/CT imaging biomarkers for time point-m atched response \ncharacterization of experimental melanomas to anti- PD-L1/anti-CTLA-4 \nimmunotherapy \n*M. J. Antons*, S. Kloiber-Langhorst, H. Hirner-Epp eneder, F. Herr, S. Ziegler, \nM. Brendel, J. Ricke, M. Heimer, C. C. Cyran; Munic h/DE \n(melissa.antons@med.uni-muenchen.de) \n \nPurpose or Learning Objective: Three-time point [18F]FDG-PET/CT imaging \nallows for in vivo monitoring of a combined anti-PD -L1/anti-CTLA-4 \nimmunotherapy in a murine melanoma model validated by time point-matched \nmultiparametric immunohistochemical reference stand ard \nMethods or Background: Melanoma cells (B16-F10) were injected \nsubcutaneously into the abdominal flank of C57BL/6 mice (n=40). Following a \nbaseline scan after day 7, the therapy group receiv ed 5 injections (i.p.) of anti-\nPD-L1 and anti-CTLA-4 antibodies on days 7, 9, 11, 13 and 15. The control \ngroup received sham treatment. Follow-up scans were  performed on day 13 \nand 19. Tumor allografts were harvested for time po int-matched \nimmunohistochemistry (CD8, Ki-67, TUNEL) to validat e PET/CT parameters \n(MTV, SUVmax) as imaging biomarkers of early therap y response. \n \n \nResults or Findings: At follow-up 1 (FU-1), the therapy group exhibited \nsignificantly lower MTV compared to the control gro up (p=0.0037). By follow-up \n2 (FU-2), both MTV and SUVmax were significantly lo wer in the therapy group \nversus the control group (MTV: p=0.0078; SUVmax: p= 0.00034). Ex vivo \nanalysis revealed significant anti-tumor effects in  the therapy group, with a \nsignificantly higher apoptosis rate at FU-1 (p= 0.0 12) and FU-2 (p= 0.001). \nMoreover, the therapy group demonstrated a signific ant increase in CD8-\npositive T-cells at FU-2 (p=0.0027), while tumor ce ll proliferation was \nsignificantly lower at both follow-up time points ( FU-1: p=0.012; FU-2: \np=0.012). \nConclusion: Multi-time point [18F]FDG-PET/CT allowed for the ea rly non-\ninvasive monitoring of a combined immunotherapy wit h anti-PD-L1/anti-CTLA-4 \nin experimental melanomas, validated by multiparame tric \nimmunohistochemistry. The significantly lower tumor  glucose metabolism was \nparalleled by significant pro-immunogenic, pro-apop totic and anti-proliferative \neffects of the combined immunotherapy. \nLimitations: The allograft model of melanoma may have limited tr anslational \nrelevance to human tumor pathophysiology. Secondly,  the observation period \nwas relatively short, and no clinical endpoints suc h as overall survival of the \nanimals were determined. \nFunding for this study: None \nEthics committee - additional information: All animal experiments were \nperformed in accordance with the guidelines for the  use of living animals in \nscientific studies and the animal study was officia lly approved (ROB-55.2-\n2532.Vet_02-19-32). \nAuthor Disclosures:  \nClemens C. Cyran: Nothing to disclose \nMaurice Heimer: Nothing to disclose \nFelix Herr: Nothing to disclose \nSibylle Ziegler: Nothing to disclose \nMelissa J. Antons: Nothing to disclose \nHeidrun Hirner-Eppeneder: Nothing to disclose \nSandra Kloiber-Langhorst: Nothing to disclose \nMatthias Brendel: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nIdentification of intratumoral clusters in breast c ancer xenograft tumors \nby simultaneous multiparametric [18F]FMISO-PET/MRI \n*S. J. Bartsch*¹, J. Friske¹, M. Hacker¹, D. Laimer -Gruber¹, D. Prinz¹,  \nT. Wanek¹, T. H. Helbich¹, K. Pinker-Domenig²; ¹Vie nna/AT, ²New York, NY/US \n \nPurpose or Learning Objective: Hypoxia is a driver of breast cancer (BC) \nprogression, inducing more aggressive phenotypes an d intratumoral \nneovascularization. The quantification of hypoxia a nd neovascularization with \nsimultaneous multiparametric [18F]FMISO-PET/MRI wou ld benefit the \ncharacterization of the hypoxic tumor microenvironm ent in BCs. We aim to \ncombine simultaneous multiparametric [18F]FMISO-PET /MRI biomarkers and \nuse them for the identification of intratumoral clu sters for a holistic assessment \nof hypoxia and neovascularization in BCs. \nMethods or Background: Female athymic nude mice (n = 32) were \ninoculated with luminal A, HER2+ or triple negative  BC cells. PET/MRI was \nperformed on a Bruker 94/30USR system, combined wit h a Bruker PET-insert. \nHypoxia was evaluated using [18F]FMISO-PET, and by hyperoxic blood \noxygen level dependent (BOLD) MRI. Neovascularizati on was assessed via \ndynamic contrast enhanced MRI, and non-contrast-enh anced intravoxel \nincoherent motion MRI. Intratumoral clusters were i dentified based on a \nhierarchical cluster analysis using R (version 4.2. 3). \nResults or Findings: The cluster analysis of [18F]FMISO-PET/MRI revealed  \ndistinct clusters in all BC subtypes. Clusters corr esponding to hypoxia showed \nelevated SUVbw values and the lowest ktrans and ve,  indicating dense tissue \nand limited vessel permeability, along with little change in R2* following \nhyperoxic BOLD-MRI. The least hypoxic cluster had t he lowest SUVbw values \nand the highest ktrans and ve of all clusters and h igh change in R2*. Hypoxic \nclusters were more prevalent in HER2+ and triple ne gative BCs than in luminal \nA BCs, while less hypoxic clusters were most common  in luminal A BCs. \nConclusion: Simultaneous multiparametric [18F]FMISO-PET/MRI pro vides a \nholistic perspective on hypoxia and neovascularizat ion in three BC molecular \nsubtypes. These insights enable a non-invasive char acterization of BC, and \nmay be used for assessing treatment response. \nLimitations: The use of intratumoral cluster identification for the assessment \nof treatment response has to be confirmed in upcomi ng studies. \nFunding for this study: This work was funded by the Vienna Science and \nTechnology Fund (WWTF), project number LS19-018. \nEthics committee - additional information: This animal study was approved \nby Austrian Federal Ministry of Education, Science and Research \n[66.009/0284-WF/V/3b/2017; 2020-0.363.124; 2022-0.7 26.820] and the \nIntramural Committee for Animal Experimentation of the Medical University of \nVienna. \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 103  \nAuthor Disclosures:  \nSilvester Julian Bartsch: Nothing to disclose \nDaniela Laimer-Gruber: Nothing to disclose \nKatja Pinker-Domenig: Nothing to disclose \nThomas H. Helbich: Nothing to disclose \nDaniela Prinz: Nothing to disclose \nMarcus Hacker: Nothing to disclose \nJoachim Friske: Nothing to disclose \nThomas Wanek: Nothing to disclose \n \n \nComparative Radiomics and Feature Consistency: In V ivo CT vs. Ex Vivo \nMicro-CT in Classifying Lung Cancer Subtypes \n*L. Brizzi*, L. Preda, C. Bortolotto, S. Megalizzi,  D. Malerba, F. Checchin; \nPavia/IT \n(md.brizzileonardo@gmail.com) \n \nPurpose or Learning Objective: The study aimed to compare radiomic \nfeatures between conventional in vivo CT and micro- CT of ex vivo lung tumor \nblocks following lobectomy. This analysis included 60 patients with lung \ncancer, comprising 30 adenocarcinomas (ADK) and 30 squamous cell \ncarcinomas (SCC). The goal was to assess the variat ion and significance of \nfeatures to improve predictive accuracy in distingu ishing between tumor types. \nMethods or Background: The dataset comprised 107 radiomic features, \ncompliant with IBSI standards, extracted using Pyra diomics software. \nStatistical analyses, including a percentage variat ion calculation and t-tests, \nwere performed to evaluate the correlation and cons istency of features \nbetween in vivo CT and ex vivo micro-CT. The analys is focused on features \nwith potential utility in discriminating between AD K and SCC tumors. \nResults or Findings: Out of the 107 radiomic features, 82 showed less th an \n10% variation between the two imaging modalities, w ith 46 features exhibiting \na variation of less than 1%. Shape features and app roximately 90% of GLCM \nfeatures demonstrated strong consistency. T-tests r evealed that 21 radiomic \nfeatures had a p-value < 0.05, indicating statistic al significance. These features \nincluded various shape characteristics and first-or der statistics that are crucial \nfor tumor classification. \nConclusion: The findings indicate a strong correlation between the radiomic \nfeatures extracted from in vivo CT and micro-CT, pa rticularly in shape and \nGLCM features. The identified significant features offer promising potential for \nimproving predictive models for lung cancer classif ication between ADK and \nSCC. \nLimitations: The study was limited to a specific subset of lung tumors (ADK \nand SCC), and the generalizability to other lung ca ncer subtypes remains \nuncertain. Additionally, technical variations in CT  acquisition parameters might \naffect feature extraction consistency. Further vali dation with larger datasets is \nnecessary. \nFunding for this study: Founding are provided by Research Foundings on AI \nof the IRCCS Policlinico San Matteo \nEthics committee - additional information: Protocol 25657/2024 \nAuthor Disclosures:  \nLeonardo Brizzi: Nothing to disclose \nSilvia Megalizzi: Nothing to disclose \nDavide Malerba: Nothing to disclose \nFilippo Checchin: Nothing to disclose \nLorenzo Preda: Nothing to disclose \nChandra Bortolotto: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n12:30-13:30 Research Stage 4 \nResearch Presentation Session: Chest \nRPS 904 \nImaging of pulmonary embolism and \npulmonary hypertension \n \nModerator \nN. J. Screaton; Cambridge/UK  \n(n.screaton@nhs.net) \n \n \nRadiomics parameters of epicardial adipose tissue p redict mortality in \nacute pulmonary embolism \n*H-J. Meyer*¹, S. Zimmermann¹, J. Borggrefe², A. Su rov²; ¹Leipzig/DE, \n²Minden/DE \n \nPurpose or Learning Objective: Accurate prediction of short-term mortality in \nacute pulmonary embolism (APE) is very important. T he aim of the present \nstudy was to analyze the prognostic role of radiomi cs values of epicardial \nadipose tissue (EAT) in APE. \nMethods or Background: Overall, 508 patients were included into the study,  \n209 female (42.1%), mean age, 64.7 ± 14.8 years. 4.6%and 12.4% died (7- \nand 30-day mortality, respectively). For external v alidation, a cohort of 186 \npatients was further analysed. 20.2% and 27.7% died  (7- and 30-day mortality, \nrespectively). CTPA was performed at admission for every patient before any \nprevious treatment on multi-slice CT scanners. A tr ained radiologist, blinded to \npatient outcomes, semiautomatically segmented the E AT on a dedicated \nworkstation using ImageJ software. Extraction of ra diomic features was applied \nusing the pyradiomics library. Patients were random ly assigned to a training \nand a validation cohort with a ratio of 7:3. We cha racterized two models (30-\nday and 7-day mortality). \nResults or Findings: We fitted the characterized models to a validation cohort \n(n = 169) in order to test accuracy of our models. We observed an AUC of \n0.776 (CI 0.671-0.881) and an AUC of 0.724 (CI 0.62 8-0.820) for the prediction \nof 30-day mortality and 7-day mortality, respective ly. The overall percentage of \ncorrect prediction in this regard was 88% and 79% i n the validation cohorts. \nLastly, the AUC in an independent external validati on cohort was 0.721 (CI \n0.633-0.808) and 0.750 (CI 0.657-0.842), respective ly. \nConclusion: Radiomics parameters of EAT are strongly associated  with \nmortality in patients with APE. \nLimitations: It is a retrospective study, which should be evalua ted in a \nprospective multi center analysis. \nFunding for this study: None \nEthics committee - additional information: Nr. 145/21, Ethics Committee, \nOtto-von-Guericke University of Magdeburg, Magdebur g, Germany) \nAuthor Disclosures:  \nAlexey Surov: Nothing to disclose \nSilke Zimmermann: Nothing to disclose \nHans-Jonas Meyer: Nothing to disclose \nJan Borggrefe: Nothing to disclose \n \n \nFeasibility Study on the Use of 6ml Iodine Contrast  Agent in Pulmonary \nArtery CT \n*H. Shang*, Y. Gao, D. D. Tian, K. Li, X. Zhang, P.  Cao; Xi'an/CN \n(727537956@qq.com) \n \nPurpose or Learning Objective: To assess the viability of using a low dose of \niodine contrast medium (CM) and slow injection rate  in single-energy 40keV CT \nimaging for pulmonary artery angiography (CTA). \nMethods or Background: Seventy patients, clinically suspected of pulmonary  \nembolism and treated between January and September 2024, were randomly \nassigned to either an experimental group (35 patien ts) or a control group (35 \npatients). The experimental group underwent imaging  at 100kV with 6ml CM at \nan injection rate of 2.5ml/s, triggered at a thresh old of 60HU. The control group \nwas imaged at 120kV with 35-40ml CM at an injection  rate of 3.5ml/s, triggered \nat 80HU. All scans targeted the pulmonary artery tr unk and utilized the non-\nionic, water-soluble iodine contrast agent iomeprol  (400mg/ml). Both groups \nwere scanned under free breathing conditions. Image  quality was subjectively \nrated on a 5-point scale, and objectively assessed based on pulmonary \nvascular enhancement, CT values, signal-to-noise ra tio (SNR), contrast-to-\nnoise ratio (CNR), and radiation dose metrics inclu ding dose-length product \n(DLP) and volumetric CT dose index (CTDIvol). \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 104  \nResults or Findings: The CT values of the main pulmonary artery showed n o \nsignificant difference between groups (P>0.05). How ever, the experimental \ngroup demonstrated significantly higher CT values i n the right upper, middle, \ninterlobar, and lower lobe arteries, as well as the  left upper and lower lobe \narteries (P < 0.05). Although the experimental grou p exhibited higher noise, \nSNR, and CNR, these differences were not statistica lly significant (P>0.05). \nThe experimental group also had significantly lower  CTDIvol and DLP values \ncompared to the control group (P<0.05). Subjective image quality evaluations \nrevealed no significant difference between the two groups (P>0.05). \nConclusion: CTA of the pulmonary artery at 40keV with 6ml CM an d an \ninjection rate of 2.5ml/s can provide diagnostic im age quality. \nLimitations: Not applicable \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nPan Cao: Nothing to disclose \nDong Dong Tian: Nothing to disclose \nXiaoyue Zhang: Nothing to disclose \nKai Li: Nothing to disclose \nHaitao Shang: Nothing to disclose \nYanjun Gao: Nothing to disclose \n \n \nDouble-low protocol CTPA for Pulmonary Embolism det ection: Enabling \nartificial intelligence with deep learning-reconstr uction based Images \n*L. Shen*, J. Lu, C. Zhou, Z. Bi, X. Ye, M. Zeng, W . Mingliang; Shanghai/CN \n(809797357@qq.com) \n \nPurpose or Learning Objective: To assess the effectiveness of artificial \nintelligence software (AI) in detecting pulmonary e mbolism (PE) using low-dose \nCT pulmonary angiography (CTPA) enhanced by deep le arning reconstruction \n(DLR) and contrast-enhancement boost (CE-boost) tec hnique. \nMethods or Background: This prospective two-center study included 180 \npatients who underwent CTPA for suspected PE. Patie nts were randomly \ndivided into two groups: the routine CTPA group wit h 50 mL contrast medium \n(CM) was reconstructed using HIR, and the low-dose CTPA group with 25 mL \nCM was reconstructed using DLR. The CE-boost was ad ditionally performed to \ngenerate DLR-boost images in the low-dose group. Si gnal-to-noise ratio (SNR) \nand contrast-to-noise ratio (CNR) of pulmonary arte ries were quantitatively \nassessed. For qualitative image quality assessment,  two experienced \nradiologists independently rated CT images (5, best ; 1, worst). A subset of 46 \nrandomly selected patients in each group (1:1 ratio ) were evaluated by the AI \nsoftware (Discover PE, uAI) for the presence of PE.  Reference standard was \nestablished by expert consensus. The diagnostic acc uracy (sensitivity and \nspecificity) of the AI interpretations were compare d between methods by \nbootstrapping. \nResults or Findings: DLR-boost images produced lower noise, higher SNR \nand CNR, and superior subjective image quality comp ared to HIR images in \nthe routine group (p < 0.05). For detecting PE, com pared to HIR images, DLR-\nboost images showed comparable sensitivity (97.67% vs. 93.02%, p < 0.001), \nand equivalent specificity of (both 100.00%, p > 0. 05). The effective dose of the \ndouble-low group and the routine group was 1.19 ± 0.45 mSv and 2.69 ± 0.49 \nmSv, respectively. \nConclusion: DLR-boost significantly enhances CTPA image quality  at reduced \nradiation and contrast doses. AI software achieves diagnostic performance \ncomparable to traditional reconstruction methods, s upporting its use in clinical \npractice. \nLimitations: N/A \nFunding for this study: This study has received funding by Shanghai \nAnticancer Association EYAS PROJECT (Grant NO. SACA  CY22C15). \nEthics committee - additional information: This study was approved by the \nEthics Committee of Shanghai Geriatric Medical Cent er(B2024-009). \nAuthor Disclosures:  \nXiaodan Ye: Nothing to disclose \nLeilei Shen: Nothing to disclose \nZhenghong Bi: Nothing to disclose \nWang Mingliang: Nothing to disclose \nChun Zhou: Nothing to disclose \nMengsu Zeng: Nothing to disclose \nJinjuan Lu: Nothing to disclose \n \n \nAI-driven pulmonary vascular analysis with computed  tomography in \npatients with chronic thromboembolic pulmonary dise ase \n*A. Cisarri*, A. Valentini, E. M. Bassi, I. Fiorina , A. D'Onorio De Meo,  \nG. Rodolico, M. Zacchino, K. Ellena, L. Preda; Pavi a/IT \n(andrea.ciso94@gmail.com) \n \nPurpose or Learning Objective: Chronic thromboembolic pulmonary disease \n(CTEPD) is a rare condition in which patients may o r may not develop \npulmonary hypertension. Although both groups are tr eated in the same way, \nthrough pulmonary thromboendarterectomy (PEA), the pathogenic \nmechanisms by which some patients develop pulmonary  hypertension while \nothers do not have yet to be clarified. The study a ims to investigate vascular \nchanges using AI-driven quantification of pre- and post-operative CT scans in \npatients undergoing PEA. \nMethods or Background: The study analyzed 35 patients with chronic \nthromboembolic disease, divided into three groups b ased on preoperative \nmean pulmonary arterial pressure (mPAP), from 2017- 2022. Pre- and post-\noperative CT scans of 22 of these patients were ana lyzed with AI-software to \nquantify vascular morphology, including vessel numb er, diameter, and blood \nvolume at various pleural depths. Hemodynamic param eters such as mPAP \nand pulmonary vascular resistance (PVR) were also a ssessed. \nResults or Findings: No significant differences in vascular quantificati on \nparameters were observed between patient groups pre -operatively. Post-\ntreatment analysis showed a reduction in the number  of small vessels \n(p=0.0065) and blood volume of small vessels (BV5) at 24 mm depth \n(p=0.036), with an increase in mean vessel diameter  (p=0.0005). A significant \ncorrelation was found between BV5 reduction and PVR  improvement (p=0.01). \nConclusion: CT quantification revealed significant post-operati ve vascular \nchanges in CTEPD patients, especially in the medium -caliber vasculature. The \nfindings suggest that PVR reduction is primarily li nked to the reperfusion of \nlarger vessels, with limited improvement in smaller  vessel volumes. Future \nstudies should explore these parameters as potentia l biomarkers for diagnosis \nand prognosis in CTEPD. \nLimitations: The small sample size and limited postoperative fol low-up restrict \nthe study’s ability to generalize findings. Further  prospective studies with larger \ncohorts are needed. \nFunding for this study: No funding \nEthics committee - additional information: Approved \nAuthor Disclosures:  \nGiuseppe Rodolico: Nothing to disclose \nEmilio Maria Bassi: Nothing to disclose \nIlaria Fiorina: Nothing to disclose \nAlessandro D'Onorio De Meo: Nothing to disclose \nAndrea Cisarri: Nothing to disclose \nMichela Zacchino: Nothing to disclose \nLorenzo Preda: Nothing to disclose \nKatia Ellena: Nothing to disclose \nAdele Valentini: Nothing to disclose \n \n \nAutomated Quantified CT analysis of Morphological D ifferences in \nChronic Thromboembolic Pulmonary Disease and chroni c \nthromboembolic pulmonary hypertension \n*W. Xu*, L. Xi, A. Liu, M. Liu, S. Zhao; Beijing/CN  \n(wenqingxu2021@163.com) \n \nPurpose or Learning Objective: We aim to study the new morphological \nmarkers of Chronic Thromboembolic Pulmonary Disease  (CTEPD) and chronic \nthromboembolic pulmonary hypertension (CTEPH) on co mputed tomography \npulmonary angiography (CTPA). \nMethods or Background: We retrospectively enrolled CTEPH, CTEPD \npatients, and control group from January 2019 to Oc tober 2023 in our hospital. \nThe morphological metrics including pulmonary blood  volume, tortuosity, and \nfractal dimension (FD) on CTPA were automatically q uantified on an Artificial \nIntelligence workstation. We compared these metrics  among three groups and \nassessed their correlation with hemodynamics. \nResults or Findings: A total of 190 participants (97 men, 56.2±10.9 year s old) \nincluding 116 CTEPH patients ,54 CTEPD patients and  20 control enrolled in \nthis study. The pulmonary artery tortuosity in the control group, CTEPD group, \nand CTEPH group showed a gradually increased progre ssively (1.07 [1.06–\n1.10] vs. 1.10 [1.07–1.14] vs. 1.14 [1.10–1.18], P< 0.01). There was a positive \ncorrelation between pulmonary artery tortuosity and  mean pulmonary artery \npressure (r=0.47, P<0.01), pulmonary vascular resis tance (r=0.44, P<0.01). \nAdditionally, the volume of small and medium-sized pulmonary arteries was \nsignificantly higher in CTEPD patients compared to those with CTEPH \n(P<0.01). FD among three groups was comparable(p>0. 05). \nConclusion: Pulmonary arterial tortuosity on CTPA is a crucial imaging \nbiomarker for distinguishing between CTEPH and CTEP D. The preservation of \nnormal volumes in the small and medium-sized pulmon ary arteries observed in \nCTEPD patients implies that this feature could be a  key determinant in \nmaintaining normal resting pulmonary artery pressur e. \nLimitations: First, it was a single-center investigation with a small cohort of \nCTEPD patients and controls, and the non-normal dis tribution limits the \ngeneralizability of our findings. Larger studies wi th more diverse populations \nare needed to validate our results. Secondly, there  is a need for further \nrefinement in the precision of pulmonary vessel seg mentation. \nFunding for this study: This work was supported by This study is supported \nby the National Natural Science Foundation of China  (82272081), Chinese \nAcademy of Medical Sciences Innovation Fund for Med ical Sciences (2021-\nI2M-1-049, 2022-I2M-C&T-B-109). All authors have no thing to disclose. \n\n \n \nThursday \nAbstract-based Programme \n \n 105  \nEthics committee - additional information: This single-center study was \napproved by the hospital’s Ethics Committee (2022-K Y-048) and was \nperformed in accordance with the Declaration of Hel sinki. \nAuthor Disclosures:  \nMin Liu: Nothing to disclose \nWenqing Xu: Nothing to disclose \nLinfeng Xi: Nothing to disclose \nAnqi Liu: Nothing to disclose \nShihua Zhao: Nothing to disclose \n \n \nComparison of dual energy CT and V/Q SPECT in diagn osis of chronic \nthromboembolic pulmonary hypertension \nE. Pershina, D. Shchekochikhin, *A. Oganesyan*; Mos cow/RU \n(talilen@mail.ru) \n \nPurpose or Learning Objective: To assess diagnostic value of DECT versus \nV/Q SPECT for CTEPH detection in PAH patients in Mo scow Pulmonary \nHypertension Center. \nMethods or Background: DECT with the calculation of iodine maps and V/Q \nSPECT were performed in 29 patients (f/m - 9/13; ag e 65 ±10) with PAH and \nrisk factors for CTEPH. All patients underwent righ t heart catheterization for \nPAH confirmation. CTA analysis included the number and level of vessel \nocclusions together with the presence and size of l ung perfusion defects on \niodine maps and CT signs of right heart failure. \nResults or Findings: 21 of 29 patients (72,4%) demonstrated CTA features  of \npossible CTEPH such as intraluminal defects, enlarg ed pulmonary trunk, \npulmonary mosaic patten. Iodine maps revealed perfu sion defects of 19 \npatients (86%). Two patients (7%) had intraluminal irregular defects without \nperfusion defects by DECT. V/Q SPECT determined per fusion defects in 20 \npatients (69%). There was one case (3, 4%) with mis match between iodine \nmaps and V/Q SPECT. However, it was explained by th e enlarged pressure on \nthe right heart chambers due to the cava-caval (cav a superior – cava inferior) \nvenous anastomosis and hepatic veins abnormalities.  Sensitivity for perfusion \ndefects by DECT was 95% and specificity 100%. Furth ermore, the mean \nradiation dose was significantly lower for DECT vs.  V/Q-SPECT (p = 0.006). \nSpreading of perfusion defect had strong correlatio n with severity of right heart \nfailure (r=0.2, p<0.05). \nConclusion: Pulmonary CTA with iodine mapping improves the dete ction of \nCTEPH. Iodine maps demonstrate high sensitivity and  specificity in comparison \nwith V/Q SPECT. According to our preliminary data D ECT could be potentially \nimplemented in future diagnostic PH algorithms at l east on par with V/Q-\nSPECT. Further research is needed. \nLimitations: Pregnant \nOlder 18 years \nGFR>30 ml/min \nIodine allergic reaction \nFunding for this study: No funding \nEthics committee - additional information: The study was approved by local \nethical committee. \nAuthor Disclosures:  \nEkaterina Pershina: Nothing to disclose \nAnait Oganesyan: Other: Analysis datas Author: Has wrote study Speaker: \npresents study \nDmitry Shchekochikhin: Nothing to disclose \n \n \nDevelopment of a CT based prognostic predictive mod el in chronic \nthromboembolic pulmonary hypertension (CTEPH) \n*S. Gowda*, R. Kothari, V. Raj; Bengaluru/IN \n(srjgwd@gmail.com) \n \nPurpose or Learning Objective: Chronic thromboembolic pulmonary \nhypertension (CTEPH) is a severe complication of pu lmonary embolism \ncharacterized by persistent pulmonary arterial hype rtension due to unresolved \nthromboembolic obstructions. This study explores th e development and \nvalidation of a CT based prognostic predictive mode l designed to simulate \nhemodynamic parameters and guide management in CTEP H patients. \nMethods or Background: We retrospectively analysed the data of 1100 \npatients diagnosed with CTEPH. Integrated data cons isted of clinical history, \nechocardiogram reports, right heart catheterisation  reports if any, CT \npulmonary angiogram studies and proposed management  strategies. CT clot \nburden score, CT parameter score and Perfusion defe ct score was obtained in \nall the patients. Patients were contacted to review  the clinical outcomes. \nResults or Findings: Parameter scoring was seen to be useful in predicti ng \nhemodynamic parameters and outcomes. A cut-off of 3  is close to normal \nhemodynamic status (sPAP, accuracy of 76%), while a  cut-off of 5 represents \nhigher than normal values of mPAP. A score of ≥7 indicated increased risk of \nmortality. \n \n \nConclusion: The CT based prognostic predictive model represents  a valuable \ntool in the management of CTEPH by integrating deta iled imaging with clinical \nfactors. We suggest this model can be a standard co mponent of CTEPH \npatient evaluation and management especially when r ight heart catheterisation \nis contraindicated or not possible. Future studies should focus on prospective \nvalidation and integration into clinical practice t o further refine its utility and \nimpact in the management. It can be used to avoid r ight heart catheterisation \nand ventilation-perfusion scans in selected patient s who meet the cut-off \nvalues, thereby reducing additional cost burden to the patients. \nLimitations: Single centre study, which may also have an in-buil t case \nselection bias. Small sample size to ascertain like lihood of mortality \nNeeds prospective validation to integrate into clin ical practice \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The study received institutional \nreview board approval. \nAuthor Disclosures:  \nRicha Kothari: Nothing to disclose \nVimal Raj: Nothing to disclose \nSuraj Gowda: Nothing to disclose \n \n \nPerformance of chest contrast-enhanced CT in pulmon ary hypertension \nclinical grouping: a dual-center, expert-blinded an alysis \n*L. Nardone*¹, L. Cereser¹, G. Agati¹, P. Ciolli², T. Nadarević³, C. Cicciò⁴,  \nA. Borghesi², R. Girometti¹, C. Zuiani¹; ¹Udine/IT,  ²Brescia/IT, ³Rijeka/HR, \n⁴Negrar/IT \n(luiginardone1996@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the diagnostic \naccuracy of contrast-enhanced CT imaging in classif ying patients with \npulmonary hypertension (PH) across international gu idelines-derived clinical \ngroups I-V. The analysis focused on quantifying the  utility of CT in a blinded \nsetting, comparing the results with those from the PH multidisciplinary team \nmeeting (PH-MDTM). \nMethods or Background: We retrospectively included 172 contrast-enhanced \nCT studies from patients with PH performed in two t ertiary referral centers. \nThree chest-devoted radiologists, blinded to the cl inical data, independently \nreviewed all the CTs, assigning probability percent ages for each of the five PH \ngroups. A consensus grouping hypothesis was reached  by averaging the \nprobabilities across readers, and this was compared  with the PH-MDTM \ngrouping. Accuracy and Cohen’s Kappa (k)-derived in ter-reader agreement \nvalues with 95% confidence intervals (95% CI) were calculated. The readers’ \ndiscriminatory power between individual PH groups w as evaluated through \nareas under the receiver operating characteristic c urve (AUC) analyses. \nResults or Findings: The expert-blinded consensus diagnosis agreed with the \nPH-MDTM in 124/172 cases (accuracy, 72%; k, 0.62; 9 5%CI, 0.50-0.70). \nWhen including the second most probable group, the readers correctly \ngrouped 148/172 cases (accuracy, 86%; k, 0.81; 95%C I, 0.74-0.88). \nDiscriminatory power analysis for individual groups  demonstrated good \nreaders’ performance, with AUC values ranging from 0.79-0.90 depending on \nthe group. \nConclusion: This study highlights the value of contrast-enhance d CT in \nclassifying PH according to clinical groups, with e xpert readers achieving \nsubstantial-to-almost perfect agreement with the PH -MDTM grouping and good \ndiscriminatory power for individual PH groups. The results suggest that CT \nimaging can serve as a reliable tool in the clinica l work-up of PH patients, \nsupporting its integration into multidisciplinary e valuations. \nLimitations: Retrospective design, low number of group V PH case s. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number IRB 250/2023. \nAuthor Disclosures:  \nAndrea Borghesi: Nothing to disclose  \nChiara Zuiani: Nothing to disclose \nRossano Girometti: Nothing to disclose \nGiorgio Agati: Nothing to disclose \nPietro Ciolli: Nothing to disclose  \nTin Nadarević: Nothing to disclose \nLorenzo Cereser: Nothing to disclose  \nLuigi Nardone: Nothing to disclose \nCarmelo Cicciò: Nothing to disclose \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 106  \n14:00-15:30 Research Stage 1 \nResearch Presentation Session: \nGenitourinary \nRPS 1007 \nImaging of renal and bladder \nmalignancies: advances in diagnosis and \ncharacterisation \n \nModerator \nN. Cowan; Portsmouth/UK  \n \n \nEnhanced U-Net for Precise Auto-Segmentation of Bla dders and Tumors \nin CT Urography Imaging \n*L. Chen*, L. Mao, X. Li, X. Zhang, X. Bai, G. Zhan g, H. Xue, Z. Jin, H. Sun; \nBeijing/CN \n(li.cherish@outlook.com) \n \nPurpose or Learning Objective: This study developed a deep learning model \nfor bladder and tumor segmentation from CT Urograph y images (CTU), integral \nto a system aiding bladder cancer management. \nMethods or Background: A 381-case dataset from two centers, approved by \nthe Institutional Review Board, was utilized. It co mprised 280 training cases \n(275 with masses, 5 normal), 56 internal validation  cases (54 with masses, 2 \nnormal), and 45 external validation cases (44 with masses, 1 normal). The set \nincluded nephrographic phase CTU scans with 0.625mm  and 1mm slice \nthicknesses for patients with pathologically confir med lesions. A reference \nstandard with manual contours was provided by an ex perienced radiologist and \nreviewed by a senior one. The nnU-Net framework tra ined a U-Net-based \nsegmentation model using an ensemble prediction in a five-fold cross-\nvalidation and test-time augmentation. Performance was assessed using DSC, \n95% HD, and MSD on the testing set. \nResults or Findings: Results show that our approach achieves superb \nsegmentation accuracy. In the internal validation s et, the U-Net-based model \nshowed strong performance with a DSC of 97.9%, 95% HD of 0.48mm, and \nMSD of 3.43mm for bladder segmentation. It excelled  in the external set with a \nDSC of 98.4%, 95% HD of 0.34mm, and MSD of 2.52mm. For tumor \nsegmentation, the internal set results were a DSC o f 76.6%, 95% HD of \n3.70mm, and MSD of 19.15mm, while the external set showed a slight \ndecrease to a DSC of 74.4%, 95% HD of 3.73mm, and M SD of 22.08mm. \nConclusion: Though the tumor segmentation was less precise than  bladder \nsegmentation, the U-Net-based model still provided satisfactory accuracy for \nboth, excelling in bladder delineation. This model proves valuable for detecting \nbladder cancer and evaluating treatment efficacy. \nLimitations: The limited sample size of the external validation cohort limited \nthe generalizability. \nFunding for this study: This work was supported by National High-Level \nHospital Clinical Research Funding(2022-PUMCH-A-035 ), National High-Level \nHospital Clinical Research Funding(2022-PUMCH-B-069 ), National High-Level \nHospital Clinical Research Funding(2022-PUMCH-A-033 ), the Natural Science \nFoundation of China (Grant No.81901742), the Beijin g natural Science \nFoundation (Grant No. L232133), and the CAMS Innova tion Fund for Medical \nSciences (2022-12M-C&T-B-019). \nEthics committee - additional information: The Pecking Union College \nHospital Institutional Review Board approval was ob tained (ethical approval \nnumber: I-22PJ887). \nAuthor Disclosures:  \nXiuli Li: Nothing to disclose \nLi Mao: Nothing to disclose \nGumuyang Zhang: Nothing to disclose \nHao Sun: Nothing to disclose \nXin Bai: Nothing to disclose \nZhengyu Jin: Nothing to disclose \nLi Chen: Nothing to disclose \nXiaoxiao Zhang: Nothing to disclose \nHuadan Xue: Nothing to disclose \n \n \n \n \n \n \n \nRadiogenomics of renal cell carcinoma: using MRI tu mor features to \npredict mir-15a expression \n*Y. Mytsyk*¹, P. Kowal², Y. Kobilnyk³, I. Dutka¹, I . Komnatska¹, A. Górecki⁴; \n¹Lviv/UA, ²Wroclaw/PL, ³Przemyśl/PL, ⁴Przeworsk/PL \n(mytsyk.yulian@gmail.com) \n \nPurpose or Learning Objective: The aim of this study was to evaluate the \neffectiveness of MRI tumor parameters in predicting  tissue expression of miR-\n15a in renal cell carcinoma (RCC) patients. \nMethods or Background: The study involved 64 patients with histologically \nconfirmed conventional RCC, where miR-15a expressio n was measured, and \npreoperative contrast-enhanced MRI (1.5 T) was perf ormed. MiR-15a \nexpression was determined using reverse transcripti on and real-time PCR. A \npolynomial regression model assessed associations b etween miR-15a \nexpression and radiological tumor parameters, with accuracy evaluated by the \nFisher method (adjusted R²). \nResults or Findings: It was found that radiological features of the cyst ic \ncomponent, exophytic growth, necrosis, macroscopic fat, and nodular contrast \nenhancement of the tumor were observed in 29.69%, 2 3.44%, 32.81%, \n20.31%, and 37.5% of patients, respectively. The me an levels of miR-15a \nexpression in the presence of these features were 0 .35±1.02 U, 0.34±1.09 U, \n4.01±3.42 U, 0.29±0.87 U, and 2.91±3.24 U, respectively. In the absence of \nthese features, the mean expression of miR-15a was 2.01±2.93 U, 1.88±2.85 \nU, 0.82±1.85 U, 1.83±2.83 U, and 0.68±1.72 U, respectively (p<0.05). The \nhighest miR-15a expression levels were observed wit h necrosis, and the \nlowest with macroscopic fat (p<0.05). Tumor size st rongly correlated with miR-\n15a expression (r=0.724; p<0.001). Tumor size alone  predicted miR-15a \nexpression with an adjusted R² of 0.8281, and combi ning tumor size with other \nradiological features predicted 85% of miR-15a expr ession (R²=0.8336; \np<0.001). The study developed a predictive formula for miR-15a expression \nbased on RCC radiological features. \nConclusion: The findings suggest that MRI parameters can accura tely predict \nmiR-15a expression, which holds diagnostic and prog nostic value in RCC. \nLimitations: The main limitation was the inclusion of only conve ntional RCC. \nFunding for this study: No funding. \nEthics committee - additional information: The study was approved by the \nEthics Committee of the Danylo Halytsky Lviv Nation al Medical University \n(protocol No. 5 dated May 25, 2021). The work was c onducted in accordance \nwith accepted standards for conducting research in the field of biology and \nmedicine, based on the guidelines of the World Heal th Organization, the \nInternational Council of Medical Scientific Societi es, the International Code of \nMedical Ethics (1983), the Helsinki Declaration ado pted by the General \nAssembly of the World Medical Association, the Conv ention on Human Rights \nand Biomedicine of the Council of Europe (1997), an d the requirements and \nstandards of ICH GCP (2002). In each specific case,  patients or responsible \nindividuals provided written consent for the surgic al intervention. \nAuthor Disclosures:  \nYulian Mytsyk: Nothing to disclose \nIryna Komnatska: Nothing to disclose \nYuriy Kobilnyk: Nothing to disclose \nAndrzej Górecki: Nothing to disclose \nPaweł Kowal: Nothing to disclose \nIhor Dutka: Nothing to disclose \n \n \nAn Artificial Intelligence Framework Based On Contr ast-Enhanced CT For \nPreoperative Predicting WHO/ISUP Nuclear Grade Of C lear Cell Renal Cell \nCarcinoma: A Multicenter Study \n*J. Han*¹, T. Liu², J. Li¹, Y. Zhang³; ¹Hohhot/CN, ²Guangzhou/CN, ³Zhuhai/CN \n(handaile@163.com) \n \nPurpose or Learning Objective: To determine whether the artificial \nintelligence integrated model based on automatic se gmentation of CT images \ncan provide a robust prediction of clear cell renal  cell carcinoma (ccRCC) \nISUP/WHO grade. \nMethods or Background: Pretreatment CT scans were retrospectively \nacquired in patients with surgically proven ccRCC a t multiple centers from \nJanuary 2017 to September 2023.The proposed framewo rk comprised five \nmodules, including a 3D tumor segmentation model by  3D-UNet, a deep \nlearning feature extraction module, a radiomic feat ure extraction module, a \nclinical-radiological feature screening module, and  a fully-connected \nclassification module that combines features from d ifferent sources to classify \nlow-grade (I and II) and high-grade (III and IV) cc RCC. The Grad-CAM method \nand SHAP method are used to analyze the interpretab ility of the artificial \nintelligence model. \n \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 107  \nResults or Findings: The training data set was comprised of 335 patients  \nfrom three centers, and 110 and 84 patients were in cluded in the two external \ntest data sets. The average Dice coefficient of the  3D-UNet automatic \nsegmentation network in the test sets was 0.86 and 0.82. Synchronous distant \nmetastasis, Planned nephrectomy type, and tumor lon g axis as independent \npredictors of high-grade ccRCC. In the test sets, t he AUC and accuracy of \nintegrated model were 0.85-0.92, 78-85%, respective ly, which were exceeded \nthose of clinical-radiological feature model (0.85 vs0.75 [P = 0.039], 0.93 vs \n0.76 [P = 0.043], 78% vs 65% [P < 0.001], 85% vs 71 % [P = 0.035]. \nConclusion: An integrated model based on clinical features, rad iological \nfeatures, radiomics features, and deep learning fea tures provided reliable \nprediction of WHO/ISUP grade for ccRCC, which outpe rformed the clinical-\nradiological feature model. \nLimitations: This is a retrospective study and we only included ccRCC with a \npathological diagnosis of WHO/ISUP grading after ne phrectomy , with some \nselection bias. \nFunding for this study: This work was supported in part by the National \nNatural Science Foundation of China under Grant Nos . 81801809, 82371917, \n81830052, 81971691, 12126610, 62371476; in part by the Basic and Applied \nBasic Research Foundation of Guangdong Province und er Grant Nos. \n2020A1515010572, and in part by the Zhuhai Basic an d Applied Basic \nResearch Foundation under Grant Nos. ZH220170032000 01PWC. \nEthics committee - additional information: This study was approved by the \ninstitutional review boards of the Fifth Hospital o f Sun Yat- sen University. \nAuthor Disclosures:  \nTao Liu: Nothing to disclose \nYaqin Zhang: Nothing to disclose  \nJunlin Li: Nothing to disclose \nJiayue Han: Nothing to disclose \n \n \nPrevalence of Venous Extension in Malignant Adrenal  Neoplasia: \nIdentification of a Novel Imaging Sign \nL. Melges, C. Torres, F. Chahud, D. F. Maia, L. Col li, J. Elias,  \nC. A. Fernandes Molina, M. Castro, *V. F. Muglia*; Ribeirao Preto/BR \n(fmuglia@fmrp.usp.br) \n \nPurpose or Learning Objective: Adrenal vein involvement is a common \nfeature of adrenocortical carcinomas, but its preva lence in metastatic adrenal \nlesions remains unknown. Our goal was to assess the  prevalence of adrenal \nvein involvement in primary and metastatic adrenal lesions and to determine if \nmorphological changes in tumor shape precede venous  extension. \nMethods or Background: This retrospective, single-center observational \nstudy evaluated 102 patients: 28 adrenal cortical c arcinoma (ACC) patients; \nand 74 non-ACC cancer patients that presented adren al metastasis (82 \nmetastatic adrenal lesions). Two readers reviewed c ross-sectional imaging to \nassess tumor size, laterality, venous invasion, and  the presence of the \"edge \nsign.\" Surgical and histopathological confirmation was the reference standard \nfor ACCs, while for metastases, sequential imaging or PET-CT results showing \nhypermetabolism were used when histopathology was u navailable. \nResults or Findings: Of the 28 ACC patients, 82.1% were female, with \nbalanced laterality. Metastases primarily originate d from the lung (24.4%), \ncolorectal (13.4%), and breast (12.2%) cancers and had a left-side dominance \n(61.7%). Venous extension was present in 14.6% of m etastases and 21.4% of \nACCs. The \"edge sign\" was more frequently observed in metastatic lesions \n(26.8%) than in ACCs (17.8%). Interobserver agreeme nt was almost perfect for \nvenous extension (κ = 0.9256) and substantial for the edge sign ( κ = 0.7844). \nConclusion: Venous extension was less prevalent in metastatic a drenal \nlesions compared to ACCs. The edge sign may precede  venous extension, \nespecially in metastatic cases. Although these find ings could impact clinical \nevaluation, prospective multicenter studies are nee ded to confirm the clinical \nsignificance of the edge sign. \nLimitations: Retrospective, single-center study. \nFunding for this study: FAEPA - Foundation for the development of learning,  \nassistance and research of Ribeirao Preto School of  Medicine Hospital. \nEthics committee - additional information: The study was approved by the \nEthics and Research Committee of the Clinical Hospi tal at Ribeirao Preto \nMedical School under number CAAE 78221024.2.0000.54 40, with a waiver \nfrom the Informed Consent Form \nAuthor Disclosures:  \nJorge Elias: Nothing to disclose \nMargaret Castro: Nothing to disclose \nDavid Freire Maia: Nothing to disclose \nFernando Chahud: Nothing to disclose \nLais Melges: Nothing to disclose \nCarlos Augusto Fernandes Molina: Nothing to disclos e \nLeandro Colli: Grant Recipient: Bristol Myers Squib b - Immunotherapy for \nRenal Cell Carcinnoma \nValdair Francisco Muglia: Nothing to disclose \nCecilia Torres: Nothing to disclose \n \n \nIRM K01 study: Diagnostic Value of Multiparametric MRI for Small Solid \nRenal Tumors \n*E. Jambon*¹, N. Grenier¹, C. Marcelin¹, A. Crombé¹ , G. Margue¹, \nJ-C. Bernhard¹, F. H. Cornelis²; ¹Bordeaux/FR, ²New  York, NY/US \n(eva.fourage@gmail.com) \n \nPurpose or Learning Objective: Small renal tumors are increasingly being \ndetected incidentally, posing diagnostic challenges . Up to 23% of small renal \ntumors result in non-contributive biopsies. This st udy evaluates the diagnostic \nvalue of multiparametric MRI (mpMRI) in the charact erization of small solid \nrenal tumors. The aim was to assess the diagnostic accuracy of mpMRI in \ndifferentiating malignant from benign small solid r enal tumors in patients with \nsuspected malignancy but no evident signs of metast asis. \nMethods or Background: This is a prospective multicentric French study. A \ncohort of 387 patients in 17 centers with non-hered itary, solid renal masses \nbetween 1.5 and 4 cm in diameter was enrolled betwe en November 2018 and \nMay 2022. MRI protocols included T1w, T2w, diffusio n-weighted imaging, and \ndynamic contrast-enhanced sequences. Radiologists p erformed blinded \nreadings with a centralized review in case of disco rdance. The primary \nendpoint is the negative predictive value (NPV) of a dichotomized Likert scale \nscore, targeting a 98% NPV. \nResults or Findings: The study found a 45% NPV for mpMRI, falling short of \nthe expected 98% due to difficulties in distinguish ing clear cell renal cell \ncarcinoma (ccRCC) from oncocytomas, which constitut ed 80% of the benign \ntumors in the cohort. Despite this, mpMRI influence d clinical management \ndecisions, increasing \"surveillance without biopsy\"  by 25% and reducing \nbiopsies by 42%. However, it also led to a 25% incr ease in partial/total \nnephrectomies. \nConclusion: While mpMRI showed limited ability to accurately di stinguish \ncertain benign lesions from malignant tumors under predefined criteria, it \nproved effective in identifying malignant cases. Th is led to a shift in clinical \nmanagement favoring surgical interventions over bio psies. The study highlights \nthe need for more objective imaging criteria and fu rther research into \nquantitative measures and radiomic analysis for bet ter tumor characterization. \nLimitations: Predefined criteria \nFunding for this study: No funding \nEthics committee - additional information: PHRC-K \nApproval from the French Ethics Committee (CPP) \nAuthor Disclosures:  \nJean-Christophe Bernhard: Nothing to disclose \nGaelle Margue: Nothing to disclose \nFrançois H Cornelis: Nothing to disclose \nEva Jambon: Nothing to disclose \nAmandine Crombé: Nothing to disclose \nClément Marcelin: Nothing to disclose \nNicolas Grenier: Nothing to disclose \n \n \nADC measurement may improve the diagnostic performa nce of bi-\nparametric bladder MRI in predicting detrusor muscl e invasion of bladder \ncancer \n*M. N. Tasdemir*, U. Eryürük, S. Aslan; Giresun/TR \n(mervetsdmr@gmail.com) \n \nPurpose or Learning Objective: To assess the diagnostic performance of a \nmodified biparametric VIRADS (mbp-VIRADS), derived from a combination of \nADC measurements and biparametric MRI (bp-MRI), in predicting detrusor \nmuscle invasion in bladder cancer (BC). \nMethods or Background: Patients with histopathologically confirmed BC \nbetween June 2020 and May 2024 were analyzed retros pectively. Two image \nsets, biparametric MRI (set 1) and multiparametric (mp) MRI (set 2), were \nformed. Tumors were categorized using both the bp-V IRADS and mp-VIRADS \nsystems. The optimal ADC value to differentiate mus cle-invasive bladder \ncancer (MIBC) from non-muscle-invasive bladder canc er (NMIBC) was \ndetermined using a receiver operating characteristi c (ROC) curve. To assess \nthe mbp-VIRADS category, for cases with a bp-VIRADS  score of 2-4, scores \nwere upgraded for those below the ADC cut-off value  and downgraded for \nthose above it. \nResults or Findings: A total of 182 patients with BC met the study crite ria. Of \nthese patients, 146 had NMIBC and 36 had MIBC. Comp aring VIRADS \ncategories with MIBC detection, AUC of the ROC anal ysis was 0.896, 0.940, \nand 0.941 for the bp-MRI, mbp-MRI, and mp-MRI proto cols, respectively. The \nsensitivity for bp-VIRADS, mbp-VIRADS, and mp-VIRAD S scores (with a cutoff \n≥4) were 78%, 88%, and 90%, specificity were 91%, 95 %, and 93%; and \noverall accuracy were 88%, 92%, and 93%, respective ly. Using bp-MRI, there \nwere 12 false-positive and 9 false-negative cases f or predicting muscle \ninvasion. With mp-MRI, false positives decreased to  9, and false negatives to \n4. When using mbp-MRI, false positives further decr eased to 6, with 5 false \nnegatives. \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 108  \nConclusion: By combining ADC measurements with bp- MRI features , the \ndiagnostic performance of bp-MRI in predicting musc le invasion of bladder \ncancer can be significantly improved, approaching t hat of mp- MRI, while \nreducing the false-positive and false-negative rate s. \nLimitations: This was a retrospective study, \nFunding for this study: None \nEthics committee - additional information: The study was approved by the \nlocal ethics commitee. \nAuthor Disclosures:  \nUluhan Eryürük: Nothing to disclose \nMerve Nur Tasdemir: Nothing to disclose \nSerdar Aslan: Nothing to disclose \n \n \nEvaluating VI-RADS Score Performance in the Post-TU RBT setting: \nExploring the Need for Modification \n*A. Dehghanpour*, M. Pecoraro, L. Laschena, M. Bicc hetti, C. Catalano,  \nV. Panebianco; Rome/IT \n(ad48ad@yahoo.com) \n \nPurpose or Learning Objective: The aim of this study was to assess the \ndiagnostic accuracy of the VI-RADS score and its in dividual MRI categories, \nstructural category (T2W), diffusion category (DWI) , and contrast-enhanced \ncategory (DCE), in patients who underwent diagnosti c transurethral resection \nof bladder tumor (TURBT). Additionally, we correlat ed the diagnostic accuracy \nof VI-RADS and its individual sequences with the ti me interval between TURBT \nand MRI, to suggest the optimal timing for MRI afte r TURBT. We also \ncalculated the inter-reader agreement in scoring VI -RADS after TURBT. \nMethods or Background: This retrospective single-center study included 150  \npatients who underwent mpMRI after TURBT at varying  intervals. Four \nexperienced readers in bladder MRI, independently a nd blinded to \nclinicopathological information, evaluated the scan s, providing both VI-RADS \nscores and local staging. Each evaluation was perfo rmed twice: once with DWI \nas the dominant sequence and once with DCE as the d ominant sequence. The \nonly exclusion criterion was prior systemic therapy . Histopathological results \nfrom therapeutic TURBT or radical cystectomy were u sed as the reference \nstandard. \nResults or Findings: The AUC for VI-RADS in detecting muscle-invasive \nbladder cancer was 0.88 (95% CI 0.84-0.92) for the most experienced reader. \nOn a per-sequence analysis, DWI showed the highest AUC (0.83 [95% CI \n0.78-0.87]), followed by DCE (0.68 [95% CI 0.63-0.7 4]). The diagnostic \naccuracy of VI-RADS improved when the time between TURBT and MRI \nexceeded 2 weeks and became optimal after 4 weeks, regardless of whether \nDWI or DCE was the dominant sequence. \nConclusion: The earliest acceptable timing for MRI after TURBT is at least 2 \nweeks, with the optimal timing being after 4 weeks.  In scoring VI-RADS after \nTURBT, DWI should be considered the dominant sequen ce, due to its high \nsensitivity and specificity. \nLimitations: Retrospective design and readers being from the sam e center. \nFunding for this study: None. \nEthics committee - additional information: Institutional Ethical Committee. \nAuthor Disclosures:  \nMarco Bicchetti: Nothing to disclose \nValeria Panebianco: Nothing to disclose \nAilin Dehghanpour: Nothing to disclose \nMartina Pecoraro: Nothing to disclose \nLudovica Laschena: Nothing to disclose \nCarlo Catalano: Nothing to disclose \n \n \nUncertainty-Aware Interactive Deep Learning System for Predicting \nPathological T3a Upstaging in Renal Cell Carcinoma Using CT Images \n*E. Yuan*¹, Q. Zhou², Y. Chen¹, C. He¹, L. Ye¹, J. Yao¹, B. Song¹; \n¹Chengdu/CN, ²Shanghai/CN \n \nPurpose or Learning Objective: To develop and validate a deep learning \nsystem for predicting pathological T3a upstaging in  renal cell carcinoma while \nintegrating prediction uncertainty to facilitate mo re reliable clinical decision-\nmaking. \nMethods or Background: In this retrospective study, we collected pathology -\nconfirmed RCC patients who underwent surgical resec tion from three tertiary \nacademic medical centers. The data in first center were split into training and \ntesting datasets. Three DenseNet-121 models were tr ained to predict the \noverall T3a invasion, the inner invasion, and the o uter invasion. The \nuncertainty was quantified by ensemble-based and vo ting-based methods. For \nuncertain cases, manual interpretation was performe d to obtain the final \nprediction. The performance of the pure model and u ncertainty-aware \ninteractive system were evaluated and compared on t he testing dataset and \ntwo external datasets using area under the ROC curv e (AUC). \nResults or Findings: The data of 1329 patients (975:235:119) were collec ted \nand analyzed. The DL system performed worse in the uncertain group \ncompared to the certain group of testing dataset (A UC 0.73 (95% CI: 0.61, \n0.85) vs 0.81 (95% CI: 0.71, 0.90)), the external_1  dataset (AUC 0.50 (95% CI: \n0.33, 0.68) vs 0.95 (95% CI: 0.90, 0.99)), and the external_2 dataset (AUC \n0.82 (95% CI: 0.62, 1.00) vs 0.94 (95% CI: 0.89, 0. 99)). The net reclassification \nindex for the DL system were 0.11, 0.18, and 0.13 i n testing, external_1, and \nexternal_2 datasets. \nConclusion: The uncertainty-aware interactive deep learning sys tem \neffectively predicts pathological T3a upstaging in renal cell carcinoma, with \nmanual interpretation improving performance in unce rtain cases. This \napproach enhances diagnostic reliability, demonstra ting potential for improved \nclinical decision-making across multiple datasets. \nLimitations: The sample size in external validation datasets wer e limited. \nFunding for this study: None \nEthics committee - additional information: The written consent was waived \nfor the retrospective design. \nAuthor Disclosures:  \nJin Yao: Nothing to disclose \nQing Zhou: Nothing to disclose \nChunlei He: Nothing to disclose \nLei Ye: Nothing to disclose \nEnyu Yuan: Nothing to disclose \nBin Song: Nothing to disclose \nYuntian Chen: Nothing to disclose \n \n \nEnhancing Diagnostic Accuracy in Renal Tumor Identi fication: Impact of \nStructured Training on the Clear Cell Likelihood Sc ore \n*M. Cosenza*, G. Brembilla, G. Imperiale, A. Larche r, U. Capitanio,  \nF. Montorsi, F. De Cobelli; Milan/IT \n(Cosenza.michele@hsr.it) \n \nPurpose or Learning Objective: This study aimed to evaluate the \nimprovement in diagnostic performance of radiologis ts in identifying clear cell \nand papillary renal tumors using the Clear Cell Lik elihood Score (CCLS) before \nand after a structured training program. \nMethods or Background: This monocentric study analyzed 60 MRI scans, \nincluding 28 cases of clear cell carcinoma and 16 c ases of papillary carcinoma, \nall confirmed by histopathological examination. Fiv e radiologists evaluated the \nscans twice: the first assessment was based solely on their prior knowledge, \nusing a cutoff of ≥4 for likelihood of clear cell carcinoma; the secon d \nassessment followed a training session that introdu ced the CCLS, where a \nscore of ≥4 was assigned for clear cell tumors and ≤1 for papillary tumors. A \nwashout period of 4 weeks was implemented between a ssessments. \nResults or Findings: Clear cell carcinoma overall sensitivity improved f rom \n56% (95%CI: 47-64) pre-training to 86% (95%CI: 79-9 1) post-training, with a \ncorresponding AUC for the ROC curve enhancing from 0.81 (95%CI: 0.76-\n0.86) to 0.85 (95%CI:0.81-0.90). Papillary carcinom a sensitivity increased from \n52% (95%CI:41-64) pre-training to 65% (95% CI: 54-7 5) post-training, with \nspecificity rising from 90% (95%CI:85-93) to 95% (9 5%CI:91-97). The AUC for \nthe ROC curve for papillary carcinoma rose signific antly from 0.79 (95%CI:73-\n85) to 0.89 (95%CI: 85-94). Additionally, the agree ment improved for clear cell \ntumors, with a K of Conger increasing from 0.293 (9 5%CI: 0.181-0.406) to \n0.594 (95%CI:0.469-0.718), and for papillary tumors , from 0.360 \n(95%CI:0.211-0.51) to 0.489 (95%CI:0.35-0.628). \nConclusion: Structured training and the application of the CCLS  significantly \nenhance the diagnostic accuracy of radiologists in identifying clear cell and \npapillary renal tumors on MRI, underscoring the imp ortance of targeted \neducation in improving radiological interpretations . \nLimitations: Limitations include monocentric design, small sampl e size, limited \nradiologist cohort, and no longitudinal follow-up, impacting generalizability and \nsustainability. \nFunding for this study: None \nEthics committee - additional information: IRB approved \nAuthor Disclosures:  \nFrancesco Montorsi: Nothing to disclose \nMichele Cosenza: Nothing to disclose \nGiulio Imperiale: Nothing to disclose \nAlessandro Larcher: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \nGiorgio Brembilla: Nothing to disclose \nUmberto Capitanio: Nothing to disclose \n \n \nDifferentiating solid from friable tumor thrombus i n renal cell carcinoma \nusing MRI ADC volumetric analysis \nP. Kowal¹, *Y. Mytsyk*², K. Ratajczyk¹, W. Bursiewi cz¹, M. Trzciniecki¹,  \nK. Marek-Bukowiec¹, J. Rogala¹; ¹Wrocław/PL, ²Lviv/ UA \n(mytsyk.yulian@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the utility of \nfirst-order radiomic features derived from MRI appa rent diffusion coefficient \n(ADC) maps using volumetric analysis in distinguish ing solid from friable \nthrombus in patients with renal cell carcinoma (RCC ). \n\n \n \nThursday \nAbstract-based Programme \n \n 109  \nMethods or Background: A cohort of 27 patients with conventional histologi c \nsubtype of RCC and tumor thrombus in the renal vein  or inferior vena cava \n(IVC) was included. All patients underwent surgical  intervention, comprising \nnephrectomy and thrombectomy, and received preopera tive abdominal MRI \nwith diffusion-weighted imaging sequences at b-valu es of 50, 200, 800 s/mm². \nThe ADC map was used for volumetric analysis, calcu lating various radiomic \nfirst-order features across the thrombus volume, in cluding ADC mean, median, \nrange, 10th percentile, 90th percentile, interquart ile range, entropy, kurtosis, \nskewness, uniformity, and variance. Tumor thrombi w ere histologically \nclassified as solid or friable, and associations be tween the radiomic features \nand thrombus consistency were analyzed. \nResults or Findings: Solid and friable tumor thrombi were identified in 51.9% \nand 48.1% of patients, respectively. Inverse associ ation noted between RCC \nthrombus cellularity and skewness (r=-0.799, p<0.00 1). No significant \ndifferences were observed in the mean values of ran ge, 90th percentile, \ninterquartile range, kurtosis, uniformity, and vari ance between groups. For \ndistinguishing solid from friable thrombus, the ADC  mean, median, and entropy \nshowed equal sensitivity (93%) and specificity (69% ), with entropy yielding the \nhighest area under the curve (AUC) at 0.808. Skewne ss demonstrated a \nsensitivity of 86% and specificity of 92%, with an AUC of 0.931. \nConclusion: In RCC patients with tumor thrombus in the renal ve in or IVC, \nvolumetric analysis of first-order radiomic feature s using ADC mapping \nfacilitates accurate differentiation between solid and friable thrombus variants. \nLimitations: The primary limitation of this study is that only c onventional \nhistologic subtype of RCC was included in the analy sis. \nFunding for this study: No funding. \nEthics committee - additional information: This study was approved by the \nLocal Bioethical Committee in the Research and Deve lopment Center, \nRegional Specialist Hospital in Wroclaw (no. KB/12/ 2021). All procedures \nconducted followed the ethical guidelines set by th e institutional and/or national \nresearch committee, adhering to the principles outl ined in the 1964 Helsinki \nDeclaration and its subsequent revisions, or equiva lent ethical standards. The \nwritten informed consent for enrolment in the study  was signed by all patients. \nAuthor Disclosures:  \nYulian Mytsyk: Nothing to disclose \nKrzysztof Ratajczyk: Nothing to disclose \nMaciej Trzciniecki: Nothing to disclose \nWiktor Bursiewicz: Nothing to disclose \nJoanna Rogala: Nothing to disclose \nPaweł Kowal: Nothing to disclose \nKarolina Marek-Bukowiec: Nothing to disclose \n \n \nDoes the Bosniak 2019 classification really provide  an objective \nassessment among the radiologists \nŞ. Evrimler, *E. Cigdem Karatayli*; Ankara/TR \n(elifcigdem95@hotmail.com) \n \nPurpose or Learning Objective: The Bosniak Classification system was \nrevised in 2019 to reduce subjectivity and prevent unnecessary nephron loss. \nIn this plot study, we aimed to evaluate the inter- reader agreement in Bosniak \nclassification among radiologists with different ex perience levels. \nMethods or Background: Out of 320 patients imaged between 01.2022 and \n04.2024 in our hospital, 12 patients were randomly selected among those with \npathology results, 22 patients were randomly select ed among those without \npathology results, a total of 34 patients were sele cted. Eight residents, two \nradiology specialists, and one abdominal radiologis t evaluated abdominal CT \nand MRI scans of these patients. All participants w ere trained on the 2019 \nBosniak classification prior to the assessment. The y assessed each criteria \n(septation presence, nodularity characteristics, co ntrast enhancement,etc.) \nseparately and ultimately identified the Bosniak ty pe. The abdominal \nradiologist's classifications served as the referen ce for Kappa analysis. \nParticipants also completed a questionnaire regardi ng difficulties in evaluating \nthe classification and their use of objective value s. \nResults or Findings: Agreement with the reference increased with radiolo gy \ntraining duration (min kappa: 0.47, max kappa: 0.79 , p<0.01). Substantial \nagreement was noted between specialists and the ref erence (kappa: 0.68-\n0.72, p<0.01). Agreement was better for non-measura ble parameters (e.g., \npresence of septation) compared to measurable ones (e.g., thickness of \nseptation). Survey results indicated the most chall enging parameter was \nseptation thickness, with 100% of participants unce rtain between categories 2 \nvs. 2F and 2F vs. 3. While 53% used objective measu rements during \nassessments, 81% relied on them as a guide rather t han exclusively. \nConclusion: Moderate-substantial agreement was found among radi ologists, \nimproving with experience. Thickness of septations was particularly confusing, \nespecially between certain categories. \nLimitations: The study's small sample size suggests further rese arch with \nlarger, diverse groups is needed. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: AEŞH-EK1-2024-54 numbered \nfile, Ethical Committee, Etlik City Hospital, Ankar a \n \nAuthor Disclosures:  \nElif Cigdem Karatayli: Nothing to disclose \nŞehnaz Evrimler: Nothing to disclose \n \n \n14:00-15:30 Research Stage 2 \nResearch Presentation Session: Oncologic \nImaging \nRPS 1016 \nRadiologic advances in hepatobiliary and \npancreatic cancer \n \nModerator \nS. De Vuysere; Leuven/BE  \n(sofie.devuysere@gmail.com) \n \n \n3D Fractal Dimension Analysis of CT Imaging for Mic rovascular Invasion \nprediction in Hepatocellular Carcinoma \nB. Song, *F. Che*; Chengdu/CN \n(chefeng2020@163.com) \n \nPurpose or Learning Objective: This study aimed to assess the potential role \nof 3-dimensional (3D) fractal dimension (FD) derive d from contrast-enhanced \ncomputed tomography (CT) images in predicting micro vascular invasion (MVI) \nin hepatocellular carcinoma (HCC) patients. \nMethods or Background: This retrospective study included 655 patients with  \nsurgically confirmed HCC from two medical centers ( training set: 406 patients; \ninternal test set: 170 patients; external test set:  79 patients). Box-counting \nalgorithms were employed to compute the 3D FD value s from portal venous \nphase images. Univariable and multivariable logisti c regression analyses were \nused to determine independent clinical and imaging predictors. Diagnostic \nperformance of the model was assessed using receive r operating \ncharacteristic (ROC) curve analysis. Recurrence-fre e survival (RFS) and \noverall survival (OS) curves were generated via Kap lan-Meier method and \ncompared using the log-rank test. \nResults or Findings: Patients with MVI-positive HCC exhibited significan tly \nhigher FD values compared to those with MVI-negativ e HCC (P< 0.01). The \ncombined model, integrating alpha-fetoprotein level , tumor size, number, and \nFD, demonstrated superior diagnostic performance fo r MVI prediction \ncompared to the clinical model alone, with AUCs of 0.80 (95% CI: 0.75, 0.88) \nand 0.75 (95% CI: 0.67, 0.83) for the internal test  set, and 0.83 (95% CI: 0.72, \n0.92) and 0.74 (95% CI: 0.61, 0.85) for the externa l test set. Patients predicted \nto have high-risk MVI showed worse RFS and OS outco mes than those \npredicted to have low-risk MVI (all P<0.05). \nConclusion: The 3D FD value significantly differed between MVI- positive and \nMVI-negative HCC patients. Integration of FD into t he clinical model enhances \nMVI prediction accuracy and may identify patients a t high risk. \nLimitations: Firstly, the retrospective nature of our study intr oduces inherent \nlimitations. Secondly, our study only analyzed 3D f ractal features on the PVP \nphase. \nFunding for this study: None \nEthics committee - additional information: Institutional Review Board \napproval was obtained by West China Hospital and He nan Provincial People’s \nHospital \nAuthor Disclosures:  \nBin Song: Nothing to disclose \nFeng Che: Nothing to disclose \n \n \nThe impact of pretreatment body composition on conv ersion surgery \nfeasibility and survival in pancreatic cancer patie nts undergoing \nneoadjuvant therapy \n*H. Y. Chen*, B-B. Chen; Taipei/TW \n \nPurpose or Learning Objective: This study aims to evaluate the impact of \npretreatment body composition, assessed via CT or M RI, on the likelihood of \nachieving conversion surgery and overall survival ( OS) in pancreatic cancer \npatients undergoing neoadjuvant therapy (NAT). \nMethods or Background: A retrospective analysis was conducted on 154 \npatients with pancreatic cancer receiving NAT befor e planned conversion \nsurgery between May, 2018 and February, 2024. Clini cal and laboratory data, \nincluding carcinoembryonic antigen (CEA) and carboh ydrate antigen 19-9 \n(CA19-9), were collected. Body composition paramete rs such as skeletal \n\n \n \nThursday \nAbstract-based Programme \n \n 110  \nmuscle volume (SM), skeletal muscle index, subcutan eous adipose tissue \n(SAT), and visceral adipose tissue (VAT) at the L3 vertebral level were \nautomatically calculated using the TotalSegmentator  software. Independent \npredictors of conversion surgery were identified us ing multivariable logistic \nregression, and Kaplan-Meier curves and Cox regress ion models assessed the \nimpact of these variables on OS. \nResults or Findings: Patients who underwent conversion surgery group had  \nsignificantly smaller tumor sizes (P=0.033), lower rates of vascular invasion \n(P=0.042), lower log-transformed CA19-9 levels (P=0 .002), higher SM \n(P=0.049), and higher VAT (P=0.048). Multivariate a nalysis revealed that log-\ntransformed CA19-9 (P=0.002) and higher SM (P=0.049 ) were independent \npredictors of conversion surgery. Conversion surger y (P<0.001) and higher \nSAT (P=0.001) were associated with better OS. SAT w as identified as an \nindependent predictor of OS (P=0.04) after adjustin g for age, conversion \nsurgery status, and TNM stage in the multivariate m odel. \nConclusion: Pretreatment body composition, as evaluated through  imaging, \nmay be associated with the feasibility of conversio n surgery and overall \nsurvival in patients with pancreatic cancer receivi ng NAT. Specifically, higher \nskeletal muscle volume and SAT appear to be favorab le prognostic factors, \nhighlighting the importance of body composition in clinical decision-making and \npatient outcomes. \nLimitations: The limitations of the study are single center stud y and relatively \nsmall patient group. \nFunding for this study: No funding was received for this study \nEthics committee - additional information: Research Ethics Committee of \nNational Taiwan University Hospital \nAuthor Disclosures:  \nBang-Bin Chen: Nothing to disclose \nHsin Yu Chen: Nothing to disclose \n \n \nIodine quantification and LI-RADS classification of  hepatocellular \ncarcinoma lesions in contrast-enhanced spectral CT studies \n*A. Celestino*, P. Marra, A. Barbaro, C. Gargiulo, R. Muglia, G. Muscogiuri,  \nP. A. Bonaffini, S. Sironi; Bergamo/IT \n(a.celestino1@campus.unimib.it) \n \nPurpose or Learning Objective: The LI-RADS classification is universally \nemployed for hepatocellular carcinoma (HCC) risk st ratification of liver nodules \nin cirrhosis, but it relies on a qualitative evalua tion. This study aims to \ninvestigate the potential role of material density (MD) parameters in the iodine \nmaps of Spectral Computed Tomography (SCT), to disc riminate between LI-\nRADS categories in cirrhotic patients, therefore in creasing the radiologists’ \nconfidence in LI-RADS class allocation. \nMethods or Background: Dual-energy SCT scans of cirrhotic patients with \nnodules between March 2022 and September 2023 were retrospectively \nreviewed. All the images were reviewed by trained r adiologists to classify \nnodules as LI-RADS 3, 4 or 5 by consensus. MD maps were generated in the \nhepatic arterial phase (HAP), portal venous (PVP) a nd equilibrium phase (EP). \nIodine concentration density (ICD) of nodules (ICDn odule) and non-nodular \nliver parenchyma (ICDliver) were measured to calcul ate lesion-to-non-nodular \nliver ICD ratio (LNR), their differences (ΔICD) and ratios (rLNR). Results were \ncorrelated with LI-RADS categories. \nResults or Findings: 69 patients were included and 79 DECT exams were \nexamined. 197 nodules (size 24.67 ± 23.11 mm, mean ± SD) were categorised \nunder different LI-RADS classes, as follows: 44 as LI-RADS 3 (22.3%), 14 as \nLI-RADS 4 (7.1%), and 139 as LI-RADS 5 (70.6%). Art erial LNR, arterial \nICDnodule, ΔICD and rLNR between HAP and PVP could discriminate  \nbetween LI-RADS 3 and LI-RADS 4+5 nodules (p < 0.00 1). All the calculated \nMD parameters showed similar, or slightly higher di agnostic accuracy rates (all \nAUCs = 70-73%) compared to those previously reporte d by non-spectral CT \n(up to 70%). \nConclusion: MD parameters of liver nodules measured in SCT scan s are \nviable diagnostic tools that may increase the radio logist’s confidence in LI-\nRADS class allocation in cirrhotic patients. \nLimitations: The lack of an adequate number of LI-RADS 4 nodules  \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nAntonio Celestino: Nothing to disclose \nSandro Sironi: Nothing to disclose \nPaolo Marra: Nothing to disclose  \nPietro Andrea Bonaffini: Nothing to disclose \nAlessandro Barbaro: Nothing to disclose \nRiccardo Muglia: Nothing to disclose \nCarlotta Gargiulo: Nothing to disclose \nGiuseppe Muscogiuri: Nothing to disclose \n \n \n \n \nSimple Cystic Lesions of the Pancreas: Image Qualit y and Diagnostic \nAccuracy of Photon-Counting Detector Computed Tomog raphy \n*S. Rau*¹, T. Stein¹, A. Rau¹, C. Wilpert¹, F. B. P allasch¹, B. Bogner¹, S. Faby², \nJ. Weiß¹; ¹Freiburg/DE, ²Forchheim/DE \n(stephan.rau@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: To evaluate image quality and diagnostic \naccuracy of photon-counting detector (PCD) computed  tomography (CT) for \nthe detection of PCLs compared to conventional ener gy-integrating detector \n(EID) CT with MRI serving as reference standard. \nMethods or Background: In this prospective study, we included consecutive \npatients who underwent clinically indicated contras t-enhanced PCD-CT of the \nabdomen and for whom an additional abdominal EID-CT  was available. \nMultiparametric MRI served as the reference standar d. CT images were \nassessed for the presence of PCLs by three radiolog ists independently in a \nblinded reading. In addition, image quality, lesion  conspicuity, and diagnostic \nconfidence were rated on a 5-point Likert scale (5= excellent). The coefficient-\nof-variation (CV) and the density difference (in Ho unsfield units [HU]) between \nPCLs and visually normal pancreatic parenchyma were  calculated as \nquantitative imaging measures. Radiation dose was a ssessed using CTDIvol \n[mGy]. \nResults or Findings: Among 106 included patients (age 62.7±12.6 years; 4 5 \n[42.5%] male), 46 had MRI-confirmed cystic lesions (mean size 8.7±7.4mm; \nrange: 2-45 mm). Diagnostic accuracy for PCLs was s ignificantly higher for \nPCD-CT vs. EID-CT (area under the curve: 0.81 vs. 0 .74; p=0.002; sensitivity: \n76.8% vs. 59.4%; specificity 84.4% vs. 88.3%, respe ctively). Image quality, \nlesion conspicuity and diagnostic confidence were r ated superior for PCD-CT \nvs. EID-CT (all p<0.001). Quantitative analyses rev ealed a significantly lower \nCV (0.19 vs. 0.24; p=0.002) and a higher density di fference (94.1 HU vs. 76.6 \nHU p<0.001) between PCLs and visually normal pancre atic parenchyma at \nlower radiation doses (7.13 vs. 8.68 mGy; p<0.001) for PCD-CT vs. EID-CT. \nConclusion: PCD-CT provided significantly higher diagnostic acc uracy and \nsuperior image quality for the detection of PCLs co mpared to conventional \nEID-CT at lower radiation dose. \nLimitations: No long-term follow-up and/or histopathological cor relation of the \ndetected PCLs were omitted. \nFunding for this study: None \nEthics committee - additional information: The local Institutional Review \nBoard (Ethics Committee of the University Medical C enter Freiburg, case \nnumber 21-2469) approved this prospective study and  written informed \nconsent was obtained from all patients prior to stu dy inclusion. \nAuthor Disclosures:  \nCaroline Wilpert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nAlexander Rau: Nothing to disclose \nBalazs Bogner: Nothing to disclose \nFabian Bernhard Pallasch: Nothing to disclose \nSebastian Faby: Employee: Siemens Healthineers \nThomas Stein: Nothing to disclose \nStephan Rau: Nothing to disclose \n \n \nPrognostic Value of RECIST, mRECIST, and LI-RADS TR A Early \nResponse in Predicting Survival in Hepatocellular C arcinoma Treated \nwith Selective Internal Radiation Therapy \nM. Dupuis, A. Dupont, S. Pizza, V. Vilgrain, A. Ban do Delaunay, R. Lebtahi,  \nM. Bouattour, M. Ronot, *J. Gregory*; Clichy/FR \n(jules.gregory@aphp.fr) \n \nPurpose or Learning Objective: This study evaluates the prognostic value of \ntumor response at three months on CT, assessed by R ECIST, mRECIST, and \nLI-RADS Treatment Response Algorithm (LR-TRA) in pa tients with \nhepatocellular carcinoma (HCC) treated with selecti ve internal radiation \ntherapy (SIRT). \nMethods or Background: A retrospective analysis was conducted on 102 \nHCC patients treated with SIRT between 2018 and 202 0. RECIST, mRECIST, \nand LR-TRA were assessed at 3 months post-SIRT. Ove rall survival (OS) and \nprogression-free survival (PFS) were assessed using  Kaplan-Meier analysis \nand Cox proportional hazards models. \nResults or Findings: Median age was 71 years, most patients (90%) had \nadvanced-stage tumors (BCLC-C). After a median foll ow-up of 32.0 months \n(95% CI: 16.8-60.9), 60/102 patients died (59%), an d 90/102 patients showed \ntumor progression (88%). Median OS was 20.4 months (95% CI: 15.4-33.0), \nand median PFS was 14.5 months (95% CI: 6.5-24.5); 1-year OS and PFS \nrates were 65.6% and 50.7%. Multivariable analysis revealed that early \nresponse according to RECIST 1.1 (HR 1.66, p=0.30),  mRECIST (HR 1.40, \np=0.215), and LR-TRA (HR 0.67, p=0.30) were not pre dictors of OS. Disease \nprogression evaluated by RECIST (HR 2.55, p<0.001) and mRECIST (HR \n2.53, p<0.001), bilirubin levels (HR 1.03, p<0.001)  and prothrombin time (HR \n0.98, p=0.005) were predictors of OS. For PFS, neit her RECIST nor mRECIST \nresponse, disease progression, nor LR-TRA viability  were predictors. \n\n \n \nThursday \nAbstract-based Programme \n \n 111  \nConclusion: In this advanced-stage HCC population, early respon se \nassessed by RECIST, mRECIST, and LR-TRA criteria di d not predict OS or \nPFS after SIRT. However, early disease progression and liver function \nindicators were prognostic factors for OS. \nLimitations: Several limitations exist in this single-center, re trospective study, \nincluding a small sample size, which may reduce the  generalizability of the \nfindings. Additionally, tumor heterogeneity, a key prognostic-factor for poor \ntreatment response and shorter PFS, was not conside red. \nFunding for this study: None \nEthics committee - additional information: This single-center retrospective \nclinical study was approved by the local Institutio nal Review Board (IRB \n00006477), and informed consent was waived due to i ts retrospective nature. \nAuthor Disclosures:  \nAxelle Dupont: Nothing to disclose \nMichel Dupuis: Nothing to disclose \nMohamed Bouattour: Nothing to disclose \nJules Gregory: Nothing to disclose \nMaxime Ronot: Nothing to disclose \nSilvia Pizza: Nothing to disclose \nRachida Lebtahi: Nothing to disclose \nValérie Vilgrain: Nothing to disclose \nAurélie Bando Delaunay: Nothing to disclose \n \n \nContrast-enhanced Ultrasound using Perfluorobutane for Diagnosing \nSmall HCC (≤20mm) in High-risk Patients: Comparison with MRI LI -RADS \nVersion 2018 \n*J. Zhou*, Y. Li, L. Lii; Guangzhou/CN \n(zhoujh@sysucc.org.cn) \n \nPurpose or Learning Objective: The sensitivity of Contrast-enhanced \nultrasound (CEUS) for diagnosing hepatocellular car cinoma (HCC) is lower \nthan MRI, especially in small liver nodules measuri ng 20mm or less. This study \naimed to compare the diagnostic performance between  CEUS with \nperfluorobutane (P-CEUS) and MRI Liver Imaging Repo rting and Data System \n(LI-RADS) version 2018 (v2018) for small liver nodu les in high-risk patients. \nMethods or Background: This multi-center retrospective study included high -\nrisk patients with newly detected liver nodules mea suring 20mm or less from \nMarch 2020 to November 2023. Patients underwent CEU S with \nperfluorobutance (P-CEUS) and concurrent MRI at int ervals of no more than 1 \nmonth. The reference standard was pathological conf irmation or 24-month \nimaging follow-up (only for benign lesions). The di agnostic performance of \nCEUS LI-RADS v2017 for P-CEUS, modified criteria fo r P-CEUS, and MRI LI-\nRADS v2018 was calculated and compared. For the mod ified criteria for P-\nCEUS, LR-4 observations measuring 10 mm or larger w ith nonrim APHE and \nno washout were reclassified as LR-5 if aslo showin g a Kupffer defect; LR-M \nobservations measuring 10 mm or larger with nonrim APHE and early washout \nwere reclassified as LR-5 if aslo showing a mild Ku pffer defect. \nResults or Findings: A total of 367 participants (301 males; mean age, 5 4±11 \nyears) with 401 observations (mean diameter, 15±4 m m) were included. Using \nLR-5 for HCC daignosis, MRI LI-RADS v2018 had highe r sensitivity (73% vs \n57%, P<0.001) with no significant lower specificity  (90% vs 94%, P=0.18) \ncompared with CEUS LI-RADS v2017. The modified crit eria for P-CEUS and \nMRI LI-RADS v2018 showed no significant difference (P>0.05) in sensitivity \n(70% vs 73%) and specificity (92% vs 90%). \nConclusion: CEUS using perfluorobutane with modified criteria d emonstrated \na diagnostic performance comparable to MRI in HCC d iagnosis for small liver \nnodules (≤20mm). \nLimitations: Retrospective study \nFunding for this study: This study was not supported by any funding. \nEthics committee - additional information: Approved \nAuthor Disclosures:  \nLingling Lii: Nothing to disclose \nJianhua Zhou: Nothing to disclose \nYu Li: Nothing to disclose \n \n \nEvaluation of dual-layer spectral CT compared with conventional CT for \nthe diagnosis of hepatocellular carcinoma \n*M. Youssef Francis*, T. Broussaud, M. Wagner, J. B enzimra,  \nN. Brillat-Savarin, O. Lucidarme; Paris/FR \n(marieyousseffrancis@hotmail.com) \n \nPurpose or Learning Objective: To investigate whether the use of dual-layer \nspectral CT (DLSCT) improves the diagnostic accurac y of hepatocellular \ncarcinomas (HCCs) compared with conventional CT. \n \n \n \n \n \nMethods or Background: 127 patients were included in this retrospective \nstudy. All patients underwent multiphase DLSCT (IQo n, Philips Healthcare) for \nthe initial diagnosis of HCC. Arterial phase hypere nhancement (APHE) and \nportal and delayed washout of each lesion were qual itatively assessed by two \nradiologists independently using three-point Likert  scales and the lesion-to-liver \ncontrast-to-noise ratio (LLCNR) was quantified usin g ROI in conventional (CIs) \nand 40keV virtual monoenergetic images (VMIs). For qualitative analysis, \ninterobserver agreement was assessed using the kapp a statistic. For \nquantitative analysis, LLCNRs were compared using W ilcoxon and Scheirer-\nRay-Hare tests. \nResults or Findings: 163 HCCs were independently analyzed. The APHE \nwas considered significantly higher for both observ ers in VMIs compared to \nCIs: mean 2.36 vs 1.32 p<0.001. The washout was con sidered significantly \nbetter seen in VMIs compared to CIs in the delayed phase: mean 1.99 vs 1.33, \np<0.001; but not in the portal phase: mean 1.74 vs 1.36, NS. Interobserver \nagreement was good for APHE and washout in both CIs  and VMIs. The \nLLCNR was significantly higher in the arterial phas e (6.18vs2.05;p<0.001) and \nsignificantly lower in the delayed phase (-2.02 vs -1.10; p<0.001) in 40 Kev-\nVMIs compared to CIs. As in the subjective analysis , LLCNR was not \nsignificantly different in the portal phase (-1.89 vs -1.52;NS). \nConclusion: Multiphasic DLCT with 40 keV VMIs increases the vis ibility of \nboth APHE and washout in the delayed phase of focal  liver lesions compared \nto CIs, leading to the final diagnosis of a higher number of HCCs. \nLimitations: This is a single-center, retrospective study. DLSCT  was not \ncompared with MRI, which is currently the gold stan dard in the detection of \nHCC. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was IRB approved \n(number CRM-2304-339). \nAuthor Disclosures:  \nOlivier Lucidarme: Nothing to disclose \nMarie Youssef Francis: Nothing to disclose \nThomas Broussaud: Nothing to disclose \nMathilde Wagner: Nothing to disclose \nNina Brillat-Savarin: Nothing to disclose \nJulie Benzimra: Nothing to disclose \n \n \nBone Mineral Density as a Prognostic Indicator for Overall Survival in \nHepatocellular Carcinoma Patients with Portal Venou s Tumor Thrombus \nL. Müller¹, R. Kloeckner², L. Heim¹, F. Stoehr¹, F.  Hahn¹, T. Bäuerle¹,  \nA. Weinmann¹, D. Pinto Dos Santos³, *A. Mähringer-K unz*¹; ¹Mainz/DE, \n²Lübeck/DE, ³Frankfurt/DE \n(aline.maehringer.kunz@googlemail.com) \n \nPurpose or Learning Objective: Low bone mineral density (BMD) has \nrecently emerged as a risk factor in hepatocellular  carcinoma (HCC). However, \nits role in patients with HCC complicated by portal  vein tumor thrombosis \n(PVTT) is unclear. This study explores the potentia l of BMD as a prognostic \nindicator within this subgroup, which is characteri zed by an exceptionally poor \nprognosis. \nMethods or Background: This retrospective study included 462 patients with  \nHCC and PVTT diagnosis at our tertiary care center between January 2005 \nand December 2020. BMD was measured by mean Hounsfi eld units (HUs) at \nthe midvertebral core of the first lumbar vertebra in computed tomography \nusing the established cut-off of 160 HU. Analysis w as performed at two points \nin time: initial HCC diagnosis and PVTT onset. We a nalyzed the impact of BMD \non median overall survival (OS) and conducted multi variate analysis with \nestablished survival predictors. \nResults or Findings: Median BMD was 136 HU (IQR, 113–160 HU) at HCC \ndiagnosis and 134 HU (IQR, 109–159 HU) at PVTT diag nosis. At initial HCC \ndiagnosis, patients with BMD ≥ 160 HU had a median OS of 10.4 months, \ncompared to 5.5 months in those with BMD < 160 HU ( p < 0.001). At PVTT \ndiagnosis, those with higher BMD had a median OS of  8.5 months, versus 4.7 \nmonths in patients with lower BMD (p < 0.001). BMD remained an independent \nprognostic factor in multivariate analysis, alongsi de growth type and ALBI \ngrade. \nConclusion: BMD serves as an independent prognostic marker for survival in \npatients with HCC and PVTT. Incorporating BMD into existing classification \nand scoring systems could enhance the accuracy of s urvival predictions and \ninform clinical decision-making processes. \nLimitations: The primary limitation is the retrospective study d esign, \nnecessitating validation of these findings in a pro spective framework. \nFunding for this study: None \nEthics committee - additional information: The study was approved by the \nresponsible ethical body: Ethics committee of the M edical Association of \nRhineland Palatinate, Mainz, Germany (Permit number : 15913). \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 112  \nAuthor Disclosures:  \nDaniel Pinto Dos Santos: Other: Chair of the ESR eH ealth and Informatics \nSubcommittee \nLorena Heim: Nothing to disclose \nFabian Stoehr: Grant Recipient: Research grant, cli nician scientist program, \nElse Kröner-Fresenius-Foundation \nAline Mähringer-Kunz: Nothing to disclose  \nFelix Hahn: Nothing to disclose \nRoman Kloeckner: Speaker: Astra Zeneca, Boston Scie ntific, , BTG, EISAI, \nGuerbet, Ipsen, Siemens, and SIRTEX Other: Chair of  the ESR Audit and \nStandards Subcommittee Advisory Board: Boston Scien tific, Bristol-Myers \nSquibb, Guerbet, MSD Sharp & Dohme, Roche, and SIRT EX \nLukas Müller: Grant Recipient: Research grant, clin ician scientist program, Else \nKröner-Fresenius-Foundation \nTobias Bäuerle: Nothing to disclose \nArndt Weinmann: Nothing to disclose \n \n \nSolid pancreatic neoplasms: bridging radiology and cytopathology for \naccurate diagnosis \n*D. J. A. D. C. E. Aragão*, S. Santos, J. Nobre, L.  M. Cabral; Lisboa/PT \n(diogoaragao97@gmail.com) \n \nPurpose or Learning Objective: To retrospectively evaluate a cohort of \npatients with solid pancreatic lesions who underwen t endoscopic ultrasound-\nguided fine-needle aspiration (EUS-FNA). To review the radiologic features of \ncommon solid pancreatic neoplasms and correlate the m with cytopathological \nfindings. \nMethods or Background: A retrospective review was performed at our \ninstitution, encompassing 100 EUS-FNA procedures du ring 18 consecutive \nmonths. Patients with solid pancreatic lesions or l esions with a solid \ncomponent (n=48) were included in the analysis. Cli nical data, imaging studies, \nand cytopathology reports were collected. The most illustrative cases were \nselected to demonstrate the radiologic and cytopath ologic characteristics of \nthese neoplasms. \nResults or Findings: The most frequently diagnosed solid pancreatic \nneoplasm was adenocarcinoma (median age: 70 years; 61% male; 71% \nlocated in the pancreatic head), typically presenti ng as a hypovascular mass. \nNeuroendocrine tumors, often hypervascular, were th e second most common \nneoplasm. Rare cases included one case of pancreati c lymphoma and one \ncase of metastatic disease involving the pancreas. \nConclusion: As expected, pancreatic adenocarcinoma, particularl y in elderly \nmales and localized in the pancreatic head, was the  most commonly \ndiagnosed neoplasm. EUS-FNA combined with cytopatho logic analysis, along \nwith multimodality imaging, remains essential in th e accurate diagnosis, \nstaging, and management of solid pancreatic neoplas ms. The integration of \nthese diagnostic tools ensures a comprehensive mult idisciplinary approach, \nleading to more effective patient care and treatmen t planning. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study is educational. \nAuthor Disclosures:  \nSofia Santos: Nothing to disclose \nLuís Monteiro Cabral: Nothing to disclose \nJoão Nobre: Nothing to disclose \nDiogo José Afonso Da Cruz E Aragão: Nothing to disc lose \n \n \nImproving Diagnostic Confidence in Assessing Pancre atic Tissue: \nProspective Evaluation of Mechanical Elastography \n*V. Koch*¹, M. Cimprich¹, L. D. Grünewald¹, C. Booz ¹, T. Vogl¹, O. Darwish²,  \nJ. Gotta¹, S. Mahmoudi¹, R. Sinkus³; ¹Frankfurt/DE,  ²London/UK, ³Paris/FR \n \nPurpose or Learning Objective: This study aimed to evaluate 2D/3D \nmagnetic resonance imaging (MRE) utilizing the grav itational transducer \nconcept compared to the current acoustic product so lution (2D-MRE \nResoundant) to further characterize pancreatic carc inoma and its potential to \nprovide imaging biomarkers for outcome prediction. \nMethods or Background: In this prospective study, 40 patients with confirm ed \npancreatic cancer undergoing MRI of the upper abdom en were enrolled \nbetween June 2023 and September 2024. Furthermore, 15 healthy volunteers \nwere included as a healthy reference standard. All participants underwent two \nexaminations using a 40Hz mechanical vibration freq uency (Aera 1.5T, \nSiemens Healthineers, Germany): initially with the acoustic MRE (Resoundant, \n2D-MRE, SE-EPI sequence, 11s breath hold [BH]), and  subsequently with the \ngravitational MRE (2D-MRE and 3D-MRE, GRE sequence,  TE=9.53ms [in-\nphase], and fractional motion encoding at 30mT/m, 1 4s BH). Two experienced \nreaders independently conducted data analysis. Addi tionally, superimposed \nanalytic plane waves with known properties at vario us amplitudes and temporal \nnoise levels were employed to investigate biases in  stiffness reconstruction \n(2D/3D) and suggest Quality Indices for 2D/3D. \nResults or Findings: Significant differences were observed in stiffness values, \nshear wave speed, and phase angle between healthy v olunteers and patients \nwith pancreatic cancer using both MRE approaches (p <.05). Additionally, \npatients who showed clinical response to chemothera py exhibited differences \nin stiffness (p<.05). However, the Bland-Altman plo ts exhibited a notable bias, \nwith 2D-MRE tending to overestimate stiffness value s. 3D-MRE provided \nseveral imaging biomarkers that correlated with dis ease progression and \nresponse to therapy. Proposed Quality Indices enabl ed the identification of \npixels exhibiting a deviation exceeding 10% from ac tual stiffness values in 3D-\nMRE. \nConclusion: Gravitational MRE proves to be an accurate techniqu e for \nnoninvasively characterizing pancreatic tissue. In particular, 3D MRE can \nprovide pertinent functional imaging markers, advan cing functional abdominal \nimaging. \nLimitations: Single-center study. \nFunding for this study: This study has been funded by the Doktor Robert \nPfleger Foundation. \nEthics committee - additional information: The institutional ethical review \nboard approved this prospective study. Written info rmed consent was obtained \nfrom all participants. \nAuthor Disclosures:  \nChristian Booz: Nothing to disclose \nOmar Darwish: Nothing to disclose \nThomas Vogl: Nothing to disclose \nVitali Koch: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nRalph Sinkus: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \nJennifer Gotta: Nothing to disclose \nMarina Cimprich: Nothing to disclose \n \n \nImaging features, management and outcomes of Solita ry Necrotic Nodule \nof the Liver: a case series and review of literatur e \nL. Asmundo, L. Giaccardi, A. Soro, C. Buonomenna, R . Vigorito, F. G. Greco, \nA. Casale, *M. Vaiani*; Milan/IT \n(marta.vaiani@gmail.com) \n \nPurpose or Learning Objective: Solitary Necrotic Nodule of the Liver (SNNL) \nis a rare, benign hepatic lesion frequently misdiag nosed as malignant, leading \nto unnecessary invasive procedures. This study pres ents a case series of \npatients diagnosed with SNNL, focusing on their ima ging characteristics, \nclinical management and outcomes. \nMethods or Background: This retrospective case series analyzed data from \npatients diagnosed with SNNL at a tertiary referral  center. Collected data \nincluded demographics, imaging studies, and follow- up outcomes. A radiologist \nwith 20 years of experience reviewed all imaging st udies. The reference \nstandard was histological examination or follow-up imaging \nResults or Findings: Among 13 patients (54% female; median age 54 years) , \nMRI was the preferred imaging modality, with 3 to 5  scans performed over a \nmedian follow-up of 25 months (range 22-70). SNNL t ypically presented as a \nnecrotic core surrounded by a fibrotic capsule, oft en with a nodular, elongated, \nor C-shaped appearance that mimicked biliary dilata tion. T1-weighted images \nshowed isointensity in 53.8% of cases, with hypoint ensity of the core in 84.6%. \nT2-weighted images revealed isointensity (38.5%) or  hypointensity (61.5%) of \nthe core. Nodules remain hypovascular after contras t media injection. \nDiffusion-weighted imaging displayed no restricted diffusion. During follow-up, \nmost nodules (92.3%) showed a reduction in size, wi th increased calcification \non CT (from 38.5% to 69.2%). \nConclusion: SNNL presents diagnostic challenges due to its rese mblance to \naggressive hepatic lesions, particularly cholangioc arcinoma and metastases. \nHowever, careful interpretation of MRI and CT findi ngs can prevent \nunnecessary invasive procedures. \nLimitations: The small sample size and single-center design limi t the \ngeneralizability of the findings. While only a few cases had histological \nconfirmation (30%), long-term imaging follow-up sup ports the benign nature of \nSNNL, reducing the need for biopsies and surgeries.  \nFunding for this study: none. The authors did not receive support from any \norganization \nEthics committee - additional information: Local ethic committee approved \nthe study protocol \nAuthor Disclosures:  \nFrancesca Gabriella Greco: Nothing to disclose \nLuca Giaccardi: Nothing to disclose \nCiriaco Buonomenna: Nothing to disclose \nRaffaella Vigorito: Nothing to disclose \nAlberto Soro: Nothing to disclose \nLuigi Asmundo: Nothing to disclose \nMarta Vaiani: Nothing to disclose \nAlessandra Casale: Nothing to disclose \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 113  \n14:00-15:30 Research Stage 3 \nResearch Presentation Session: \nInterventional Radiology \nRPS 1009 \nAdvances in musculoskeletal and soft \ntissue interventions \n \nModerator \nM. Szmygin; Lublin/PL  \n(mszmygin@gmail.com) \n \n \nEfficacy and safety of image-guided bone biopsies: Insights gained from \nthe German Society for Interventional Radiology and  Minimally Invasive \nTherapy (DeGIR) registry data 2018-2022 \n*S. Zensen*¹, F. Behr¹, M. Opitz¹, D. Bos¹, M. Holt kamp¹, L. Salhöfer¹,  \nJ. Haubold¹, M. Moche², B. M. Schaarschmidt¹; ¹Esse n/DE, ²Leipzig/DE \n(sebastian.zensen@uk-essen.de) \n \nPurpose or Learning Objective: Image-guided bone biopsies are critical for \nthe diagnosis of a wide range of bone lesions, yet there is limited large-scale \ndata on the technical success, diagnostic yield, an d safety of these \nprocedures. This study aims to evaluate the efficac y and safety of image-\nguided bone biopsies using data from the German Soc iety for Interventional \nRadiology and Minimally Invasive Therapy (DeGIR, De utsche Gesellschaft für \nInterventionelle Radiologie und minimal-invasive Th erapie) registry. \nMethods or Background: In this retrospective observational study, 17,397 \nbone biopsies from 214 centers between 2018 and 202 2 were analyzed. \nTechnical success was defined as the visually succe ssful placement of the \nbiopsy needle within the target lesion. Complicatio ns were classified according \nto the Society of Interventional Radiology (SIR) gu idelines. \nResults or Findings: About one-third of biopsies were performed as \noutpatient procedures (34%,5,924/17,397). Most biop sies were conducted \nunder local anesthesia (86.6%,15,072/17,397). CT-gu idance was used in the \nmajority (68.7%,11,952/17,397). The technical succe ss rate was 98.9% \n(17,201/17,397), with histological representativene ss of 93.2% \n(10,316/11,073). Outpatient biopsies had a slightly  higher technical success \nrate (99.32%,5,884/5,924) than inpatient biopsies ( 98.63%,11,316/11,473), \nthough histological representativeness was lower (9 1.06%,1,284/1,410 vs. \n93.48%,9,031/9,661, p=0.001). The overall complicat ion rate was low at 0.62% \n(108/17,397), with major complications in 23.1% (25 /108) of cases. Patients \nwith abnormal coagulation parameters had higher com plication rates. Sub-\n/solid lesions had higher histological representati veness compared to necrotic-\ncystic lesions (94.01%,7,846/8,346 vs. 90.32%,1,558 /1,725, p<0.0001). \nConclusion: Image-guided bone biopsies are highly effective and  safe, even \nin outpatient settings. The data support their cont inued use as a minimally \ninvasive diagnostic tool, with low complication rat es and high diagnostic \naccuracy. \nLimitations: This study is limited by the nature of registry dat a, which is \nsubject to reporting biases and lacks external vali dation. Incomplete data from \nsome centers also restricted the scope of certain a nalyses. \nFunding for this study: This research received no specific grant from any \nfunding agency in the public, commercial, or not-fo r-profit sectors. \nEthics committee - additional information: Ethical approval for this \nretrospective registry study was granted by the eth ics committee of the \nUniversity of Duisburg-Essen, Germany (22-10893-BO) . \nAuthor Disclosures:  \nDenise Bos: Nothing to disclose \nBenedikt Michael Schaarschmidt: Nothing to disclose  \nJohannes Haubold: Nothing to disclose \nSebastian Zensen: Nothing to disclose  \nLuca Salhöfer: Nothing to disclose \nFlorian Behr: Nothing to disclose \nMichael Moche: Nothing to disclose \nMarcel Opitz: Nothing to disclose \nMathias Holtkamp: Nothing to disclose \n \n \n \n \n \n \n \n \nEarly results of a prospective study on palliative arterial embolization for \nbone metastases: the EMBONEMET study \n*N. Papalexis*, G. Peta, S. Quarchioni, M. Di Carlo , L. Campanacci, M. Carta, \nM. Miceli, G. Facchini; Bologna/IT \n(nicolaspapalexis@gmail.com) \n \nPurpose or Learning Objective: To evaluate the clinical and radiological \neffect of arterial embolization using N-butyl cyano acrylate (NBCA) as palliation \nfor bone metastases. \nMethods or Background: This study analyzes the early results of a \nprospective study “EMBONEMET”, designed to prospect ively evaluate the \nsafety and efficacy of palliative arterial emboliza tion for bone metastases. \nThirty-three patients were enrolled from June 2023 to June 2024. The primary \ngoal was pain control, measured in VAS score at 3, 6 and 12 months follow-up. \nThe secondary goal was the size reduction of the le sion. Data on technical \nsuccess were also collected. All embolization-relat ed complications were \nevaluated according to the CIRSE classification sys tem for complications. \nResults or Findings: The average follow-up was 9 months (range 2 to 12 \nmonths). Baseline VAS scores averaged 5,3 (SD 3,19) , decreasing to 3.56 (SD \n3,1) at three months. Progressive reduction was obs erved at 6 and 12 months \nwith scores of 3,52 (SD 3,33), and 2,8 (SD 2,7) res pectively. Metastatic tumor \nsize was reduced from a mean of 196.8 cm3 (range 39 .5 to 486.4 cm3) pre-\nembolization to a mean of 179.4 cm3 (range 38.2 to 458.6 cm3) at the 6-month \nfollow-up (p<0.05). Twenty-nine patients experience d post embolization-related \npain, that resolved within 15 days. Three patients experienced sensory loss of \nthe lower leg, paresthesia and pain in the femoral catheter access point. \nConclusion: The preliminary results are promising, suggesting t hat arterial \nembolization could be a safe and effective tool for  pain management and \ndisease control in metastatic bone disease. \nLimitations: Small sample size, lack of control group \nFunding for this study: None \nEthics committee - additional information: Prospective study approved by \nthe local ethics committee of Emilia Romagna, Italy . \nAuthor Disclosures:  \nMarco Miceli: Nothing to disclose \nMichela Carta: Nothing to disclose \nLaura Campanacci: Nothing to disclose \nGiancarlo Facchini: Nothing to disclose \nGiuliano Peta: Nothing to disclose \nSimone Quarchioni: Nothing to disclose \nMaddalena Di Carlo: Nothing to disclose \nNicolas Papalexis: Nothing to disclose \n \n \nPercutaneous cryoablation of progressing extra-abdo minal desmoid \ntumours \n*A. Vanzulli*, L. V. Sciacqua, L. Saggiante, T. Cas cella, C. Colombo,  \nE. Palassini, C. Morosi, S. Stacchiotti, A. Gronchi ; Milan/IT \n(andrea.vanzulli@unimi.it) \n \nPurpose or Learning Objective: To evaluate the safety and efficacy of \npercutaneous cryoablation for the treatment of extr a-abdominal desmoid \ntumours (DT) progressing after active surveillance/ first-line treatments or \nthreatening to life/function/quality-of-life. \nMethods or Background: We retrospectively evaluated baseline and post-\ntreatment MR and CT imaging of 27 consecutive proce dures performed at our \nInstitution between May 2021 and July 2024. Treatme nt response was \nassessed both with standard and modified (m-) RECIS T 1.1 (employing also \nT2WI and DWI to define tumour viability), with the underlying premise that \ndimensional reduction alone does not adequately cap ture tissue viability and \nmay incompletely depict responses to local treatmen ts. Three different \ntimepoints were considered: approximately 40 days a fter treatment, 5-10 \nmonths after treatment and 11-16 months after treat ment. \nResults or Findings: The study cohort comprised 25 patients (21 females and \n4 males; median age at treatment: 36 years; age ran ge 13-59 years) affected \nby extra-abdominal treatment eligible for local abl ative therapies. Tumour \nlocations included the head & neck (3), the thoraci c wall (4), the abdominal wall \n(19) and the lower extremities (1), with an average  maximum diameter of 81,2 \nmm (range 32-162,3 mm). At approximately 40 days af ter treatment, mRECIST \nresponses were: 2/19 (10,5%) Stable Disease (SD), 6 /19 (31,6%) Partial \nResponse (PR) and 11/19 (57,9%) Complete Response ( CR). At 5-10 months \nafter treatment, mRECIST responses were: 9/18 (50%)  SD, 2/18 (11,1%) PR \nand 7/18 (38,9%) CR. At 11-16 months after treatmen t, mRECIST responses \nwere: 6/16 (37,5%) SD, 4/16 (25%) PR and 6/16 (37,5 %) CR. At 11-16 months \nafter treatment, standard RECIST responses were: 1/ 16 (6,3%) Progressive \nDisease (PD), 14/16 (87,5%) SD and 1/16 (6,3%) PR. \nConclusion: Percutaneous cryoablation represents a feasible tre atment for \nextra-abdominal DT requiring treatment. mRECIST 1.1  outperform standard \nRECIST1.1 in this clinical scenario. \nLimitations: Retrospective \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 114  \nFunding for this study: Nothing to disclose \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nLorenzo Saggiante: Nothing to disclose \nLucilla Violetta Sciacqua: Nothing to disclose \nElena Palassini: Nothing to disclose \nChiara Colombo: Nothing to disclose \nAlessandro Gronchi: Nothing to disclose  \nAndrea Vanzulli: Nothing to disclose \nTommaso Cascella: Nothing to disclose \nCarlo Morosi: Nothing to disclose \nSilvia Stacchiotti: Nothing to disclose \n \n \nTransarterial Embolization for Adhesive Capsulitis:  outcome assessment \nusing MRI \n*B. Wang*¹, K-W. Liang², H. Y. Lin²; ¹Tainan/TW, ²T aichung/TW \n(wangbow1227@gmail.com) \n \nPurpose or Learning Objective: To assess the efficacy of transarterial \nembolization (TAE) for adhesive capsulitis (AC) by evaluating clinical outcomes \nand changes in inflammatory status using magnetic r esonance imaging (MRI). \nMethods or Background: Patients diagnosed with AC, undergoing TAE, and \nwith baseline and 3-month contrast-enhanced MRI eva luations, were included. \nMRI results were analyzed to assess periarticular c apsule/ligament \ninflammation. Clinical assessments included pain sc ores using the Numeric \nRating Scale (NRS) and functional scores using the Quick Disabilities of the \nArm, Shoulder, and Hand (Quick DASH) questionnaire.  \nResults or Findings: Twenty-five patients with AC were included. Signifi cant \nreductions in average NRS pain scores, and signific ant improvements in Quick \nDASH scores and ROM, including anterior flexion and  abduction, were \nobserved at 1, 3, and 6 months after TAE (all P < 0 .001). MRI analyses \nrevealed that TAE significantly decreased the grade  of axillary recess (AR) \ncapsule enhancement, the grade of rotator interval (RI) capsule T2 signal \nintensity, and the grade of RI capsule enhancement (all P ≤ 0.004). \nConclusion: TAE is a promising and safe therapeutic approach fo r AC, \nimproving pain alleviation and functional recovery.  The observed MRI findings \nsuggest that the mechanism of TAE for AC may involv e the reduction of \ninflammation and the elimination of angiogenesis. \nLimitations: First, it was a single-arm study without a control group, meaning \nthat clinical outcomes such as pain and functional scores were subjective and \ncould be influenced by a placebo effect, necessitat ing cautious interpretation. \nSecond, the study involved a relatively small numbe r of patients. Third, the \nclassification of hypersignality and enhancement in tensity is arbitrary. Finally, \ndifferent MRI machines with varying magnetic field strengths (1.5T and 3T) \nwere used across patient examinations, introducing potential discrepancies in \nimage interpretation. \nFunding for this study: Nil \nEthics committee - additional information: Institutional Review Board of \nNational Cheng Kung University Hospital: IRB No: B- ER-113-038 \nAuthor Disclosures:  \nBow Wang: Nothing to disclose \nKeng-Wei Liang: Nothing to disclose \nHsuan Yin Lin: Nothing to disclose \n \n \nA novel treatment for persistent symptoms following  spinal surgery: \npercutaneous ct-guided trans-facetal screw fixation  \n*K. Desalos*, V. Sala, P-A. Ranc, L. J. Pavan, T. V ivarrat-Perrin, N. Amoretti; \nNice/FR \n(kevindesalos@gmail.com) \n \nPurpose or Learning Objective: Segmental spinal instability after \nlaminectomy and adjacent segment disease (ASD) foll owing arthrodesis often \nrequires repeated surgical interventions, usually b y complex surgical \nprocedures such as surgical arthrodesis under gener al anesthesia which can \nbe often demanding in patients with associated como rbidities. Trans-facetal \nfixation (TFF) under local anesthesia and CT guidance is a minimally invasive \ntechnique which involves placement of percutaneous screws through the \nposterior facet joints to improve spinal stability.  Our retrospective study is \naimed at evaluating the efficacy of pain reduction  and improvement of daily \nactivities following CT guided TFF, in patients wit h symptoms related to focal \ninstability secondary to laminectomy or ASD. \nMethods or Background: TFF were performed in 43 symptomatic patients \nwith previous history of spinal surgery (laminectom y and/or surgical \narthrodesis) at Nice University Hospital between 20 17 and 2024. The pre and \npostoperative pain and disability levels were colle cted prospectively, using the \nvisual analogue scale (VAS) and the Oswestry Disabi lity Index (ODI), at 6-\nmonths and 1-year. Long term outcomes were assessed  by telephone \nconsultations. \n \nResults or Findings: There was a mean decrease of VAS by 3.4 points at 6  \nmonths and by 3. 6 points at 1 year. Mean decrease of ODI was 17.1 points at \n6 months (47.7 +/- 13.4 versus 30.6 +/- 17.5, P=0.0 009)). All the screws were \nsatisfactorily positioned and the pain tolerance un der local anesthesia was very \ngood. The average duration of the procedure was 46 minutes, without any \nsignificant complications. \nConclusion: TFF under CT guidance and local anesthesia resulted  in a \nsignificant pain reduction and improvement in daily  activities, without any \ncomplications. TFF under CT guidance and local anes thesia is a safe and \neffective alternative for symptomatic instability after previous surgical \nlaminectomy and/or arthrodesis. \nLimitations: Small cohort \nMonocentric study \nFunding for this study: CHU Nice \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nLuca Jacopo Pavan: Nothing to disclose \nKevin Desalos: Nothing to disclose \nThomas Vivarrat-Perrin: Nothing to disclose \nNicolas Amoretti: Nothing to disclose \nPaul-Alexis Ranc: Nothing to disclose \nVincent Sala: Nothing to disclose \n \n \nGenicular Artery Embolization for the Treatment of Symptomatic Knee-\nOA using resorbable particles: a pilot study of 66 patients \n*F. N. N. Fleckenstein*, B. Gebauer, F. Collettini;  Berlin/DE \n(florian.fleckenstein@charite.de) \n \nPurpose or Learning Objective: Genicular artery embolization (GAE) is an \ninnovative, minimally invasive therapy for patients  with symptomatic knee \nosteoarthritis (OA) that is unresponsive to conserv ative treatments. Despite its \npotential, there is ongoing debate regarding the op timal embolic material for \nthis procedure. This study assesses both the safety  and efficacy of GAE using \na novel, resorbable particle in treating symptomati c knee OA. \nMethods or Background: A single-center study was conducted at our \ninstitution. Participants were aged between 40 and 90 years, all presenting with \nmoderate to severe knee OA (Kellgren-Lawrence grade s 2 to 4) and a history \nof failed conservative treatments. Baseline pain wa s measured using the visual \nanalog scale (VAS), and symptoms were assessed usin g the Knee Injury and \nOsteoarthritis Outcome Score (KOOS). After femoral arterial access was \nachieved with a 4 Fr sheath, embolization was perfo rmed using Nexsphere-F \nparticles (100-300 µm). Target vessels were identif ied through digital \nsubtraction angiography, correlating with patients'  pain locations. Adverse \nevents and symptom improvements were evaluated at 6  weeks, 3 months, and \n6 months following the procedure. \nResults or Findings: 66 patients were enrolled, with a median age of 68y . OA \nseverity was grade 2 (14%), 3 (44%), and 4 (42%). T echnical success was \n100%. Skin discoloration and mild knee pain, were r eported in 9% of cases. No \nmajor complications occurred. At the 6-month, the K OOS quality-of-life index \nshowed an 89% improvement, while VAS score indicate d an 82% reduction in \npain, median baseline 52/100 and 8/10, respectively . \nConclusion: This study demonstrates that using resorbable parti cles for GAE \nis both safe and effective in alleviating symptoms associated with OA that do \nnot respond to conservative treatments. \nLimitations: This is a single-center study without a control gro up, yet the size \nof the cohort is considerably large. \nFunding for this study: No \nEthics committee - additional information: Ethics approved \nAuthor Disclosures:  \nBernhard Gebauer: Nothing to disclose \nFederico Collettini: Nothing to disclose \nFlorian Nima Nima Fleckenstein: Nothing to disclose  \n \n \nResults of US-guided Hyaluronic Acid in patients wi th ankle osteoarthritis \nand osteochondral lesions of the talus \n*M. De Albert De Delas-Vigo*, E. A. Vargas Meouchi,  I. Benegas, A. Sallent, \nG. Duarri, Y. Lara Taranchenko, S. Roche, M. Veinte millas, J. Alonso; \nBarcelona/ES \n(tisodealbert@gmail.com) \n \nPurpose or Learning Objective: Describe the effects regarding pain relief and \npossible complications of hyaluronic acid (HA) inje ction in patients with ankle \nOA and OLT. \nMethods or Background: Observational study of patients with ankle OA and \nOLT that had an ultrasound guided HA injection by t he radiology department \nbetween January 1st 2020 and December 1st 2023 in o ur center and a \nminimum follow-up of 6 months. Patients' visual ana log scale (VAS) at \nbaseline, 3 months and 6 months after injection wer e recorded. Exclusion \ncriteria were administration of another medication (except mepivacaine) or \nbiological therapy and incomplete data collection d uring follow-up. \n\n \n \nThursday \nAbstract-based Programme \n \n 115  \nResults or Findings: 137 patients were referred to the radiology departm ent \nfor an ultrasound guided injection. Fifty-eight pat ients (63 ankles) that received \nHA injection for ankle OA and OLT were identified. Thirty-seven (42 ankles) \nwere included. Baseline VAS score was 7.98 ± 1.37; 5.76 ± 2.14, and 6.64 ± \n2.07 at 3 and 6 months post injection respectively (p<0.05). Patients reported a \nmean of 7.9 ± 7.81 months with some pain relief. Patients mean age was 59.6 \n(range 32-83) and mean follow-up was 18.8 ± 12.1 mo nths. Eleven (26.2%) \npatients received a second HA injection and 7 (16.6 %) underwent surgery \nduring follow-up. No complications were recorded in  this series. \nConclusion: HA injections is a safe treatment that may provide a temporary \npain improvement in patients with ankle OA and OLT.  \nLimitations: Retrospective studio \nFunding for this study: None \nEthics committee - additional information: Prospective clinical study \nAuthor Disclosures:  \nYuri Lara Taranchenko: Nothing to disclose \nSarai Roche: Nothing to disclose \nMatias De Albert De Delas-Vigo: Nothing to disclose  \nGemma Duarri: Nothing to disclose \nMaite Veintemillas: Nothing to disclose \nJaime Alonso: Nothing to disclose \nEnrique A. Vargas Meouchi: Nothing to disclose  \nIker Benegas: Nothing to disclose \nAndrea Sallent: Nothing to disclose \n \n \nPercutaneous Hydrodissection Technique For Anterior  Cervical \nApproach of the spine \nI. Ben Rejeb¹, J. Lavigne¹, S. Grange², J-B. Noel³,  *J-B. Pialat*¹, N. Stacoffe¹; \n¹Pierre-Bénite/FR, ²Saint-Étienne/FR, ³Lyon/FR \n(jean-baptiste.pialat@chu-lyon.fr) \n \nPurpose or Learning Objective: Hydrodissection aims to reduce the risks \nassociated with traditional open surgery by enablin g minimally invasive access \nto the cervical spine. To evaluate the feasibility,  effectiveness, and safety of \nusing hydrodissection in percutaneous anterior cerv ical spine procedures. \nMethods or Background: A retrospective analysis was conducted on 32 \nhydrodissection procedures (30 patients) performed between 2020 and 2024 in \nthree medical centers. The patient cohort included individuals with advanced \noncological conditions, benign tumors, trauma, and infections. The \nhydrodissection technique involved gradual injectio n of a mixture of normal \nsaline and contrast medium under CT guidance, displ acing the jugulo-carotid \nstructures and creating a safe space for interventi onal procedures such as \nodontoid osteosynthesis, cementoplasty, biopsy, and  thermoablation. \nResults or Findings: The procedure was successfully completed in all 30 \npatients. The right-sided approach was predominantl y used to avoid \nesophageal injury. The mean volume of dissection fl uid used was around 250 \nml, with continuous hydrodissection being essential  to maintain the created \nspace. A few patient with multiple level procedure reach to 500 ml. The \nextubation following the procedure was shortly dela yed to avoid any \ncompression risk. No perioperative or postoperative  complications were \nreported. \nConclusion: The use of hydrodissection in anterior cervical spi ne procedures \noffers a safe and effective alternative to open sur gery, particularly for high-risk \npatients. This technique minimizes neurovascular co mplications and ensures \nsafe procedural access, with no recorded complicati ons in this short serie. \nLimitations: Further studies with larger cohorts are warranted t o standardize \nthe technique and confirm its safety and efficacy. \nFunding for this study: None \nEthics committee - additional information: Actually evaluated by a comittee \nto be approved for multiple center retrospective in clusion \nAuthor Disclosures:  \nSylvain Grange: Nothing to disclose \nNicolas Stacoffe: Nothing to disclose \nIlyess Ben Rejeb: Nothing to disclose \nJoris Lavigne: Nothing to disclose \nJean-Baptiste Pialat: Nothing to disclose \nJean-Baptiste Noel: Nothing to disclose \n \n \nPatient reported outcomes and return to work after CT-guided \npercutaneous lumbar discectomy: a prospective study  \n*P-A. Ranc*, N. Amoretti; Nice/FR \n(passi_ranc@msn.com) \n \nPurpose or Learning Objective: To evaluate the efficacy percutaneous \nlumbar discectomy (PLD) under computed tomography ( CT) guidance on pain, \nfunctional capacities and to estimate the speed of recovery by assessing the \ntime before return to work. \n \n \nMethods or Background: Patients treated by PLD were prospectively \nincluded between December 2019 and April 2021. Data  collected consists in \npain, duration of the symptoms, analgesia intakes, time of absence from work \nand the Oswestry Disability Index (ODI). Patients w ere followed-up during six \nmonths. Duration of hospitalization and time before  return to work were \nreported. The Fisher test was used to compare nomin al variables, the Kruskal-\nWallis test for ordinal variables, and the Student test to compare quantitative \ncontinuous variables. \nResults or Findings: A total of 87 patients were evaluated (median age, 56; \ninterquartile range [IQR], 42.5-66). The median ODI  decreased from 44 (IQR, \n33-53) to 7 (IQR, 2-16.5) at six months (p<0,001). The median visual analog \nscale (VAS) decreased from 8 (IQR, 8-9) to 2 (IQR, 0-3) within six months \n(p<0,001). In 96,5% of cases, patients were dischar ged on the day of the \nprocedure, and 3,5% the following day. No severe ad verse events were \nreported according to the society of interventional  radiology (SIR) classification \nsystem. Out of the 57 patients at work, 50 were abl e to return to work during \nthe follow-up with a median time of 8 days (IQR, 0- 20). \nConclusion: Symptomatic lumbar disc herniations can be successf ully treated \nby PLD, resulting in significant improvement of sym ptoms, functional \ncapacities, and a fast return to work. \nLimitations: The limitations of the study are the lack of a cont rol group, so no \nreal comparison among patients treated with surgica l methods. Full follow-up \nwas not obtained for 20 patients who had to be excl uded from the analysis. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study received approval \nfrom the institutional review board and reference o n the clinicaltrials.gov \ndatabase. \nAuthor Disclosures:  \nNicolas Amoretti: Author: Last Author \nPaul-Alexis Ranc: Author: First Author \n \n \nImmediate imaging findings after positioning of a n ew percutaneous \ninterspinous process spacer \n*L. J. Pavan*, P-A. Ranc, T. Vivarrat-Perrin, K. De salos, V. Kilani, E. Dien,  \nN. Amoretti; Nice/FR \n(lucajpavan@gmail.com) \n \nPurpose or Learning Objective: To evaluate the immediate changing in spine \nimaging after positioning of a percutaneous intersp inous process spacer (IPS) \nfor symptomatic degenerative lumbar spinal stenosis  (DLSS). \nMethods or Background: All patients treated in our Centre from January 202 1 \nto December 2023 with a new percutaneous IPS (Lobst er®-Diametros \nMedical®) were retrospectively reviewed. Patients u nderwent this procedure \nfor treating a symptomatic DLSS nonresponding to lu mbar epidural injection. \nAll procedures were performed with combined CT-scan  and fluoroscopy \nguidance under general anesthesia. For each patient  neural foramina area on \nsagittal plane, as well as zygapophyseal articular space on axial plane, were \nindependently measured by two operators on the trea ted level of preoperative \nand postoperative CT-scans. \nResults or Findings: Thirty-four consecutive patients were treated in th e \nconsidered period, and all were retrospectively inc luded in the study (mean \nage 79.2 ± 8.3 years [72-92], 27 males, 17 females). Mean neural foramina \narea increased from 73 ± 19 to 93 ± 24 mm2 (+ 27%; p < 0.01). Mean facet \njoint articular space increased from 2,2 ± 0,9 to 3,1 ± 1,1 mm (+ 40%; p < \n0.01). Inter-observer reliability was very good (Cr onbach's alpha = 0.9). No \nprocedural complication was reported. \nConclusion: These imaging changes may explain the clinical effe cts of IPS. \nThe widening of foramina may be related to a decrea se pression on the middle \ncolumn and the posterior portion of the interverteb ral disk, with reduction of its \nprotrusion. The widening of the zygapophyseal space  is linked to a stretching \nof flavum ligaments with a consequent reduction of its bulging into the spinal \ncanal. Both factors will play a major role in reduc ing clinical symptoms of \nDLSS. \nLimitations: The retrospective design of the study. \nFunding for this study: None \nEthics committee - additional information: Not neede since observational \nand retrospective \nAuthor Disclosures:  \nVictor Kilani: Nothing to disclose \nLuca Jacopo Pavan: Nothing to disclose \nKevin Desalos: Nothing to disclose \nEmmanuel Dien: Nothing to disclose \nThomas Vivarrat-Perrin: Nothing to disclose \nNicolas Amoretti: Nothing to disclose \nPaul-Alexis Ranc: Nothing to disclose \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 116  \nEvaluation of the therapeutic value of conventional  lymphography for the \ntreatment of inguinal lymphatic fistulas after lymp hadenectomy \n*C. Wolfram*, T. J. Vogl, K. Eichler, T. Gruber-Rou h; Frankfurt/DE \n(Christian.Wolfram@unimedizin-ffm.de) \n \nPurpose or Learning Objective: There is critics that advancements in \ntechnology, particularly in cross-sectional imaging  techniques such as \ncomputed tomography (CT) and magnetic resonance ima ging (MRI) are better \nthan conventional lymphography. The aim of the pres ent study is to prove that \na lipiodol-based conservative lymphography seals pe rsistent lymphatic fistulas, \nproviding a safe and effective alternative to conse rvative therapies and surgical \ninterventions. \nMethods or Background: A group of 39 patients who underwent \nlymphadenectomy resulting into inguinal lymphatic f istulas between 2003 and \n2023 was selected. Participants were eligible if th ey had persistent lymphatic \nleakage after inguinal lymphadenectomy and were unr esponsive to \nconservative treatment. Of these 39 patients, four could not be statistically \nevaluated in our retrospective study due to various  technical problems. The \nremaining 35 patients were evaluated. Lipiodol lymp hography was performed \nusing transpedal lymphatic vessel cannulation. Data  were collected through \nclinical records (RIS/PACS) and imaging follow-ups.  Statistical analyses \nincluded the Wilcoxon–Mann–Whitney test using BiAS software. Success was \ndefined as the complete occlusion of lymphatic leak age, and data on \ncomplications and additional interventions were col lected. \nResults or Findings: Therapeutic success was achieved in 22 patients \n(62.86%), with a mean resolution time of 7.13 days.  For four patients there was \nno data on further course, while 13 required additi onal interventions (three \nsurgical, six radiotherapy). Statistical analysis s howed no significant correlation \nbetween the volume of administered iodized oil and therapeutic success (p = \n0.51), nor lymphatic drainage volume (p = 0.69). \nConclusion: Our results highlight that conventional lymphograph y is a \nsuccessful therapy. These findings could inform fut ure studies aiming to refine \npatient selection criteria and optimize treatment p rotocols for complex \nlymphatic conditions. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number UID 2023-1444. \nAuthor Disclosures:  \nChristian Wolfram: Nothing to disclose \nKathrin Eichler: Nothing to disclose \nThomas J. Vogl: Nothing to disclose \nTatjana Gruber-Rouh: Nothing to disclose \n \n \n14:00-15:30 Research Stage 4 \nResearch Presentation Session: Breast \nRPS 1002 \nExploring the role of artificial intelligence \nin breast imaging \n \nModerator \nI. Thomassin-Naggara; Paris/FR  \n(isabelle.thomassin@aphp.fr) \n \n \nA 10-year image-derived AI risk model for use in pr imary prevention of \nbreast cancer \n*M. Eriksson*¹, K. Czene¹, C. Scott², P. Hall¹, C. Vachon²; ¹Stockholm/SE, \n²Rochester, MN/US \n(mikael.eriksson@ki.se) \n \nPurpose or Learning Objective: Image-derived artificial intelligence (AI) risk \nmodels have shown significant potential in enhancin g breast cancer (BC) \nscreening through short-term risk assessment. A lon g-term image-derived AI \nrisk model for primary prevention has yet to be dev eloped and externally \nvalidated. \nMethods or Background: This study utilized a case-cohort approach, \nincluding women aged 35-94 recruited between 2009-2 017 from population-\nbased screenings in Olmsted County, Minnesota (U.S. ), and the KARMA \ncohort in Sweden. Median follow-up was 10 years, wi th BCs diagnosed before \n5/2022. An image-derived AI risk model, initially d eveloped in a Swedish \npopulation, was validated independently in the Olms ted/KARMA cohorts. At \nstudy entry, 10-year absolute risks were estimated.  Time-dependent \ndiscriminatory performance (AUC(t)) and expected-to -observed event ratios \n(E/O) were calculated. \nResults or Findings: The combined Olmsted/KARMA cohorts included 8,721 \nwomen, with a mean age of 54.4 years (±10.6) in the  subcohort and 1,633 \nincident BC cases with a mean age of 57.0 years (±10.6). The AI-derived 10-\nyear average risks were 3.83% and 3.14%, with E/O r atios of 0.99 (95%CI \n0.94-1.05) in Olmsted and 0.99 (95%CI 0.91-1.08) in  KARMA. The 10-year \nAUC(t) values were 0.70 (95%CI 0.68-0.73) for Olmst ed and 0.73 (95%CI \n0.69-0.77) for KARMA. Using the U.S. Preventive Ser vices Task Force \n(USPSTF) guidelines, 41% of cases in KARMA were ide ntified as high-risk, \ncompared to 15% with Tyrer-Cuzick-v8 and 5.1% with BCSC-v3 (p<0.01). \nUnder the National Institute for Health and Care Ex cellence (NICE) guidelines, \nthese figures were 31%, 7.4%, and 0.2%, respectivel y. \nConclusion: The 10-year image-derived AI risk model demonstrate d strong \npredictive performance in both U.S. and Swedish cas e-cohorts, outperforming \ntraditional clinical risk models in KARMA. This AI model holds significant \npotential for clinical application in primary preve ntion, targeting up to 40% of \nBCs. \nLimitations: The study population was mainly White women. \nFunding for this study: Swedish Research Council \nEthics committee - additional information: Mayo Clinic and Olmsted \nMedical Center Institutional review board and the S wedish Ethical Review \nAuthority \nAuthor Disclosures:  \nMikael Eriksson: Patent Holder: Patent on \"system a nd method for assessing \nbreast cancer risk using imagery\" with a licence to  iCAD medical, Nashua, NH, \nU.S. Patent on \"compositions and methods for monito ring the treatment of \nbreast disorders\" with a licence to Atossa Therapeu tics, Seattle, WA, U.S. \nCeline Vachon: Nothing to disclose \nChristopher Scott: Nothing to disclose \nKamila Czene: Patent Holder: Patent on \"system and method for assessing \nbreast cancer risk using imagery\" with a licence to  iCAD medical, Nashua, NH, \nU.S. Patent on \"compositions and methods for monito ring the treatment of \nbreast disorders\" with a licence to Atossa Therapeu tics, Seattle, WA, U.S. \nPer Hall: Patent Holder: Patent on \"system and meth od for assessing breast \ncancer risk using imagery\" with a licence to iCAD m edical, Nashua, NH, U.S. \nPatent on \"compositions and methods for monitoring the treatment of breast \ndisorders\" with a licence to Atossa Therapeutics, S eattle, WA, U.S. \n \n \nCracking the Code: Predicting Pathogenic Mutations in Breast Cancer \nwith Ultrasound Radiomics \n*R. M. Pintican*, N. Antone; Cluj-Napoca/RO \n(roxana.pintican@gmail.com) \n \nPurpose or Learning Objective: To evaluate the potential of US-based in the \nprediction of pathogenic mutational status of breas t cancer patients, relevant to \nprophylactic mastectomy recommendations. \nMethods or Background: This retrospective study included 73 breast cancer \npatients tested with multigene panel tests includin g all seven pathogenic \nmutations (BRCA1, BRCA2, TP53, PTEN, CDH1, PALB2, a nd STK11 \nmutations). US images were acquired prior to any tr eatment and tumoral and \nperitumoral areas were used to extract radiomics da ta. The study population \nwas divided into testing and validation group, each  with pathogenic- and non-\npathogenic mutation population. Radiomics features were analyzed using \nmachine learning models, alone and in combination w ith clinical features ( \nki67%). \nResults or Findings: We observed significant differences in radiomics \nfeatures between pathogenic- and non-pathogenic mut ation driven tumors. \nUsing a three-step feature selection process we dev elop the prediction models \n(The Mann-Whitney U test, Spearman Correlation and LASSO Regression); \nthe Rad-score 1 ( tumor) achieved an accuracy of 78 .6% in identifying \npathogenic mutation carriers, while Rad-score 2 (tu mor+peritumoral) increased \nthe model's accuracy to 85%. The Rad-Clin 1 and Rad -Clin 2 achieved 83% \nand 95% acuracy in predicting mutational status. On  validation cohort we \nobtained the following AUCs: Rad-score 1 = 66%; Rad -score 2 = 91%; Rad-\nClin 1 = 58%; Rad-clin 2 = 83%. \nConclusion: Radiomics models based on US images of breast tumor s may \nprovide a promising alternative for predicting path ogenic mutation status in BC \npatients. The highest accuracy was reached when we combined radiomics \ndata extracted from the tumor and peritumoral area.  This approach could \nreduce dependence on costly genetic testing and exp edite the diagnostic \nprocess. \nLimitations: Small sample size \nUnicentric study \nFunding for this study: No funding \nEthics committee - additional information: Retrospective study - the \ninformed consent was waived. \nAuthor Disclosures:  \nRoxana Maria Pintican: Nothing to disclose \nNicoleta Antone: Nothing to disclose \n \n\n \n \nThursday \nAbstract-based Programme \n \n 117  \nDo we still need to double read the most suspicious  screening \nmammograms when using AI for decision support? A su b-analysis from \nthe AITIC breast cancer screening prospective trial  \n*E. Elías Cabot*¹, A. Rodriguez Ruiz², J. L. Raya P ovedano¹,  \nS. Romero Martin¹, M. Álvarez Benito¹; ¹Cordoba/ES,  ²Nijmegen/NL \n(eeliascabot@gmail.com) \n \nPurpose or Learning Objective: To evaluate the differences between single \nand double reading of the most suspicious mammogram s after the introduction \nof AI in breast cancer screening. \nMethods or Background: This was a sub-analysis of the AITIC paired \nprospective trial in the breast cancer screening pr ogram in Cordoba, Spain. In \nthis trial, between March 2022 and January 2024, 31 ,301 women (age 50-71) \nwere included and imaged with either DM or DBT base d on equipment \navailability. Two reading strategies were independe ntly applied to each exam: \nDouble blind and non-consensual reading of all exam s (standard strategy) and \nAI-based triaging (AI strategy), where an AI system  (Transpara v1.7, \nScreenPoint Medical) evaluated the cancer risk of a ll exams. Cases identified \nby AI as Low risk were automatically assessed as ne gative, while exams with \nIntermediate or Elevated risk were double read with  concurrent AI-support. For \nthe latter group, cancer detection (CDR) and false positive rates (FPR) were \ncompared between single and double reading. P value s using McNemar and \nbinomial confidence intervals (CI) were computed. \nResults or Findings: The AI strategy, double reading only 36% of the tot al \nscreening mammograms, resulted in 228 screen-detect ed cancers \n(CDR=7.3/1000, CI: 6.4-8.2/1000) and 1,723 recalls (FPR=4.8%, CI: 4.5-\n5.0%). Should these exams have been single read wit h AI support, there would \nhave been 190 screen-detected cancers (CDR = 6.0/10 00, CI: 5.2-7.0/1000), \nand 1,082 recalls (FPR = 2.9%, CI: 2.7-3.0%), a -17 % (P<0.05) and -42% \nreduction (P<0.05) with respect to double reading. The standard strategy \nresulted in 1,501 recalls (FPR=4.2%, CI: 4.0-4.4%) and 198 cancers \n(CDR=6.3/1000, CI: 5.5-7.2/1000). \nConclusion: After introduction of AI for triage and decision su pport in \nscreening, increased cancer detection rates were ac hieved in comparison to \nstandard of care by still double reading a subgroup  of the most suspicious \nexams. \nLimitations: Single-site. \nFunding for this study: None. \nEthics committee - additional information: Local IRB approval. \nAuthor Disclosures:  \nAlejandro Rodriguez Ruiz: Employee: ScreenPoint Med ical \nMarina Álvarez Benito: Nothing to disclose \nSara Romero Martin: Nothing to disclose \nEsperanza Elías Cabot: Nothing to disclose \nJose Luis Raya Povedano: Nothing to disclose \n \n \nBenefits and risks of AI use for reviewing negative  screening \nmammograms \n*C. De Wolf*¹, K. Brändle², J-L. Bulliard²; ¹Geneva /CH, ²Lausanne/CH \n(c.dewolf@adsan.org) \n \nPurpose or Learning Objective: Introduction: Breast cancer remains a global \nhealth concern, with artificial intelligence (AI) offering promising advancements \nin improving screening accuracy. Traditional method s, requiring high-volume \nreadings, often lead to fatigue and reading errors.  AI addresses these \nlimitations by providing fatigue-free, reproducible  results. This study assesses \nthe benefits and costs of AI in detecting high-risk  lesions in mammograms \ninitially classified as negative by radiologists. \nMethods or Background: Methods: Risk scores (Transpara® version 1.7.3) \nwere calculated for 54’300 mammograms from a public  Swiss screening \nprogram (2018–2021). Data included screen detected (n=321) and interval \ncancers (n=94), lesion location, and double-blind r adiologist readings. We \nincluded risk score thresholds considered as elevat ed risk (61 to 90). Key \noutcomes included additional workload (additional m ammograms in consensus \nconference), avoided false-negative interval cancer s (FN-IC, n=39), and \nincreased false-positive (FP) rates. Multivariable logistic regression was used \nto predict the rise in FP cases across thresholds. \nResults or Findings: Results: The FN-IC rate reduction ranged from 8.3% \n(threshold 90) to 31.3% (threshold 61), with an add itional workload of 2 to 67 \nextra mammograms per 1,000 participants. Avoiding o ne FN-IC case required \n28 to 242 extra readings, resulting in 12 to 86 add itional false positives (FP). \nFP rates rose by 2.1% to 59.3%, with the workload i ncreasing by a third for \nevery 5-point threshold drop up to 75. With an AI t hreshold set to 85, the false \npositive rate increased by 2.3‰ (from 40.5‰ to 42.8 ‰) and the workload \nwould increase by 6 mammograms /1000 participations . \nConclusion: AI assistance may enhance mammography sensitivity. However, \nthis comes with a relatively high cost in terms of FP results and additional \nreadings. Therefore, determination of the critical threshold must be context-\nspecific to achieve optimal benefit – risk ratio. \nLimitations: Retrospective design. \nFunding for this study: No external funding \nEthics committee - additional information: All women signed an informed \nconsent that their anonymized screening data could be used for quality \nassurance purposes. \nAuthor Disclosures:  \nJean-Luc Bulliard: Nothing to disclose \nKaren Brändle: Nothing to disclose \nChristophorus De Wolf: Nothing to disclose \n \n \nPatient perceptions and attitudes towards the use o f AI in the \nsymptomatic breast unit \n*S. Singh*, R. P. Crean, H. Briody, R. Bruen, N. Ha mbly, M. Bambrick,  \nD. Duke, M. Mullooly, N. Healy; Dublin/IE \n(snehasingh2412@gmail.com) \n \nPurpose or Learning Objective: Artificial intelligence (AI) has been evaluated \nin a number of breast screening settings with favou rable results. While there \nare limited studies looking at patient attitudes to  AI in breast screening none \nhave examined perceptions of AI in the symptomatic setting. The aim of this \nstudy was to determine attitudes towards AI among p atients attending the \nsymptomatic breast unit. \nMethods or Background: An anonymous 15 question, voluntary \nquestionnaire was given to all patients attending t he symptomatic breast clinic \nimaging department of Beaumont Hospital from 01/07/ 2024 to 30/09/2024. \nResults were collated in a password protected Excel  database and descriptive \nstatistics performed. Likert responses were numeric ised so that mean of 1 \ndenotes strong agreement and 5 denotes strong disag reement. \nResults or Findings: Of the 1500 patients who were surveyed, most were \naged 40–59 years (62.1%). Almost one-quarter had ei ther a personal \n(364/1500) or family history of breast cancer (360/ 1500). 62% (927/1500) had \nsome or strong interest in AI. Regarding the use of  AI in healthcare, 46% \nagreed it was a good idea, 8% disagreed and 46% wer e indifferent. There was \nsupport for AI assisting radiologists in reading ma mmograms (Mean \n(M)=2.43,95% CI:2.39-2.48) but disapproval of AI be ing the sole reader \n(M=3.82,95% CI:3.77-3.87). Respondents strongly pre ferred human \nradiologists over AI for reading mammograms, even i f AI were more efficient \n(M=1.95,95% CI:1.90-1.99) or more accurate (M=2.17,  95% CI:2.13-2.22). \n75% of patients would blame both the AI developer a nd the human radiologist \nfor an incorrect result. All results were statistic ally significant (p<0.001). \nConclusion: Respondents hold favourable views towards the use o f AI in \nhealthcare. They welcome use of AI as an adjunct fo r radiologists but disagree \nwith AI being the only reader of their mammogram. \nLimitations: N/A \nFunding for this study: RCSI seed funding \nEthics committee - additional information: Approval has been obtained from \nthe hospital audit committee (CA2024/126). Formal e thical approval was not \ndeemed necessary as this is an anonymised, voluntar y study. \nAuthor Disclosures:  \nNiamh Hambly: Nothing to disclose \nMaeve Mullooly: Nothing to disclose \nDeirdre Duke: Nothing to disclose \nMarie Bambrick: Nothing to disclose \nRichard Bruen: Nothing to disclose \nHayley Briody: Nothing to disclose \nRory Peter Crean: Nothing to disclose \nNuala Healy: Nothing to disclose \nSneha Singh: Nothing to disclose \n \n \nThe effect of an artificial intelligence decision s upport system on \nradiologists’ screening mammography performance and  visual search \npatterns \n*J. Gommers*, S. D. Verboom, M. Broeders, I. Sechop oulos; Nijmegen/NL \n(jessie.gommers@radboudumc.nl) \n \nPurpose or Learning Objective: To investigate the effect of using a \ncommercial artificial intelligence (AI) decision su pport system on the diagnostic \nperformance and visual search patterns of radiologi sts interpreting screening \nmammograms. \nMethods or Background: A multi-reader, multi-case study was performed \nwith 12 Dutch screening radiologists interpreting 1 50 screening mammography \nexaminations (75 normal, 75 malignant). Radiologist s read the examinations \nwithout and with AI support while an eye tracker re corded their eye \nmovements. AI classified the examinations as low (m aximum region \nscores:<40), intermediate (40-59), medium-high (60- 79), or very-high risk (80-\n100). Radiologists provided a probability of malign ancy score (0-100) and \nrecall decision for each examination. The performan ce under the two reading \nconditions was compared using the area under the re ceiver operating \ncharacteristics curve (AUC), sensitivity, and speci ficity through mixed-model \nanalysis of variance. Reading time and eye tracking  outcomes were compared \nby bootstrap resampling (n=20,000). \n\n \n \nThursday \nAbstract-based Programme \n \n 118  \nResults or Findings: The average AUC increased significantly from 0.93 \nwithout AI support to 0.97 with AI support (P<.001) . There was no evidence of \na significant change in sensitivity (81.7% vs 87.2% , P=.06) or specificity \n(89.0% vs 91.1%, P=.46), although sensitivity tende d to increase for AI-\nclassified high-risk examinations (medium-high: 54. 9% vs 61.8%, very-high: \n89.5% vs 95.6%). Overall reading time did not chang e significantly (29.4 vs \n30.8 seconds, P=.32), but decreased for AI-classifi ed low-risk examinations \n(25.1 vs 20.1 seconds, P<.001). When using AI, radi ologists covered less of \nthe breast area with fixations (11.1% vs 9.5%, P=.0 05), while spending more \ntime fixating in lesion areas (4.0 vs 5.1 seconds, P<.001). \nConclusion: Reading with an AI decision support system increase d \nradiologists’ screening performance and allowed the m to focus more on lesion-\nspecific areas without increasing overall reading t ime, indicating a more \nefficient search. \nLimitations: Enriched case set and one AI system only. \nFunding for this study: aiREAD financed by KWF Dutch Cancer Society and \nthe Dutch Research Council (NWO) Domain Applied and  Engineering \nSciences (AES), as part of their joint strategic re search program Technology \nfor Oncology II. The collaboration project is co-fu nded by the PPP Allowance \nmade available by Health-Holland, Top Sector Life S ciences & Health, to \nstimulate public-private partnerships. \nEthics committee - additional information: The need for ethical approval for \nthis retrospective multi-reader multi-case study wa s waived by the Research \nEthics Committee of Radboud University Medical Cent er (registration number \n2021–13186). \nAuthor Disclosures:  \nMireille Broeders: Research/Grant Support: Screenpo int Medical, Sectra \nBenelux, Hologic, Volpara Solutions, Lunit, iCAD Sp eaker: Siemens \nHealthcare, Hologic \nSarah Delaja Verboom: Nothing to disclose \nJessie Gommers: Nothing to disclose \nIoannis Sechopoulos: Research/Grant Support: Siemen s Healthcare, Canon \nMedical, ScreenPoint Medical, Sectra Benelux, Volpa ra Solutions, Lunit, iCAD \nSpeaker: Siemens Healthcare, Canon Medical \n \n \nMammographic features of false positive AI markings  on screening \nmammograms from BreastScreen Norway \n*M. A. Martiniussen*¹, M. B. Bergan², J. Gjesvik², M. Undrum Kristiansen¹,  \nS. Hofvind²; ¹Graalum/NO, ²Oslo/NO \n(maritkarl@gmail.com) \n \nPurpose or Learning Objective: False positive AI markings are an expected \nchallenge when implementing artificial intelligence  (AI) in mammographic \nscreening and might contribute to an unsustainable increase in the workload \nfor the radiologists. The aim of this study was to gain knowledge about false \npositive AI markings from two AI systems on screeni ng mammograms. \nMethods or Background: In this retrospective study, 129 385 screening \nexaminations from BreastScreen Norway, performed at  Ostfold Hospital Trust, \n2008-2018, were run through two AI systems. System A was Lunit INSIGHT \nMMG version 1.1.7.2, and system B was a non-commerc ial system, developed \nby the Norwegian Computing Center and the Cancer Re gistry of Norway. Each \nmodel provided a score on a scale from 0-100, and m arked the most \nsuspicious areas. Higher score indicated higher ris k of cancer. Two radiologists \nperformed a consensus-based informed review of exam inations among those \nwith the 5% highest AI score from both systems, int erpreted negative at index \nscreening and without cancer diagnosed at index and  two consecutive \nscreening rounds. Mammographic features correspondi ng to the AI markings \nwere classified according to the Breast Imaging Rep orting and Data System \n(BI-RADS). The results were analyzed using descript ive statistics. \nResults or Findings: Among the examinations that met the inclusion crite ria \n(n=252), 120 examinations from 120 women were rando mly selected for \nreview. The mammographic feature corresponding to t he AI markings was \ncalcifications for 71.7% (86/120) for system A and 67.5% (81/120) for system \nB, a mass for 12.5% (15/120) for system A and 14.2%  (17/120) for system B, \nwhile asymmetry accounted for 10.8% (13/120) for sy stem A and 11.7% \n(14/120) for system B. \nConclusion: Calcifications was the main mammographic feature in  screening \nmammograms with high AI score without diagnosed can cer. \nLimitations: No limitations were identified. \nFunding for this study: The South-Eastern Norway Regional Health Authority \nOstfold Hospital Trust \nEthics committee - additional information: The study was approved by the \nRegional Committees for Medical and Health Research  Ethics (#13294, \n#11022) \nAuthor Disclosures:  \nJonas Gjesvik: Nothing to disclose \nMarie Burns Bergan: Nothing to disclose \nMarit Almenning Martiniussen: Nothing to disclose \nSolveig Hofvind: Nothing to disclose \nMerete Undrum Kristiansen: Nothing to disclose \n \nRe-attendance in BreastScreen Norway after a false positive screening \nresult \n*M. Larsen*, N. Moshina, J. Gjesvik, S. Sagstad, Å.  S. Holen, M. B. Bergan,  \nT. E. Nilsen, S. Hofvind; Oslo/NO \n(maln@kreftregisteret.no) \n \nPurpose or Learning Objective: Higher risk of breast cancer after a false \npositive versus a negative screening result has bee n reported. We aimed to \ncompare re-attendance for women with a false positi ve versus negative \nscreening result using more than 25 years of screen ing data. \nMethods or Background: BreastScreen Norway invites women aged 50-69 to \nbiennial screening. The study sample included 3 990  388 screening \nexaminations from 921 309 women where an invitation  to the subsequent \nscreening round was available (eligible for re-atte ndance). Attendance in the \nsubsequent screening round was analysed using mixed  logistic regression with \nage at screening and screening history as covariate s and screening outcome \nas exposure. Predicted probabilities (re-attendance ) and 95% confidence \nintervals (CI) were calculated using average margin al effects. \nResults or Findings: Having a false positive result after the prevalent \nscreening examination resulted in a re-attendance r ate of 88.3%. For women \nwith a negative result, re-attendance was 90.3% aft er the prevalent \nexamination. Having a false positive or negative re sult in the 9th screening \nround, gave a re-attendance rate of 89.0% and 91.1% , respectively. Predicted \nre-attendance rate was 88.9% (95% CI: 88.9-89.0) af ter a false positive result \nand 88.1% (95% CI 88.0-88.3%) after a negative resu lt. Using negative result, \nfalse positive without invasive procedure or false positive with invasive \nprocedure as exposure variable, the predicted proba bilities of re-attendance \nwere 88.9% (95% CI: 88.9-89.0), 88.4% (95% CI: 88.2 %-88.6%) and 87.6% \n(95% CI: 87.3%-88.0%), respectively. \nConclusion: Despite small differences in re-attendance after a false positive \nversus negative screening result, we consider the d ifference clinically \nimportant. Women should be informed about the impor tance of re-attending the \nscreening programme after a false positive result. \nLimitations: We do not have patient reported data on reasons for  non-\nattendance. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Programme quality assurance is \ncovered by the Cancer Registry Regulations. \nAuthor Disclosures:  \nNataliia Moshina: Nothing to disclose \nJonas Gjesvik: Nothing to disclose \nMarie Burns Bergan: Nothing to disclose \nÅsne Sørlien Holen: Nothing to disclose \nSilje Sagstad: Nothing to disclose \nTom Erik Nilsen: Nothing to disclose \nSolveig Hofvind: Nothing to disclose \nMarthe Larsen: Nothing to disclose \n \n \nThe application of artificial intelligence to enhan ce the identification of \npreviously missed non-palpable breast carcinomas \nS. A. Mansour, R. M. Kamal, S. Hussien, M. Emara, Y . Kassab, S. Taha,  \nM. M. Gomaa, *Y. M. Nada*; Cairo/EG \n(dr.yasmin.nada.nl@gmail.com) \n \nPurpose or Learning Objective: To investigate the impact of artificial \nintelligence (AI) on digital mammograms in increasi ng the chance of detection \nof missed breast cancer, study the early morphology  indictors detected by AI \nand overlooked by the radiologist and correlate wit h the missed cancer \npathological types. \nMethods or Background: Screening and diagnostic mammograms (done in \n2020-2023) presenting breast carcinomas (n = 1998) were analyzed in \nconcordance with prior one-year-ago (2019-2022) ass umed negative or \nbenign) mammograms. Present mammograms were reviewe d for the \nmammographic descriptors: asymmetry, distortion, ma ss, and \nmicrocalcifications. The AI analyzed mammograms and  presented \nabnormalities by overlaying color hue and scoring p ercentage for the degree of \nsuspicion of malignancy. \nResults or Findings: Artificial intelligence detected 555 (54%) lesions in the \nprior mammograms, and in present mammograms (year 2 020-2023) targeted \n904 (88%) carcinomas. The descriptor proportion of asymmetry was the \ncommon presentation of missed breast carcinoma (n=3 56/555, 64.1%) in the \nprior mammograms and the AI highest detection rate presented by distortion \n(100%) followed by grouped microcalcifiactions (80% ). AI performance to \npredict malignancy in previously assigned negative or benign mammograms \nshowed a sensitivity of 73.4%, a specificity of 89% , and an accuracy of 78.4%. \nConclusion: Reading mammograms with artificial intelligence enh anced the \ndetection of early cancerous changes. AI detection rate is not correlated with \ncertain pathological types of breast cancer. Close follow-up is required for AI \nabnormality scoring of low values to minimize the p otential for missed breast \ncarcinoma. \n\n \n \nThursday \nAbstract-based Programme \n \n 119  \nLimitations: The study is being limited by the retrospective stu dy design; and \nthat it was a two institutional-based study, so mul tiple institutional-based \nstudies are recommended. \nFunding for this study: The study has no source of funding \nEthics committee - additional information: The study has been ethically \napproved by the research center of the affiliated i nstitute \nAuthor Disclosures:  \nSahar Abdelkhalek Mansour: Nothing to disclose \nSamar Hussien: Nothing to disclose \nMohammed Mohamed Gomaa: Nothing to disclose \nYasmin Mohamed Nada: Nothing to disclose \nYoumna Kassab: Nothing to disclose \nMostafa Emara: Nothing to disclose \nSherif Taha: Nothing to disclose \nRasha Mohamed Kamal: Nothing to disclose \n \n \nAI-assisted Breast Mass Classification in Digital B reast Tomosynthesis \n(DBT): Applicability and insights from a single aca demic centre \n*G. Cura Curà*¹, G. Bartoli², M. Costa², E. Regini² , E. Puglisi², F. Piccione²,  \nF. Schettini², M. Durando², P. Fonio²; ¹Vercelli/IT , ²Torino/IT \n(gaiacuracura@gmail.com) \n \nPurpose or Learning Objective: In previous research, we trained a deep-\nlearning model to classify benign and malignant mas ses identified on DBT \nimages (convolutional neural network: efficientNetB 0; dataset: 448 masses, \nsize < 6 cm, 221 malignant, 227 benign; accuracy 94 %, sensitivity 95.6%, \nspecificity 91.7%). The aim of this study is to eva luate its applicability on breast \nmass lesions diagnostic work-up in clinical practic e. \nMethods or Background: In this single-centre multireader study, we \nprospectively collected DBT images from patients wi th biopsy-proven breast \nmasses (size < 6 cm). For each case, masses were ma nually delineated with \northogonal axes on the best focused slice in both D BT standard views. A \npreliminary set of 64 DBTs (46 benign, 18 malignant ) was reviewed by three \nindependent dedicated breast radiologists with diff erent experience, then \nassessed with the AI model. The software provides t he benign/malignant \nclassification combined with a prediction confidenc e score. The response of \nthe software was compared to biopsy results, focusi ng on BI-RADS \nclassification, error rates, inter-reader agreement , and reading time. \nResults or Findings: In 7% of cases, there was inter-reader disagreement  on \nthe software prediction. The model correctly classi fied 84% of masses, \nconfirming the 92.8% of lesions categorized as BI-R ADS 3 as benign. \nSoftware-assisted reading did not modify the readin g time compared to \nconventional methods. \nConclusion: In these preliminary results, the highest agreement  between \nradiologists and the AI model was observed with BI- RADS 3 lesions, \nhighlighting the benefit of software-assisted chara cterization of benign masses. \nHowever, variability in inter-reader agreement on t he software’s predictions \nlimits its reliability in real practice. Further in vestigations and model refinement \nare necessary to improve the model robustness. \nLimitations: Small sample size and single-centre study \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: Non applicable \nAuthor Disclosures:  \nManuela Durando: Nothing to disclose \nGaia Cura Curà: Nothing to disclose \nMatilde Costa: Nothing to disclose \nFrancesca Schettini: Nothing to disclose \nEugenia Puglisi: Nothing to disclose \nElisa Regini: Nothing to disclose \nFederica Piccione: Nothing to disclose \nGermana Bartoli: Nothing to disclose \nPaolo Fonio: Nothing to disclose \n \n \nMulti-site validation of an image-based AI breast c ancer risk model for \nmammography to drive personalized screening after a  negative screening \n*A. D. Lauritzen*¹, A. Rodriguez-Ruiz², N. Karsseme ijer², C. De Wolf³,  \nR. Mann², M. Nielsen¹, I. Vejborg⁴, M. Lillholm¹; ¹Copenhagen/DK, \n²Nijmegen/NL, ³Geneva/CH, ⁴Gentofte/DK \n(al@di.ku.dk) \n \nPurpose or Learning Objective: To validate the performance of an image-\nbased AI breast cancer risk model to stratify women  attending screening after \na negative screening. \nMethods or Background: Exams from women attending two European \nscreening programs (Denmark and Switzerland) and fr om a public U.S. \ndatabase (EMBED) were consecutively sampled. All ex ams were screen-\nnegative (cancer-free for 180 days) and had follow- up information of between  \n \n \ntwo and six years. Mammography exams were processed  by an AI breast \ncancer risk model (Transpara Risk, ScreenPoint Medi cal, trial version for \nresearch). The risk model computes three image biom arkers (suspicious \nfindings, volumetric breast density, breast texture ), and combined with age, it \ngenerates a five-year breast cancer risk score per exam. All exams were fully \nindependent from the development of the risk model.  Risk model AUCs were \ncomputed for each cohort along with sensitivity for  women with the highest \n10% risk and breast density, respectively. \nResults or Findings: In total, 98,084 exams were included (31,349, 17,44 5, \nand 49,290 from Switzerland, US, and Denmark, respe ctively) with 1,336 \nbreast cancers diagnosed within 5 years from screen ing. Images were \nacquired with machines from four manufacturers (Hol ogic, Siemens, GE, \nPhilips). The AUCs of the AI risk model were 0.73 ( 95% CI: 0.69-0.76), 0.74 \n(95% CI: 0.69-0.79) and 0.74 (95% CI: 0.73-0.76) fo r Switzerland, US, and \nDenmark, respectively. When simulating using risk t o offer supplemental \nimaging to 10% of women, after a negative screening , sensitivity was 37% \n(95% CI: 34%-39%), in comparison to 15% (95% CI: 13 -17%) when using \ndensity alone. \nConclusion: An image-based AI breast cancer risk model shows hi gh \naccuracy and robustness to stratify women attending  screening according to \nrisk and could support personalized screening with higher sensitivity than \nbreast density. \nLimitations: The retrospective study design is a limitation of t his study. \nFunding for this study: Supported in part by Eurostars (grant E9714 \nIBSCREEN) \nEthics committee - additional information: The Danish Patient Safety \nAuthority and Danish Data Protection Agency approve d this retrospective study \nand the use of relevant Danish data, and waived the  need for informed consent \n(ref. 3–3013–2118, addendum 2019/2023). \nAuthor Disclosures:  \nAndreas David Lauritzen: Nothing to disclose \nAlejandro Rodriguez-Ruiz: Employee: ScreenPoint Med ical \nMads Nielsen: Nothing to disclose \nMartin Lillholm: Nothing to disclose \nNico Karssemeijer: Employee: ScreenPoint Medical \nChris De Wolf: Nothing to disclose \nIlse Vejborg: Nothing to disclose \nRitse Mann: Nothing to disclose \n \n \nReplacing one radiologist with AI for independent d ouble reading in \nmammographic screening \n*M. B. Bergan*, M. Larsen, J. Gjesvik, N. Moshina, S. Sagstad, T. Hovda,  \nH. W. Koch, M. A. Martiniussen, S. Hofvind; Oslo/NO  \n(mbbe@kreftregisteret.no) \n \nPurpose or Learning Objective: The aim of this study was to explore how \nreplacing one radiologist with artificial intellige nce (AI) for independent double \nreading in mammographic screening would affect canc er detection. \nMethods or Background: This study sample consisted of 1,027,430 \nscreening examinations, including 5786 screen-detec ted cancers, that were \nindependently interpreted by two radiologists in Br eastScreen Norway, 2004-\n2018. The radiologists scored each breast from 1, n egative for abnormality, to \n5, high suspicion of malignancy, and score ≥2 was considered positive. All \nexaminations were processed by the AI system Lunit INSIGHT MMG version \n1.1.7.2, assigning a continuous malignancy score fr om 0, no risk, to 100, very \nhigh risk. Cancer detection was presented for the c ombination of one \nradiologist and AI at various AI thresholds for pos itive examinations. \nResults or Findings: Of all screen-detected cancers, 86.9% (5028/5786) w ere \nclassified as positive (score ≥2) by one radiologist. When defining 10% of the \nexaminations with the highest AI score as positive by AI, 79.9% (4622/5786) of \nthe screen-detected cancers and 7.5% (134/1783) of the interval cancers \nwould be detected. When 5% with the highest AI scor es were considered \npositive, 75.5% (4348/5786) of the screen-detected and 5.7% (102/1783) of \nthe interval cancers would be detected. In a scenar io where 1% of the \nexaminations were classified as positive by AI, 58. 2% (3369/5786) of the \nscreen-detected and 2.4% (42/1783) of the interval cancers would be detected. \nConclusion: At an AI threshold of 5%, replacing one of the radi ologists with AI \nin independent double reading of screening mammogra ms will reduce the \nreading volume by 50% at the cost of missing 24.5% of screen-detected \ncancers, but with the possibility of detecting 5.7%  of the interval cancers. \nLimitations: We assume that all cancers classified as positive b y the \nradiologist and AI were detected. \nFunding for this study: Funding was provided by the Norwegian Cancer \nSociety (Pink Ribbon) \nEthics committee - additional information: The study was approved by the \nRegional Committees for Medical and Health Research  Ethics (#2018/2574). \n \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 120  \nAuthor Disclosures:  \nNataliia Moshina: Nothing to disclose \nJonas Gjesvik: Nothing to disclose \nMarie Burns Bergan: Nothing to disclose \nHenrik Wethe Koch: Nothing to disclose \nTone Hovda: Nothing to disclose \nMarit Almenning Martiniussen: Nothing to disclose \nSilje Sagstad: Nothing to disclose \nSolveig Hofvind: Nothing to disclose \nMarthe Larsen: Nothing to disclose \n \n \n16:00-17:30 Research Stage 1 \nResearch Presentation Session: Cardiac \nRPS 1103 \nExploring cardiac imaging through \nquantitative MRI \n \nModerator \nN. Fink; Munich/DE  \n \n \nImproved and automated detection of papillary muscl e infarction using \njoint bright- and black-blood LGE MRI \n*T. Richard*, V. Nogues, T. Boulle, V. De Villedon De Naide, K. Narceau,  \nB. Durand, S. Sridi, H. Cochet, A. Bustin; Bordeaux /FR \n(theo.richard@ihu-liryc.fr) \n \nPurpose or Learning Objective: Papillary muscle infarction (PMI) has been \nlinked to significantly increased mortality, and is  a source of ventricular \narrhythmias and mitral regurgitation. Bright-blood LGE (PSIR) imaging is the \nclinical gold standard for myocardial fibrosis char acterization. However, the \nlimited contrast at the blood-scar interface makes PMI visualization often \nchallenging. Black-blood LGE imaging overcomes this  limitation by improving \nscar-to-blood contrast. Here, we aim to develop a n ovel co-registered joint \nbright- and black-blood LGE technology (SPOT) that could improve visual PMI \ndetection (visu-PMI) , while allowing an automated PMI detection algorithm \n(auto-PMI). \nMethods or Background: Short-axis 2D whole-heart PSIR and SPOT images \nwere collected on a 1.5T Siemens Aera system under breath-hold 12min post \ngadolinium injection. Auto-PMI included image acqui sition, statistics-based \nslice selection, left ventricular endocardium segme ntation, blood pool \npreprocessing, and fibrosis detection. 198 patients  participated to the study \nand were divided into an optimization dataset for a uto-PMI parameters \nselection, and a testing dataset to evaluate visu-P MI and auto-PMI \nperformances. Two radiologists assessed PMI on PSIR  and SPOT images. A \nconsensus reading was used as reference standard. N umber of patients with \nPMI were compared. Sensitivity and accuracy of both  sequences and auto-PMI \nwere measured. Inter- and intra-observer reproducib ility were reported. \nResults or Findings: Radiologists detected significantly more PMI with S POT \n(average increase: 30%). SPOT outperformed average PSIR sensitivity (93% \nvs. 75%) and accuracy (93% vs. 86%). Average inter-  and intra- reproducibility \nincreased with SPOT (79% vs. 74%, 97% vs. 88%). Aut o-PMI outperformed \nPSIR sensitivity (87%), while the accuracy equaled the SPOT average (86%). \nConclusion: Co-registered joint bright- and black-blood SPOT im aging allows \nfor improved PMI detection and opens a new door for  automated PMI \nassessment. \nLimitations: Further validation in larger cohort is warranted. V isu- and auto-\nPMI reliability depends mostly on contrast selectio n. \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nt ANR-22-CPJ2-\n0009-01, and from the European Research Council (ER C) grant \"SMHEART\" \nunder the European Union’s Horizon 2020 research an d innovation programme \n(grant agreement No101076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \n \n \n \n \n \n \n \n \nAuthor Disclosures:  \nVictor De Villedon De Naide: Nothing to disclose \nSoumaya Sridi: Nothing to disclose \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nKalvin Narceau: Nothing to disclose \nThibault Boulle: Nothing to disclose \nVictor Nogues: Nothing to disclose \nThéo Richard: Nothing to disclose \nBaptiste Durand: Nothing to disclose \n \n \nOptimizing Static B1+ Shimming for Cardiac MRI at 7  Tesla: Impact on \nImage Quality and Myocardial Strain \n*A. A. Peters*, K. Fischer, M. Hundertmark, C. Scha ub, G. Bonanno,  \nS. Schmitter, D. Günsch, C. Gräni, B. Jung; Bern/CH  \n \nPurpose or Learning Objective: The purpose of this work was to: 1) establish \nan efficient workflow for additional 7T-specific st atic B1+ shimming procedure \nas short as possible, 2) investigate whether a sing le set of B1+ shim values \ncalculated at the beginning of the exam provides re liable image quality over a \nregion of interest (ROI), 3) investigate whether sp atial resolution or field \nstrength has an influence on volumetric and myocard ial strain parameters. \nMethods or Background: Ten healthy volunteers underwent cine imaging at \n7T and 3T on the same day. The B1+ shimming process  at 7T used relative \nB1+ maps to minimize the coefficient of variation ( CV) within a ROI covering \nthe heart, with the constraint to maximize excitati on homogeneity. Image \nquality was assessed by two experienced readers usi ng a 4-point Likert scale, \nand quantitative measures such as left and right ve ntricular volumes and strain \nparameters were evaluated. \nResults or Findings: Results showed that B1+ shimming significantly \nimproved homogeneity, reducing the CV from 61.5% to  23.3%, while \nincreasing transmit efficiency. Image quality at 7T  exhibited more \ninhomogeneities compared to 3T, but these did not i mpact the clinical \nassessment of myocardial function. Quantitative ana lysis revealed higher \nmyocardial mass and smaller ventricular volumes at 7T, though these \ndifferences were minimal and insignificant regardin g clincial assessment. \nStrain parameters were comparable between 3T and 7T . \nConclusion: In conclusion, this study demonstrates that a fast and efficient \nworkflow for B1+ shimming at 7T can achieve diagnos tic image quality and \naccurate functional analysis comparable to 3T MRI. \nLimitations: - Small sample size \n- Small number of readers \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMoritz Hundertmark: Nothing to disclose \nAlan Arthur Peters: Nothing to disclose \nKady Fischer: Nothing to disclose \nChristoph Gräni: Nothing to disclose \nSebastian Schmitter: Nothing to disclose \nGabriele Bonanno: Nothing to disclose \nChristof Schaub: Nothing to disclose \nDominik Günsch: Nothing to disclose \nBernd Jung: Nothing to disclose \n \n \nChanges of myocardial extracellular volume fraction  measurements in \nacute versus chronic disease in a large animal infa rct model \n*M. C. Halfmann*¹, L. Van Der Meulen², M. W. Smulde rs², H. M. J. M. Nies²,  \nF. Prizen², C. Mihl², A. Varga-Szemes³, R. J. Holta ckers², T. S. Emrich¹; \n¹Mainz/DE, ²Maastricht/NL, ³Charleston, SC/US \n \nPurpose or Learning Objective: Cardiac MRI derived myocardial extracellular \nvolume fraction (ECV) is a reproducible imaging bio marker for myocardial \nfibrosis. However, well-controlled evidence on the influence of the timing of the \nscan in relation to the contrast injection is scarc e. Therefore, this study aimed \nto compare ECV measurements at different time point s after contrast injection \nin a large animal infarct model. \nMethods or Background: Cardiac MRI was performed at 1.5T while the \nanimals were ventilated and under general anesthesi a. Hematocrit levels were \ndrawn directly prior to the scan. Native short-axis  T1-maps of the left ventricle \nwere acquired. 7 and 30 minutes following an iv-adm inistration of 0.2 mg/kg \ngadobutrol, post-contrast T1-maps at identical slic e locations were acquired. \nECV was calculated for both global and separately f or the infarcted and remote \nmyocardium. Results were compared using Pearson’s c orrelation and paired \nsample Student’s t-tests. \n \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 121  \nResults or Findings: A total of 13 Yorkshire pigs with balloon catheter-\ninduced myocardial infarction were included in this  prospective study. \nHowever, 7 animals died before the MRI due to sever e arrhythmias and two \nanimals did not undergo the scan due to instability . Thus, MRI was successful \nin four pigs. Median time between infarction and ca rdiac MRI was 8 days [IQR \n8–8]. While there was a strong correlation between measurements at both time \npoints (r=0.94), ECV was significantly higher at 30  vs. 7 minutes (32.2±5.0% \nvs. 27.8±4.2%,P=.015). This was confirmed when asse ssing infarcted \n(56.9±11.4% vs. 43.7±9.1%,P=.018) and remote myocardium (28.4±2.8% vs. \n25.5±3.2%,P=.010) separately. \nConclusion: Myocardial ECV by cardiac MRI significantly increas es with \nincreasing time after contrast injection in a large  animal infarct model. A similar \neffect was observed in regions with only remote myo cardium and only infarcted \nmyocardium. \nLimitations: The limitation of the study is the limited number o f animals. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Animal handling complied with \nthe Dutch Law on Animal Experimentation and the Eur opean Directive on the \nProtection of Animals used for Scientific Purposes (2010/63/EU). This study \nwas approved by the Experimental Animal Committee o f Maastricht University \n(DEC2016-002). \nAuthor Disclosures:  \nCasper Mihl: Nothing to disclose \nFritz Prizen: Nothing to disclose \nRobert J. Holtackers: Nothing to disclose \nTilman Stephan Emrich: Advisory Board: Siemens Heal thineers Speaker: \nSiemens Healthineers \nMoritz Christian Halfmann: Nothing to disclose \nLara Van Der Meulen: Nothing to disclose \nMartin W. Smulders: Nothing to disclose \nAkos Varga-Szemes: Nothing to disclose \nHedwig M. J. M. Nies: Nothing to disclose \n \n \nCMR e' as a novel diagnostic biomarker of Asymptoma tic Left Ventricular \nDiastolic Dysfunction (ALVDD) \n*N. Mcveigh*¹, D. T. Ryan¹, F. Ryan², M. Ferre², J.  Mccambridge²,  \nM. Ledwidge², K. Mcdonald², J. Dodd¹; ¹Dublin 4/IE,  ²Dublin/IE \n \nPurpose or Learning Objective: Evaluate a novel cardiac MRI biomarker of \ndiastolic dysfunction, CMRe', in pre-clinical patie nts at risk of heart failure(HF). \nMethods or Background: 236patients from the PARABLE trial \n(NCT04687111) underwent CMR. Mitral annular relaxat ion velocity(CMRe’) \nwas measured at four mitral annular anchor points a nd compared with feature \ntracking analysis of radial, circumferential and lo ngitudinal diastolic strain rate \nand velocity as the gold-standard. Comparison were made with a control group \nof 25 age/gender-matched subjects. Comparisons were  made with \nindependent t-test, diagnostic accuracy was perform ed with receiver operator \ncurve analysis and predictors of diastolic dysfunct ion were analysed using \nlogistic regression. \nResults or Findings: LAVimax, LVEDVi, LVESVi and cardiac mass all \ndemonstrated significant increases between patient and control groups \n(p<0.001 for all). Peak diastolic longitudinal velo city was the only significant \nfeature tracking variable that differed between gro ups (p<0.001). LAVimax did \nnot correlate with any measured feature tracking pa rameter when adjusted for \nclinical, left ventricle and left atrial parameters . In similar multivariate analysis, \nCMRe’ correlated with diastolic radial, circumferen tial and longitudinal strains \nrates, as well as radial and longitudinal diastolic  velocity measurements \n(p<0.001). It also correlated with echo e’ (r=0.195 ,p=0.0069), LV mass \n(r=-.18,p=0.008), LAVimax (r=-.18,p=0.008) and BNP (r=-0.30,p<0.0001). \nLAVimax and total CMR e’ both exhibited high accura cy as independent \npredictors of diastolic dysfunction (AUC:0.89, 0.76 ,p<0.001 for both). \nCombined model (LAVImax and CMR e’ total) predicted  diastolic dysfunction \nwith an AUC = 0.99. LAVimax, CMR e’ and peak diasto lic longitudinal velocity \nwere independent predictors of diastolic dysfunctio n (p<0.001 for all), adjusted \nfor clinical and standard CMR parameters. \nConclusion: CMRe' is a precise imaging biomarker for ALVDD. Int egrating \nLAVimax and CMRe' holds promise in optimizing CMR m ethodologies for \nidentifying patients at risk of diastolic dysfuncti on. \nLimitations: Lack of BNP and Echo markers for the control group.  \nFunding for this study: This trial was supported by the Health Research \nBoard of the Government of Ireland, the European Co mmission Framework \nProgramme 7, the Heartbeat Trust CLG, and Novartis (the manufacturer of \nsacubitril/valsartan). Under the terms of the grant  from Novartis, the study was \nan investigator-led, Heartbeat Trust–sponsored clin ical trial. \nEthics committee - additional information: SVUH Ethics Committee \n \n \n \n \n \n \nAuthor Disclosures:  \nJonathan Dodd: Nothing to disclose \nKenneth Mcdonald: Nothing to disclose  \nFiona Ryan: Nothing to disclose \nDavid Thomas Ryan: Nothing to disclose \nMaria Ferre: Nothing to disclose \nNiall Mcveigh: Nothing to disclose \nJoe Mccambridge: Nothing to disclose \nMark Ledwidge: Nothing to disclose \n \n \nRight Ventricular Function Predicts Outcome in Hear t Failure with \nPreserved Ejection Fraction: Strain Analysis Derive d from MR Feature-\nTracking \n*L. Zhu*, J. He, S. Zhao, M. Lu; Beijing/CN \n \nPurpose or Learning Objective: To evaluate the association between right \nventricular (RV) strain parameters derived from car diac magnetic resonance \nfeature tracking (CMR-FT) and adverse outcomes in p atients with heart failure \nwith preserved ejection fraction (HFpEF). \nMethods or Background: Patients with HFpEF who underwent CMR \nexamination from January 2010 to December 2018 were  retrospectively \nenrolled. The primary endpoint was all-cause mortal ity. \nResults or Findings: A total of 1019 consecutive patients with HFpEF (ag e \n56.9 ± 12.3 years; 70% male) were enrolled in this study. During a median \nfollow-up of 7.8 years, 103 (10.1%) patients reache d the primary endpoint. In \nmultivariable Cox regression analysis, both RV glob al longitudinal and \ncircumferential strain were independent predictors of the primary endpoint \n(HRadj per 1% increase, 1.07 [95% CI: 1.02, 1.12; P  = .005] and 1.13 [95% CI: \n1.05, 1.21; P < .001], respectively). The full mode l based on clinical, \nconventional imaging, and RV strain variables for t he primary endpoint \nimproved the model discrimination (C-index = 0.794)  compared with the \nbaseline model based solely on clinical variables ( C-index = 0.716) and the \nmodel incorporating clinical and conventional imagi ng variables (C-index = \n0.760). In receiver operating characteristic analys is for the primary endpoint, \nthe addition of CMR-specific variables including la te gadolinium enhancement \nand FT RV strain yielded an improved area under the  curve for the baseline \nmodels (all P < .001). \nConclusion: RV global longitudinal and circumferential strain d erived from \nCMR-FT were independent predictors of adverse clini cal outcomes in patients \nwith HFpEF, providing incremental prognostic value over traditional clinical and \nCMR-derived risk markers. \nLimitations: This was a single-center, retrospective study. Echo cardiographic \nparameters, including E/e’, were excluded from the Cox regression analysis \ndue to missing values exceeding 50%. \nFunding for this study: The Beijing Natural Science Foundation (grant no. \n7242110) \nEthics committee - additional information: This study was approved by the \ninstitution ethics review board of Fuwai Hospital. \nAuthor Disclosures:  \nMinjie Lu: Nothing to disclose \nLeyi Zhu: Nothing to disclose \nShihua Zhao: Nothing to disclose \nJian He: Nothing to disclose \n \n \nNon-compaction Cardiomyopathy and Diastolic dysfunc tion \n*S. S. D. Dereli Bulut*, S. N. Emir; Istanbul/TR \n(ssanembulut@gmail.com) \n \nPurpose or Learning Objective: Non-compaction cardiomyopathy (NCCMP) \nis an uncommon disorder marked by increased trabecu lation of the ventricular \nwall and the existence of non-compacted myocardial regions. These \nanatomical alterations may hinder the ventricle's r elaxing capacity. Following \nthe preliminary evaluation by echocardiogram (TTE),  cardiac magnetic \nresonance imaging (CMR) is conducted for an in-dept h assessment. A non-\ncompacted to compacted myocardium ratio (N/C) excee ding 2.3 substantiates \nthe diagnosis. This study aims to assess individual s with NCCMP who \nunderwent CMR for signs of diastolic dysfunction (D D) and to explore the \ncorrelation between disease severity and DD. \nMethods or Background: This retrospective, single-center study comprised \n82 patients initially diagnosed with NCCMP based on  TTE data. Cardiac \nMagnetic Resonance imaging was conducted utilising a 1.5 Tesla MRI scanner \n(Avanto, Siemens). Morphological and functional eva luations including left \nventricular (LV) and right ventricular (RV) volume quantifications, cardiac \noutput (CO), ejection fraction (EF), LV mass, peak ejection rate (PER), and \npeak filling rate (PFR). Correlation analysis was p erformed among these \nmetrics. \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 122  \nResults or Findings: A substantial relationship was seen between the \nelevated N/C ratio and both PER and PFR (p<0.05; r= 0.31, r=0.21, \nrespectively). A positive association was noted bet ween the N/C ratio and \nvariations in LV mass assessed during the average a nd end-diastolic phases \n(r=0.35, p<0.05). No significant link was seen betw een the N/C ratio and LV EF \n(p>0.05), and no additional significant relationshi ps were detected. \nConclusion: The structural alterations in NCCMP may hinder vent ricular \nrelaxation, adversely impacting diastolic function,  which can be accurately \nidentified using CMR. Timely identification of DD i s essential for enhancing \npatients' quality of life. \nLimitations: The patient population was small, the study was pla nned \nretrospectively \nFunding for this study: The authors declared that this study has received n o \nfinancial support. \nEthics committee - additional information: This study was approved by the \nEthics Committee of the University Hospital \nAuthor Disclosures:  \nSevde Nur Emir: Nothing to disclose  \nSafiye Sanem Dereli Dereli Bulut: Nothing to disclo se \n \n \nImpact of Formalin Fixation on Biventricular Parame ters in Diffusion \nTensor CMR: Insights from a Miniature-Swine Model \n*L. Zhu*, J. Xu, H. Zhang, C. Cui, P. Sun, S. Zhao,  M. Lu; Beijing/CN \n \nPurpose or Learning Objective: To examine the impact of formalin fixation on \nbiventricular parameters derived from diffusion ten sor cardiac magnetic \nresonance (CMR) in a miniature-swine model, using h istological findings as the \nreference standard. \nMethods or Background: High-resolution ex-vivo diffusion tensor CMR data \nof one healthy miniature-swine were acquired at bas eline, and at 5- and 9-days \nafter the fixation with 10% neutral buffered formal in. Diffusion tensor CMR \nparameters were estimated using the AHA 16-segment model for the left \nventricular (LV) wall, and an 8-segment model for t he right ventricular (RV) \nwall. Histology with hematoxylin and eosin staining  was performed at 10 days \nfollowing formalin fixation to assess helix angles (HA) and HA gradients. \nResults or Findings: Diffusion tensor CMR data at baseline, and 5- and 9 -\ndays following formalin fixation were head-to-head analyzed. Subepicardial \nHAs became much more negative after fixation in bot h LV and RV walls, and \nendocardial HAs showed a positive increase, which l ed to a significant \nelevation in HA gradients. In the LV wall, mean dif fusivity values were slightly \nreduced during the first 5 days of fixation, follow ed by a marked decrease over \nthe subsequent 4 days; while in the RV wall, these values also reduced during \nthe first 5 days of fixation but did not change sig nificantly over the next 4 days. \nHAs derived from diffusion tensor CMR exhibited exc ellent consistencies with \nthose assessed with histology, among which baseline  HAs yielded the highest \ninterclass correlation coefficient of 0.953. \nConclusion: Formalin fixation had an impact on both fiber orien tations and \ndiffusion properties derived from diffusion tensor CMR, and baseline fiber \norientations assessed before fixation showed the be st consistency with \nhistology findings. \nLimitations: This study requires larger sample sizes to enhance the \nrobustness of the findings. \nFunding for this study: The Beijing Natural Science Foundation (grant no. \n7242110) \nEthics committee - additional information: Ethics approval was obtained \nfrom the Ethics Committee for Animal Study of Fuwai  Hospital. \nAuthor Disclosures:  \nMinjie Lu: Nothing to disclose \nLeyi Zhu: Nothing to disclose \nPeng Sun: Nothing to disclose \nShihua Zhao: Nothing to disclose \nJing Xu: Nothing to disclose \nHuaying Zhang: Nothing to disclose \nChen Cui: Nothing to disclose \n \n \nAutomated myocardial scar segmentation on joint bri ght- and black-\nblood late gadolinium enhancement images \n*T. Génisson*¹, V. De Villedon De Naide¹, B. Durand ¹, K. Narceau¹,  \nJ-D. Maes¹, P. Gut², H. Cochet¹, M. Stuber², A. Bus tin¹; ¹Bordeaux/FR, \n²Lausanne/CH \n(thais.genisson@ihu-liryc.fr) \n \nPurpose or Learning Objective: Bright-blood sequences are used to retrieve \nheart anatomy information, while black-blood late g adolinium enhancement \nhas shown promise for scar detection. However myoca rdial scar assessment, \ncrucial for accurate patient prognostic, is time-co nsuming, operator-dependent \nand labor-intensive. Here, we propose an artificial  intelligence-based method \nfor automated scar segmentation on joint bright- an d black-blood LGE (SPOT) \nimages. \nMethods or Background: A cohort of 70 patients (21% female, age range 28-\n81yo) with known or suspected ischemic heart diseas e was divided into a \ntraining (70%), validation (10%) and testing (20%) set. Breath-held short-axis \n2D whole-heart single-shot co-registered bright- an d black-blood SPOT, and \nreference phase-sensitive inversion recovery (PSIR)  images were collected on \na 1.5T (Siemens Aera) 12min post-contrast injection . An experienced \nradiologist manually performed scar segmentation on  SPOT and PSIR images. \nA transformer-based model automatically segmented l eft ventricular wall on \nSPOT bright-blood images and the contours were prop agated onto black-blood \nimages. Then, a U-net automatically segmented the s car within these contours. \nScar segmentation accuracy was assessed. Another ex perienced radiologist \ngraded the scar segmentation clinical quality (Like rt scale: 0=redo; 1=major \nadjustments; 2=minor adjustments; 3=no adjustments needed). Concordance \nbetween scar size assessed with manual PSIR and aut omated SPOT \nprocessing was evaluated. Scar segmentation times w ere recorded. \nResults or Findings: Scar segmentation was automatically achieved on \nSPOT in 0.14s per slice, reaching a global Dice of 76.1%. Scar segmentations \nwere rated 3 in 62%, 1 or 2 in 33% and 0 in only 5%  of the cases. No \nsignificant differences between scar sizes were fou nd when comparing with \nmanual PSIR processing (P<0.05). \nConclusion: The proposed method allows for fast, accurate and a utomated \nscar segmentation on SPOT images, achieving clinica l quality needed to better \nhelp guide therapy. \nLimitations: Validation in larger cohort is warranted. \nFunding for this study: This research was supported by funding from the \nFrench National Research Agency under grant agreeme nt ANR-22-CPJ2-\n0009-01, and from the European Research Council (ER C) grant \"SMHEART\" \nunder the European Union’s Horizon 2020 research an d innovation programme \n(grant agreement No101076351). \nEthics committee - additional information: The study was approved by the \nBiomedical Research Ethics Committee and all partic ipants provided informed \nconsent for participation. \nAuthor Disclosures:  \nVictor De Villedon De Naide: Nothing to disclose \nPauline Gut: Nothing to disclose \nAurelien Bustin: Nothing to disclose \nHubert Cochet: Nothing to disclose \nJean-David Maes: Nothing to disclose \nThaïs Génisson: Nothing to disclose \nKalvin Narceau: Nothing to disclose \nBaptiste Durand: Nothing to disclose \nMatthias Stuber: Nothing to disclose \n \n \nImpact of Reduced Spatial Resolution on Cardiac Str ain Using \nDeformable Registration and Feature-Tracking: A Pil ot Comparison \n*C. G. Glessgen*¹, T. Chitiboi², J. Wetzl², J-P. Va llee¹; ¹Geneva/CH, \n²Erlangen/DE \n \nPurpose or Learning Objective: Accelerated cardiac MRI cine sequences are \nessential for real-time imaging and for patients un able to hold their breath but \ncome with reduced spatial resolution. A comparison of strain measurements \nfrom high-resolution (HR) and low-resolution (LR) c ine data is presented, using \ntwo strain methods: deformable registration analysi s (DRA) and feature-\ntracking (FT). \nMethods or Background: Twenty patients with normal cardiac MRI findings \nwho underwent standard SSFP cines (HR; 0.8×0.8mm2) and compressed-\nsensing accelerated single-shot SSFP cines (LR; 2.4 ×2.4mm2) at 1.5 T were \nretrospectively analyzed. Breath-hold short-axis im ages were processed using \nDRA (TrufiStrain v2.4, Siemens Healthineers) and FT  (cvi42 v6.1, Circle CVI). \nGlobal radial (GRS) and circumferential strain (GCS ) were calculated for each \nimage/software pair. Correlations between HR and LR  images and between \nsoftware were assessed using Pearson's R coefficien t; concordance was \nevaluated using Bland-Altman analysis. \nResults or Findings: GCS correlation between HR and LR was stronger for \nDRA (r = 0.93, p < 0.05) than for FT (r = 0.68, p < 0.05). GRS correlations were \nsimilar for DRA and FT (r = 0.63 and r = 0.65, respectively, p < 0.05). Bland-\nAltman analysis showed a mean GCS difference betwee n HR and LR of 0.02 \n(LoA: -3.14 to 3.19) for DRA and -4.35 (LoA: -7.00 to -1.69) for FT; for GRS, a \nmean difference of 18.37 (LoA: 0.60 to 36.15) for D RA and 13.81 (LoA: 6.23 to \n21.38) for FT. \nConclusion: Cardiac strain measurements from LR and HR cines ar e well \ncorrelated but exhibit significant limits of agreem ent, which should be \nconsidered in clinical practice. DRA showed superio r correlation and narrower \nlimits of agreement for GCS, while FT demonstrated narrower but significant \nlimits for GRS. \nLimitations: This pilot work is limited by its sample size and t he absence of \npathological data. \nFunding for this study: None \nEthics committee - additional information: Waiver for informed consent  \n(ID: 01574) \n \n\n \n \nThursday \nAbstract-based Programme \n \n 123  \nAuthor Disclosures:  \nJean-Paul Vallee: Nothing to disclose \nCarl Guillaume Glessgen: Nothing to disclose \nTeodora Chitiboi: Employee: Siemens Healthineers \nJens Wetzl: Employee: Siemens Healthineers \n \n \nPrecision prediction of heart failure events in pat ients with NDLVC using \nmulti-parametric cardiovascular magnetic resonance \n*M. Jiang*, M. Lu; Beijing/CN \n(jiangmd1302@163.com) \n \nPurpose or Learning Objective: To assess whether left ventricular (LV) \nglobal longitudinal strain (GLS), derived from card iovascular magnetic \nresonance (CMR), is associated with (i) major heart  failure (HF) events, and (ii) \nsudden cardiac death (SCD) in patients with non-dil ated left ventricular \ncardiomyopathy (NDLVC). \nMethods or Background: We conducted a retrospective observational cohort \nstudy of patients with NDLVC assessed by CMR, inclu ding feature-tracking to \nassess LV GLS and late gadolinium enhancement (LGE) . Long-term \nadjudicated follow-up included (i) HF hospitalizati on, heart transplantation or \nHF death, and (ii) SCD or aborted SCD (aSCD). \nResults or Findings: Of 386 patients with NDLVC (mean age 45 years, 258 \nmen [66.8%], median LVEF 49% [46–54]) followed up f or a median 6.2 years, \n68 patients (17.6%) experienced HF events and 15 (3 .9%) experienced SCD \nor aSCD. Following adjustment in a multivariable mo del, the presence of LGE \nand LV GLS predicted the HF events (HR 1.95; 95% CI  1.17-3.27; p=0.011 vs. \nper % HR 1.14, CI 1.07–1.22, p<0.001). However, LV GLS was not associated \nwith SCD/aSCD, whereas LGE presence still was (unad justed HR 5.36, 95% \nCI 1.20–23.99, p=0.028). LVEF was neither associate d with HF events nor \nSCD/aSCD. \nConclusion: Multi-parametric CMR has utility for precision prog nostic \nstratification of patients with NDLVC. LV GLS strat ifies risk of progressive HF, \nwhile LGE stratifies both HF and SCD risk. \nLimitations: T1 mapping and extracellular volume calculation, wh ich were \nmore sensitive in detecting subtle myocardial alter ation and fibrosis, were not \nnot performed systematically in patients. Additiona l investigations are required \nto assess the potential diagnostic and prognostic s ignificance of mapping \ntechniques within this particular context. \nFunding for this study: None \nEthics committee - additional information: The Institutional Review Board \napproved this study, and written informed consent w as waived due to the \nretrospective nature of the study. \nAuthor Disclosures:  \nMinjie Lu: Nothing to disclose \nMengdi Jiang: Nothing to disclose \n \n \n16:00-17:30 Research Stage 2 \nResearch Presentation Session: \nEmergency Imaging \nRPS 1117 \nEmergency radiology: new technologies \nand workload challenges \n \nModerator \nA. Platon; Geneva/CH  \n(alexandra.platon@hcuge.ch) \n \n \nNon-invasive bullet characterization by material de composition in Photon \ncounting CT \n*B. M. Schaarschmidt*¹, J. Hegmanns¹, J. Wulff¹, V.  Haase², S. Faby²,  \nC. Bäumer¹, S. Zensen¹, J. Haubold¹, B. Hartung¹; ¹ Essen/DE, ²Forchheim/DE \n \nPurpose or Learning Objective: Gunshot deaths are a worldwide health \nconcern. Especially in patients with lodged bullets  or forensic analyses, image-\nbased bullet characterization is of major interest.  Therefore, the present study \ninvestigated bullet differentiation from different materials using photon counting \ncomputed tomography (PCCT). \nMethods or Background: Using a new research scan mode, six lead and \nthree brass bullets were analyzed on a NAEOTOM Alph a PCCT system \n(Siemens Healthineers, Forchheim, Germany). For eac h scan, a set of four \nimages was reconstructed using four different energ y thresholds of the \ndetector (20, 55, 72, and 90 keV). Two independent readers placed three \ncircular regions of interest (ROI) on the 20 keV th reshold images on an OsiriX \nWorkstation (Pixmeo SARL, Bernex, Switzerland). The se ROIs were then \nautomatically duplicated to the other threshold ima ges. Dual energy indices \n(DEIs) were computed for the energy threshold pairs  of 20/90 keV, 55/90 keV, \nand 72/90keV based on the measured HUmean and HUmax  values. \nResults or Findings: DEIs of lead and brass projectiles differed signifi cantly, \nmost notably for the 20/90 keV DEI HUmean (Qr40): l ead: -0.085±0.021, brass: \n0.024±0.048, p<0.001; HUmax (Qr40): lead: -0.093±0.011, brass: \n0.023±0.057, p<0.001). For the 55/90 keV and 72/90 keV DEIs, differences \nbetween the two projectile materials decreased, but  remained statistically \nsignificant. \nConclusion: In the DEIs derived from different energy threshold  images \nobtained by PCCT, significant differences could be observed between lead and \nbrass bullets. Therefore, PCCT might be a potential  technique for bullet \nmaterial analysis in both clinical and forensic ima ging. \nLimitations: Phantom study only. \nFunding for this study: The study was performed at Siemens Healthineers \nfacilities in Forchheim, Germany, with support from  Viktor Haase and \nSebastian Faby (employees of Siemens Healthineers A G). \nEthics committee - additional information: As no human or animal subjects \nwere investigated in this study, no approval by the  local ethics committee was \nnecessary. \nAuthor Disclosures:  \nBenno Hartung: Nothing to disclose \nViktor Haase: Employee: Siemens Healthineers AG \nBenedikt Michael Schaarschmidt: Nothing to disclose  \nJohannes Haubold: Nothing to disclose \nSebastian Zensen: Nothing to disclose \nChristian Bäumer: Nothing to disclose \nJan Hegmanns: Nothing to disclose \nJörg Wulff: Nothing to disclose \nSebastian Faby: Employee: Siemens Healthineers AG \n \n \nVirtual monoenergetic images from photon-counting d etector CT in \nthoracic trauma: Improved discriminability of sever e lung injury and \natelectasis at low-keV energy levels \nH-L. Kaatsch, *B. V. Becker*, D. Dillinger, J. Piec hotka, C. Schreyer,  \nR. Schwab, D. Overhoff, S. Waldeck; Koblenz/DE \n \nPurpose or Learning Objective: The aim of this study was to assess the \nusability of virtual monoenergetic images (VMIs) de rived from photon-counting \ndetector CT (PCD-CT) for discriminability of severe  lung injury and atelectasis \nafter thoracic trauma. \nMethods or Background: We retrospectively selected 20 polytraumatized \npatients, who underwent contrast-enhanced whole-bod y PCD-CT and showed \nsimultaneous presence of trauma-associated atelecta sis and pulmonary injury. \nVMIs were reconstructed from 40 to 120 keV at 10 ke V increments. \nQuantitative image analysis was performed based on density measurements \nand the calculation of injury-to-atelectasis contra st-to-noise ratio (CNR)). Three \nradiologists rated subjective discriminability, noi se perception and overall \nimage quality by means of a 5-point Likert scale. \nResults or Findings: CT values for severe lung injury and atelectasis di ffered \nsignificantly at each keV level (p < 0.001) with a gradual decrease for \natelectasis from 342 ± 97 HU at 40 keV to 69 ± 15 HU at 120 keV and a near \nconstant behavior for severe lung injury from 42 ± 49 HU at 40 keV to 44 ± 22 \nat 120 keV. In line with this, the highest injury-t o-atelectasis CNR was achieved \nat 40 keV (3.97) with a continuous decline down to 120 keV (1.21). CNR \nreached no significant differences between 40 and 5 0 keV as well as 110 and \n120 keV (p >0.05), whereas all other pairings were significantly different \n(p<0.05). The best ratings for subjective discrimin ability were reported for VMIs \nat 40 keV, whereas VMIs at 60-70 keV provided the o ptimal noise perception \nand overall image quality. \nConclusion: Low-keV VMIs at 40-50 keV from PCD-CT considerably improved \nthe discriminability of severe lung injury and atel ectasis after thoracic trauma. \nLimitations: Lack of a validated reference standard for lung inj ury vs. \natelectasis, retrospective study design and small s ample size. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study is a retrospective \nsingle-centre analysis that has been approved by th e local ethics committee of \nthe chamber of physicians Rhineland-Palatinate in M ainz, Germany (number \n2022-16314). \nAuthor Disclosures:  \nJoel Piechotka: Nothing to disclose \nDaniel Overhoff: Nothing to disclose \nHanns-Leonhard Kaatsch: Nothing to disclose  \nRobert Schwab: Nothing to disclose \nBenjamin Valentin Becker: Nothing to disclose \nChristof Schreyer: Nothing to disclose \nStephan Waldeck: Nothing to disclose \nDaniel Dillinger: Nothing to disclose \n\n \n \nThursday \nAbstract-based Programme \n \n 124  \nBody composition parameters in initial CT imaging o f mechanically \nventilated trauma patients: Single-centre observati onal study \n*H-J. Meyer*, T. Dermendzhiev, T. Denecke, M. Struc k; Leipzig/DE \n(jonas90.meyer@web.de) \n \nPurpose or Learning Objective: Body composition parameters provide \nrelevant prognostic significance in critical care c ohorts and cancer populations. \nPublished results regarding polytrauma patients are  inconclusive to date. The \ngoal of this study was to analyse the role of body composition parameters in \nseverely injured trauma patients. \nMethods or Background: All consecutive patients requiring emergency \ntracheal intubation and mechanical ventilation befo re initial computed \ntomography (CT) at a level-1 trauma centre over a 1 2-year period (2008-2019) \nwere reanalysed. The analysis included CT-derived b ody composition \nparameters based upon whole-body trauma CT as progn ostic variables for 30-\nday mortality, intensive care unit length of stay ( ICU LOS) and mechanical \nventilation duration. \nResults or Findings: Four hundred seventy-two patients (75% male) with a  \nmedian age of 49 years, median injury severity scor e of 26 and 30-day \nmortality rate of 22% (104 patients) met the inclus ion criteria and were \nanalysed. Regarding body composition parameters, 23 1 patients (49%) had \nvisceral obesity, 75 patients had sarcopenia (16%) and 35 patients had \nsarcopenic obesity (7.4%). After adjustment for sta tistically significant \nunivariable predictors age, body mass index, sarcop enic obesity, visceral \nobesity, American Society of Anesthesiologists clas sification ≥3, injury severity \nscore and Glasgow Coma Scale ≤ 8 points, the Cox proportional hazard model \nidentified sarcopenia as significant prognostic fac tor of 30-day mortality \n(hazard ratio 2.84; 95% confidence interval 1.38-5. 85; P = 0.004), which was \nconfirmed in Kaplan-Meier survival analysis (log-ra nk P = 0.006). \nConclusion: In a multivariable analysis of mechanically ventila ted trauma \npatients, CT-defined sarcopenia was significantly a ssociated with 30-day \nmortality whereas no associations of body compositi on parameters with ICU \nLOS and duration of mechanical ventilation were obs erved. \nLimitations: It is a single-center study with a retrospective de sign. \nFunding for this study: None \nEthics committee - additional information: Approval by the ethics committee \nat the Medical Faculty, Leipzig University, Leipzig , Germany (IRB00001750, \nproject ID 441/15ek, September 14, 2020) \nAuthor Disclosures:  \nTimm Denecke: Nothing to disclose \nManuel Struck: Nothing to disclose \nTihomir Dermendzhiev: Nothing to disclose \nHans-Jonas Meyer: Nothing to disclose \n \n \nAI Denoising Enhances Image Quality and Diagnostic Accuracy While \nReducing Radiation Exposure in Prospective LDCT Sca ns for Acute \nAbdomen \n*A. S. Brendlin*, U. Schmid, S. Afat; Tübingen/DE \n \nPurpose or Learning Objective: To identify the optimal low-dose CT protocol \nthat minimizes radiation exposure while preserving high image quality and \ndiagnostic accuracy in patients presenting with acu te abdomen. \nMethods or Background: A prospective, randomized study was conducted \nwith 180 patients assigned to one of three CT proto cols. Each protocol \nincluded high-dose (HD) and low-dose (LD) scans rec onstructed using Iterative \nReconstruction (IR2) and AI Denoising (AID). Subjec tive image quality was \nassessed by radiologists evaluating diagnostic conf idence, contrast, and \nsharpness. Objective image quality metrics, includi ng noise and contrast-to-\nnoise ratio (CNR), were measured. Diagnostic accura cy was evaluated through \nsensitivity and specificity calculations for detect ing abdominal pathologies. \nResults or Findings: Protocol 2 LD achieved the lowest Size-Specific Dos e \nEstimate (SSDE) at 4.83 mGy, significantly reducing radiation dose compared  \nto Protocols 1 and 3 (P < 0.001). AID significantly enhanced subjective image  \nquality in LD scans across all protocols, with Prot ocols 2 and 3 achieving \nratings comparable to HD scans with IR2. Objective assessments showed that \nAID substantially reduced image noise and increased  CNR in LD scans, with \nProtocol 2 LD exhibiting the highest CNR. In outcom e analysis, both Protocols \n1 and 2 demonstrated 100% sensitivity and specifici ty in LD scans with AID, \neliminating false negatives and matching the diagno stic performance of HD \nscans. Protocol 3 maintained high diagnostic accura cy across all doses and \nreconstruction methods. \nConclusion: Protocol 2 with AID emerges as the most effective l ow-dose CT \nstrategy, offering significant radiation dose reduc tion while maintaining superior \nimage quality and diagnostic accuracy. \nLimitations: - Single-Center Study with Limited Sample Size - Po tential \nObserver Bias in Subjective Assessments - Lack of L ong-Term Outcome Data \n- Equipment and Protocol Specificity \nFunding for this study: None \n \n \n \nEthics committee - additional information: University Hospital Tuebingen \nAuthor Disclosures:  \nAndreas Stefan Brendlin: Nothing to disclose \nSaif Afat: Nothing to disclose \nUlrich Schmid: Nothing to disclose \n \n \nCan perfusion-derived cerebral CT angiography repla ce routine cerebral \nCT angiography by using artificial intelligence ite rative reconstruction for \nacute ischemic stroke patients? \nJ. Xie¹, *T. Wang*², G. Zhang², J. Huang¹, M. Wang¹ , Y. Lin¹; ¹Taizhou/CN, \n²Shanghai/CN \n(tiantian.wang@cri-united-imaging.com) \n \nPurpose or Learning Objective: To investigate the feasibility and realiability \nof replacing routine cerebral CT angiography (CTA) with CT perfusion (CTP)-\nderived cerebral CTA by using artificial intelligen ce iterative reconstruction \n(AIIR) for acute ischemic stroke (AIS) patients. \nMethods or Background: Forty-nine patients (33-93 y, male: 35) with AIS \nundergoing a cerebral CTP and a routine CTA were pr ospectively collected. \nCTA images derived from CTP at the arterial phase ( 100kVp/150mAs) were \nreconstructed with hybrid iterative (Group A1) and AIIR (Group A2), whereas \nroutine CTA images (100kVp/ref. 200mAs) were obtain ed with hybrid iterative \nreconstruction (Group B). Two radiologists independ ently located the \nresponsible vessels, with digital subtraction angio graphy (DSA) as reference \nstandard. They further graded the image noise, shar pness of the vascular \nedge, small vessel visibility, and overall diagnosa bility using a five-point Likert \nscale (1: poor, 5: excellent). Objective parameters , including the SNR and CNR \nof the internal carotid artery, the middle cerebral  artery, and the basilar artery, \nwere also calculated. \nResults or Findings: The diagnosis of responsible vessels was consistent  by \nboth radiologists, where the diagnostic accuracy of  Groups A2 and B were \ncomparable (47/49, 95.92%) and higher than that in Group A1 (43/49, \n87.76%). The inter-observer agreement was excellent  (κ = 0.84).All subjective \nscores were significantly higher in Group A2 than t hose in Groups A1 and B \n(all p < 0.017), especially for the small vessel vi sibility (4.9 ± 0.2 vs.2.8± 0.4 vs. \n4.2 ± 0.5). AIIR significantly reduced noise, leading to a significantly higher \nSNR and CNR for all arteries in Group A2 than those  in Groups A1 and B (all p \n< 0.017). \nConclusion: Perfusion-derived cerebral CTA by use of the AIIR a chieves \ncomparable diagnostic image quality to that of the routine cerebral CTA for AIS \npatients. \nLimitations: Not applicable \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nJicheng Xie: Nothing to disclose \nTiantian Wang: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nMinke Wang: Nothing to disclose  \nJinbiao Huang: Nothing to disclose \nYouyou Lin: Nothing to disclose \n \n \nReal life performance of a commercially available A I for post-traumatic \nintracranial haemorrhage detection on CT-scans: a s upportive tool \n*L. Mabit*, G. Herpe; Poitiers/FR \n(leo.mabit@outlook.fr) \n \nPurpose or Learning Objective: Investigate the real-world performance of \nqER.AI, an artificial intelligence-based CT haemorr hage detection tool, in a \npost-traumatic population. \nMethods or Background: Retrospective monocentric observationnal study of \na dataset of consecutively acquired head CT scans a t the emergency radiology \nunit to explore a brain trauma. AI performance was compared to groundtruth \ndetermined by expert consensus. A subset of nighshi ft cases with radiological \nreport of junior resident was compared to AI result s and groundtruth. \nResults or Findings: 682 head CT scan were analyzed. AI demonstrated a \nsensitivity of 88.8% and specificity of 92.1% overa ll, with a positive predictive \nvalue of 65.4% and a negative predictive value of 9 8%. AI's performance was \ncomparable to junior residents in detecting ICH, wi th the latter showing a \nsensitivity of 85.7% and a high specificity of 99.3 %. Interestingly, the AI \ndetected two out of three ICH cases missed by junio r residents. When AI and \nresidents performances were combined, the sensitivi ty improved to 95.2%, and \nthe overall accuracy reached 98.8%. \nConclusion: This study shows a better performance of AI and rad iologist \nresident associated than each one alone. These resu lts are encouraging to \nrethink the radiological workflow and the future of  triage of this large population \nof brain traumatised patients in emergency unit. \n \n\n \n \nThursday \nAbstract-based Programme \n \n 125  \nLimitations: The limitations of the study are exclusion of some CTs due to \nlogistics issues (potential selection bias), and a low number of positive ICH \ncases in the nightshift subset. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was reviewed and \napproved by the Institutional Review Board of CERIM  (CRM-2401-400). \nAuthor Disclosures:  \nGuillaume Herpe: Nothing to disclose \nLéo Mabit: Nothing to disclose \n \n \nIncreasing workload in emergency radiology: A decad e-long trend \nanalysis using Large Language Models \n*M. T. Meyer*, E. M. Merkle, T. Heye; Basel/CH \n \nPurpose or Learning Objective: This study aims to identify trends explaining \nthe significant rise in CT examinations ordered by the emergency department \n(ED) of a tertiary hospital in the last decade (201 4 to 2023) using large \nlanguage models (LLM). \nMethods or Background: Radiology Information System data were extracted \nfor CT scans performed on ED patients between 2014- 2023. The open-source \nLlama 3.1 405B LLM was used to classify each report  into YES (expected \npathology confirmed), NO (no relevant pathology), a nd OTHER (unexpected \nbut relevant pathology found). Trends were analyzed  by body-region, focusing \non patients older than 65 years, as this group show ed the highest increase in \nimaging. \nResults or Findings: Over the past decade, the number of CT scans \nincreased 2.2 times for ED patients, with a more pr onounced rise for patients \nover 65 years (2.4). Preliminary results show that the LLM achieved an \naccuracy of 84% compared to a manual review of 500 randomly sampled \nreports. Scans of abdomen/pelvis (YES in 49-60%) an d thorax/abdomen/pelvis \n(YES in 54-63%) remained stable with minor annual f luctuation. CT scans of \nthe skull in polytrauma patients showed steady incr ease in NO findings from \n43% (2016) to 59% (2022). CT of the skull showed a gradual increase in NO \nfindings from 57% (2015) to 66% (2022). CT of the t horax remained relatively \nstable (YES 41-46% from 2014-2019), with notable pe aks during the COVID-\npandemic (52% in 2021). \nConclusion: Diagnostic yield across different organ groups vari es greatly with \nhigh proportions of NO findings in skull CTs (typic ally to rule-out bleeding). \nOver time, unremarkable CT scans, particularly of t he skull, have increased, \nwhile trends for other body regions (e.g., thorax/a bdomen/pelvis), were less \ndistinct. \nLimitations: The results are preliminary. Accuracy of the LLM is  limited, which \nmight lead to over-/underestimation of trends. \nFunding for this study: None \nEthics committee - additional information: The study is retrospective and \nGeneral Consent is available. \nAuthor Disclosures:  \nTobias Heye: Nothing to disclose \nElmar M. Merkle: Nothing to disclose \nManfred Tobias Meyer: Nothing to disclose \n \n \nIncreasing On-Call Workload for Radiology Trainees:  A Five-Year \nAnalysis in a Tertiary Referral Centre \n*P. Rohan*, H. Briody, C. Mccarthy, M. M. Morrin; D ublin/IE \n(pat_rohan@outlook.com) \n \nPurpose or Learning Objective: The demand for acute diagnostic radiology \nservices during on-call hours continues to increase , placing strain on radiology \ntrainees. This study aims to assess the on-call wor kload for Radiology \nSpecialist Registrars (SpRs) over a five-year perio d at a model 4 tertiary \nreferral centre in Ireland. \nMethods or Background: A retrospective review was conducted, analysing \nthe volume of computed tomography (CT) scans perfor med during on-call \nhours (weekdays 17:00-08:00, weekends, and public h olidays) across three \nmonths—January, July, and November—from 2019 to 202 3. Data were \ncategorized by study type, focusing on Emergency De partment (ED) requests \nand key scan types, including non-contrast brain CT  (NCB), CT \nabdomen/pelvis (CTAP), CT thorax/abdomen/pelvis (CT  TAP), and cerebral \nstroke protocol (\"FAST\"). Statistical analysis incl uded percentage changes and \npaired t-tests. \nResults or Findings: Between 2019 and 2023, there was a 25% increase in \nthe total number of on-call CTs performed (p = 0.09 8), with a significant 46% \nrise in ED-related studies (p < 0.05). Significant increases were observed in \nNCBs (38%, p < 0.05) and CT TAPs (220%, p < 0.05). The number of CTs \nperformed after midnight increased by 82% (p = 0.05 5), while FAST protocol \nCTs rose by 41% (p = 0.056), reflecting the growing  incidence of stroke in \nIreland. \n \n \nConclusion: The increasing on-call workload for Radiology SpRs,  particularly \ndriven by ED requests and the rising demand for str oke imaging, underscores \nthe need for careful planning and resource allocati on to manage future \ndemand. Addressing trainee workflows is essential t o sustaining radiology \nservices while maintaining high standards of care. \nLimitations: This retrospective study is limited to on-call CT w orkload. Other \nmodalities like ultrasound, MRI, and X-rays, as wel l as communication burdens \nwith clinical teams, were not assessed and should b e explored in future \nstudies. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Ethics committee approval not \nrequired. \nAuthor Disclosures:  \nCathal Mccarthy: Nothing to disclose \nMartina M Morrin: Nothing to disclose \nHayley Briody: Nothing to disclose \nPat Rohan: Nothing to disclose \n \n \nTrends in CT examination utilization in the emergen cy department during \nand after the COVID-19 pandemic \n*F. Kempter*, D. Jäschke, J. Vosshenrich, B. Ceresa , T. Heye; Basel/CH \n \nPurpose or Learning Objective: To investigate the impact of COVID-19 \npandemic-related measures on trends and volume in C T examinations \nrequested in the emergency department. \nMethods or Background: CT examinations of the head, chest, and/or \nabdomen-pelvis (n=161,008), and chest radiographs ( n=113,240) performed at \nour tertiary care hospital between 01/2014- 12/2023  were retrospectively \nanalyzed. CT examinations (head, chest, abdomen, du al- region and \npolytrauma) and chest radiographs requested by the emergency department \nduring (03/2020-03/2022) and after the COVID-19 pan demic (04/2022- \n12/2023) were compared to a pre-pandemic control pe riod (02/2018-02/2020). \nAnalyses included CT examinations per emergency dep artment visit, and \nprediction models based on pre-pandemic trends and inpatient data. A regular \nexpressions text search algorithm determined the mo st common clinical \nquestions. \nResults or Findings: The usage of dual-region and chest CT examinations \nwere higher during (+116,4% and +115.8%, respective ly; p<0.001) and after \nthe COVID-19 pandemic (+88,4% and +70.7%, respectiv ely; p<0.001), \ncompared to the control period. Chest radiograph us age decreased (-54.1% \nand -36.4%, respectively; p<0.001). The post-pandem ic overall CT \nexamination rate per emergency department visit inc reased by 4.7%. The \nprediction model underestimated (p<0.001) the growt h (dual-region CT: 22.3%, \nchest CT: 26.7%, chest radiographs: -30.4%), and th e rise (p<0.001) was \nhigher compared to inpatient data (dual-region CT: 54.8%, chest CT: 52.0%, \nCR: -32.3%). Post- pandemic, the number of clinical  questions to rule out \n“pulmonary infiltrates”, “abdominal pain” and “infe ction focus” increased up to \n235.7% compared to the control period. \nConclusion: Following the COVID-19 pandemic, chest CT and dual- region CT \nusage in the emergency department experienced a dis proportionate and \nsustained surge compared to pre-pandemic growth. \nLimitations: Single-center, retrospective design limits generali zability. CT as \nstandard imaging for suspected COVID-19 pneumonia ( 2020–2022) may \nexplain the shift. Only clinical questions, not dia gnoses, were analyzed. \nDespite a large sample size, data mislabeling is po ssible. \nFunding for this study: Not applicable. \nEthics committee - additional information: The local ethics committee of \nnorthwestern and central Switzerland (EKNZ, project  ID 2022-01016) approved \nthis study. \nAuthor Disclosures:  \nTobias Heye: Nothing to disclose \nDominik Jäschke: Nothing to disclose \nFelix Kempter: Nothing to disclose \nBenjamin Ceresa: Nothing to disclose \nJan Vosshenrich: Nothing to disclose \n \n \nAn 18-year Retrospective Analysis of Urgent Inpatie nt and Emergency CT \nReporting at a UK Local General Hospital \n*A. Gmati*, Z. Foster, M. Mobley; Warwick/UK \n(aimen.gmati@outlook.com) \n \nPurpose or Learning Objective: Radiology has transformed how acute \nmedicine is practiced, with growing imaging demands  placing significant strain \non radiology departments in the UK. This study retr ospectively examines 18 \nyears of urgent inpatient and emergency CT reportin g at South Warwickshire \nUniversity NHS Foundation Trust (SWFT) to assess tr ends in workload, \nstaffing, and operational changes over this period.  \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 126  \nMethods or Background: Data were extracted from the Radiology Information \nSystem (RIS) at SWFT, encompassing approximately 14 0,000 urgent scans \nperformed since 2007. These were analysed to show c hanges over time such \nas seasonal variability and peaks in workload throu ghout the week, as well as \nthe overall trend for increased imaging. \nResults or Findings: Urgent CT workload at SWFT has significantly increa sed \nover the study period, reflecting national trends. This timescale covers a period \nof rapid growth and operational change at SWFT, suc h as the introduction of \nweekend shifts and the implementation of outsourced  teleradiology for out-of-\nhours reporting. Typical daily reporting numbers ha ve increased from 12 scans \nper day in 2007 to over 50 in 2024. Staffing has no t kept pace, with WTE \nconsultant numbers dropping from 12 in 2019 to 9.5 in 2023. Despite \noutsourcing, many departments continue to struggle with growing diagnostic \ndemands. \nConclusion: Across the UK, CT usage surged from 1 million scans  in 1997 to \nover 6.7 million in 2021. Rising CT demand at SWFT reflects broader national \ntrends, with increasing pressure on radiology servi ces. The UK’s Royal College \nof Radiologists predicts a shortage of 1,669 radiol ogists by 2025, highlighting a \nnational critical staffing issue. Operational chang es such as weekend working \nmay provide temporary relief but place increasing d emands on the same pool \nof radiologists. Systemic reforms and increased inv estment are essential to \nmeet future diagnostic demands. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nZoe Foster: Nothing to disclose \nMark Mobley: Nothing to disclose  \nAimen Gmati: Nothing to disclose \n \n \nIs population aging behind the increasing workload in emergency \nradiology? \n*J. Sarnecki*, M. T. Meyer, E. M. Merkle, T. Heye; Basel/CH \n(jedrzej.sarnecki@gmail.com) \n \nPurpose or Learning Objective: To investigate the trends in imaging in \nemergency patients over the last decade. \nMethods or Background: 190'028 emergency CT examinations from \n1/1/2015 to 31/12/2023 of patients aged 18-100 year s, performed within 24h of \nthe request, were included. The examination time wa s stratified into dayshift \n(8-17:30h), late-shift (17:30-22h), night-shift (22 -8h) and weekend-dayshift \n(Saturday/Sunday 8-17:30h). The patients were divid ed into two age groups \n(18-64 versus 65-100 yrs.) and binned using 5-year increments. The \npercentage increase in imaging numbers was calculat ed based on 2015. \nResults or Findings: A sharp overall increase in CT imaging numbers was \nobserved in 2020 (18-64 yrs.: 127.0% vs. 65-100 yrs .: 155.3%), peaking in \n2022 (18-64 yrs.: 182.4% vs. 65-100 yrs.: 235.6%). The mean age in the 18-64 \nyrs. group increased from 44.8 (2015) to maximally 46.3 (2022), but did not \nincrease for 65-100 yrs. (79.2 vs. 79.3 yrs.). The day-shift imaging numbers \nshowed a linear steady increase, whereas the night- shift examinations showed \nthe strongest increase, peaking in 2023 with 267.5%  (18-64 yrs) and 403.0% \n(65-100 yrs) followed by changes in weekend and lat e-shift CT numbers (18-64 \nyrs.: 119.8%;127.1% vs. 65-100 yrs.: 173.1%; 139.0% ). Between 2020-2023 \nfor the younger age group, age-bins 55-59 (199.5-26 0.9%) and 60-64 (140.8-\n200.6%) showed the largest relative increase compar ed to 2015 for late-, night- \nand weekend-dayshift CTs combined (all age bins 18- 64yrs: 30-112.2%). In \nthe 65-100 yrs. group, the largest relative increas e was seen for patients 95-\n100 years old (318.2-360.0%) compared to other 5-ye ar age bins (145.7-\n276.8%). Chest, Chest-Abdomen-Pelvis and Polytrauma  CT showed the \ngreatest increase. \nConclusion: The results support that a change in imaging was tr iggered and \nsustained by the COVID-19 pandemic with a dispropor tionate increase in CT \nimaging volume overall and especially for older pat ients during non-day-shifts. \nLimitations: Retrospective, single-center study. \nFunding for this study: No funding. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nTobias Heye: Nothing to disclose \nElmar M. Merkle: Nothing to disclose \nJędrzej Sarnecki: Nothing to disclose \nManfred Tobias Meyer: Nothing to disclose \n \n \n \n \n \n \n \n \n \nThe increase and inappropriate requests for cranial  CT scans in \nemergency departments contribute to overuse and dec rease test \naccuracy \n*A. Martínez López*, G. Pagán Vicente, E. Otón Gonz ález, H. Ortiz Mayoral,  \nE. C. Cotillo Ramos, M. S. Canales, B. Molina-Lozan o, J. Plasencia Martínez, \nI. Santiago Suárez; Murcia/ES \n(amabandrea6@gmail.com) \n \nPurpose or Learning Objective: The number of non-traumatic urgent cranial \ncomputed tomography (NT-UCCT) is exponentially incr easing but limited \nresearch has been conducted on the quality of clini cal justification. We aimed \n(1) to assess how clinical information in the elect ronic NT-UCCT request \nagreed with that provided in the emergency departme nt discharge summary \nand (2) to analyze the potential effect of those di screpancies on the NT-UCCT \noverload. \nMethods or Background: Patients undergoing NT-UCCT in 2017-2021 were \nrandomly selected for this retrospective study. Sig ns and symptoms (S/S) in \nelectronic request and emergency department dischar ge summary, acute and \nrelevant computed tomography (CT) findings, and fin al diagnosis at emergency \ndepartment discharge summary were collected. Concor dance between both \nand their association with CT findings and final di agnosis were analyzed. \nResults or Findings: We recruited 156 patients: 80 men; mean age, 55. \nAcute, relevant CT findings were detected in 28 cas es (17.9%). The final \ndiagnosis was neurological disease 46 (29.5%), non- neurological disease 58 \n(37.2%), and no definitive diagnosis 51 (32.7%). Fu ll agreement between the \nelectronic request and emergency department dischar ge summary occurred in \n36 patients (23.1%). Motor deficit was the most fre quent false positive S/S (18; \n11.54%), having low positive predictive value (30.3 0%; 95%CI 15.59-48.71%) \nand worst association with acute relevant CT findin gs than when true positive \n(OR 2.54; 95%CI 0.04-6.21 vs. OR 6.26, 95%CI 2.21-1 7.78). Nausea/vomiting \nwas the third most common false negative S/S (13; 1 0.26%) and reduced the \nlikelihood of acute relevant CT findings (OR 0.126;  95%CI 0.016-0.971; p = \n0.020). \nConclusion: Discrepancies between electronic request and emerge ncy \ndepartment discharge summary were observed in >75% of patients, leading to \nunnecessary NT-UCCT tests. \nLimitations: Differences in how clinical information is recorded , the variability \namong physicians, results not applicable to large h ospitals, small sample size \nlimited detailed analysis. \nFunding for this study: No \nEthics committee - additional information: The study was approved by the \nethics committee. \nAuthor Disclosures:  \nEstefania Corina Cotillo Ramos: Author: Author \nElena Otón González: Author: Author \nIsabel Santiago Suárez: Author: Author \nMarta Sánchez Canales: Author: Author \nJuana Plasencia Martínez: Advisory Board: Author \nGonzalo Pagán Vicente: Author: Author \nBelén Molina-Lozano: Author: Author \nAndrea Martínez López: Author: Author \nHerminia Ortiz Mayoral: Author: Author \n \n \n16:00-17:30 Research Stage 3 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1105 \nArtificial intelligence and planet radiology: \nthe green machine \n \nModerator \nR. Mirón Mombiela; Herlev/DK  \n(mirona@ufm.edu) \n \n \nHow do radiology department carbon footprints contr ibute to climate \nchange? \n*S. D. Jagadeesha*¹, R. Botchu²; ¹Mysuru/IN, ²Birmi ngham/UK \n(sushdj98@gmail.com) \n \nPurpose or Learning Objective: The purpose of this study is to investigate \npaper usage in the radiology department of a single  hospital institution over the \nlast three years to forecast paper usage up to 2050 . \n\n \n \nThursday \nAbstract-based Programme \n \n 127  \nMethods or Background: This retrospective study was performed in the \nradiology department of our tertiary orthopedic hos pital. The study included \nforms used for diagnostic and interventional proced ures in various \ndepartmental modalities. Diagnostic procedures requ ire one to three forms and \ninterventional procedures require three forms each.  Based on the established \nratio that 1.2 trees are cut for every 10,000 paper s used, the study calculated \nthe number of trees cut annually over the past thre e years and projected paper \nusage and tree loss until 2050 \nResults or Findings: Paper usage was distributed between diagnostic and \ninterventional procedures, with 67% used in diagnos tics and 33% in \ninterventions. The corresponding number of trees cu t during this period \namounted to 53.7 trees, with 47.4 trees for diagnos tic procedures and 6.4 trees \nfor interventional procedures. A total of 57.8 tree s for diagnostic procedures \nand 11.7 trees for interventional procedures were f orecasted to be cut annually \nfrom 2024 to 2050, cumulatively being 1227 trees by  the year 2050. \nConclusion: Our individual department had a significant contrib ution from \npaper usage in the carbon footprint of the departme nt. Adoption of digitalized \nappointment, prescribing and patient records is imp ortant in reducing this and \nachieving NHS net-zero targets. \nLimitations: The use of paper for forms, there are other signifi cant sources of \npaper consumption within the department. For exampl e, extensive paper \npackaging used for interventional consumables, and tissue paper used for \nvarious applications, such as covering ultrasound a nd CT couches, are also \ncontributing to the overall paper usage in the radi ology department. This has \nbeen excluded in the study. \nFunding for this study: Not applicable \nEthics committee - additional information: This has obtained ethical \ncommittee clearance from the Hospital. \nAuthor Disclosures:  \nRajesh Botchu: Nothing to disclose \nSushmitha Devihalli Jagadeesha: Nothing to disclose  \n \n \nAI-driven green gains: Enhancing efficiency with en vironmental benefits \nin Imaging \n*P. Strouhal*¹, N. Khan², A. Heathcote¹, M. Darwish ³, S. Persichini³, B. Miles¹, \nM. Trumann⁴, I. Farid³; ¹Warwick/UK, ²Dubai/AE, ³Chalfont St G iles/UK, \n⁴Freiburg/DE \n(pstrouhal@alliance.co.uk) \n \nPurpose or Learning Objective: Alliance Medical Ltd (AML) provides \ndiagnostic imaging for 800,000 NHS patients annuall y via networked facilities. \nGrowing concerns over operational and energy effici encies in 2022 prompted \nAML’s implementation of GE HealthCare’s Imaging360 solution. 18 months on, \nwe showcase how such Artifical Intelligence (AI) dr iven solutions are pivotal in \nrefining patient flows, scheduling, staffing, energ y usage and logistics \nmanagement within imaging services. \nMethods or Background: Integration comprised 6 separate Imaging360 \ncomponents utilising data from various sources, inc luding HL7, DICOM, \nBusiness Intelligence software and CSV extracts; mo bile and static CT and \nMRI scanners were incorporated from multiple sites across England (with PET-\nCT scanners now being onboarded). \nResults or Findings: Using predictive analytics, AML reduced missed \nappointments from 17% to 3% per week, improving res ource utilisation. \nOptimising protocols and schedules done on-cloud al lowed reduced senior \nstaff travel (approx. 380-480 km/month) and time (3 7.5hr/week) to manage \nscanner protocols; and significantly increased scan ner efficiency: - MRI: \nThroughput rose from 21 to 27 scans per day (+33%),  with kWh/patient \nreduced from 15.5 to 11.8. This saved 3.7 kWh per e xam —enough to power \n45 average households annually. One MRI site increa sed throughput by 43%, \nachieving 410 exams per month increase. - CT: scann er throughput improved \nby 256 scans per month average, cutting idle time; and reducing protocol \nvariability for CT chest, abdomen, pelvis from 47 t o 15 standardised protocols, \nwith related radiation doses lowered from 500 to 35 0 mGy.cm. Increased \nthroughput was achieved with no extra staff or equi pment. \nConclusion: Integrating AI into radiology workflows allows tran sformative \nchanges not only of operational efficiencies and co st savings, but improved \nsustainable practice. Going forward, further eco-fr iendly innovations could \nenhance both performance and sustainability across the healthcare imaging \nsector. \nLimitations: Imaging360 optimised for GE HealthCare scanners \nFunding for this study: GE HealthCare supporting implementation of AI \nplatform \nEthics committee - additional information: N/A \n \n \n \n \n \n \n \n \nAuthor Disclosures:  \nNaeem Khan: Employee: GE HealthCare \nPeter Strouhal: Board Member: Alliance Medical \nBrad Miles: Employee: Alliance Medical \nStefano Persichini: Employee: GE HealthCare \nMariam Darwish: Employee: GE HealthCare \nImran Farid: Employee: GE HealthCare \nMarkus Trumann: Employee: GE HealthCare \nAnn Heathcote: Employee: Alliance Medical \n \n \nBalancing Sustainability and Performance: Evaluatin g Energy Use, \nCarbon Footprint and Task Performance of Locally ru n Large Language \nModels for Radiology Report Simplification \n*A. Gupta*, R. Dheeka, R. Kumar, A. Rastogi, H. Mal hotra, K. Rangarajan; \nNew Delhi/IN \n(amit.aiims2014@gmail.com) \n \nPurpose or Learning Objective: To investigate tradeoffs between \nperformance and energy use when using different loc ally-run large language \nmodels (LLMs) and prompts for patient-centric simpl ification of radiology \nreports. \nMethods or Background: This study evaluated three different open-source \nLLMs (Meta’s Llama 3.1-8B, Microsoft’s Phi-3.5-Mini  and Mistral-7B) using five \ndifferent prompts to simplify 50 computed tomograph y report impressions, \ncollected from our tertiary-care oncology centre. M odels were run on a local \nworkstation with graphic processing unit. Energy us e (in watt-hours) and \ncarbon emissions (in grams) for each inference, wer e measured using an \nopen-source tool (CodeCarbon). Readability of origi nal and generated \nsimplified reports was quantitatively assessed usin g an average score of four \nreadability indices. LLM performance for simplifica tion task was measured as \ndifference in readability scores between original r eports and LLM-generated \nreports. Energy efficiency ratios (performance per watt-hour) and carbon \nfootprint (performance per gram of emissions) were calculated for each model-\nprompt combination. \nResults or Findings: Llama-prompt 5 (multi-shot learning) demonstrated t he \nhighest task performance (7.36), best energy effici ency ratio (31.89/Wh), and \nleast carbon footprint (44.70/g). Phi-prompt 5 achi eved high simplification \n(6.14) and energy efficiency (25.87/Wh). For Mistra l, prompt 1 (no context) was \noptimal (2.15/Wh and 3.01/g), but performance (1.16 ) lagged behind Llama \nand Phi. Friedman test revealed significant differe nces among readability \nscores (p < 0.001), with post-hoc Wilcoxon tests sh owing significant \nimprovements for Llama and Phi over the original an d Mistral, and Llama \noutperforming Phi (adjusted p < 0.0033). \nConclusion: Different LLM-prompt combinations showed variabilit y in energy \nuse, carbon emissions, and simplification task perf ormance. These results \nhighlight the importance of LLM-prompt combination selection for medical \napplications, balancing sustainability and performa nce. \nLimitations: Development of test prompts has inherent potential for \nsubjectivity. Apart from prompt engineering, we did  not use other accuracy \nimproving techniques like retrieval augmented gener ation. \nFunding for this study: None \nEthics committee - additional information: Study approved by the Institute \nEthics Committee All India Institute of Medical Sci ences, New Delhi (Ref. No. - \nIEC-343/15.06.2023) \nAuthor Disclosures:  \nKrithika Rangarajan: Nothing to disclose \nRohit Kumar: Nothing to disclose \nAshish Rastogi: Nothing to disclose \nAmit Gupta: Nothing to disclose \nHema Malhotra: Nothing to disclose \nRahul Dheeka: Nothing to disclose \n \n \nGreenhouse gas emissions due to long-term data stor age of reformatted \nCT series and strategies for mitigation \nY. Jia¹, M. Deng¹, *R. Burger*¹, S. L. Sheard¹, K. Hanneman²,  \nM. Drucker Iarovich², R. Illing¹, A. G. Rockall¹; ¹ London/UK, ²Toronto, ON/CA \n(r.burger@nhs.net) \n \nPurpose or Learning Objective: Image data storage and associated \ngreenhouse gas (GHG) emissions is accelerating, yet  strategies to minimise \nthis are limited. Reducing the average file size of  CT studies by reducing the \nnumber of reformats stored could help reduce emissi ons. This study aims to \nestimate GHG emissions associated with storage of C T reformats by modelling \nmeasurements from endometrial cancer baseline stagi ng CT. Secondary aims \nwere to model the findings comparing cloud storage emissions and assess the \nhypothetical GHG mitigation impact of a data retent ion policy \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 128  \nMethods or Background: Baseline staging CT chest, abdomen, and pelvis \n(CT-CAP) in 183 endometrial cancer patients in a UK  cancer centre between \n2013-2016 were analysed (Cohort A). The number of s tored multiplanar \nreformats, maximum intensity projections images and  lung reconstructions \nwere recorded. The file size of each reformat was n oted for 30 studies (Cohort \nB). Comparison was made with an external dataset of  100 baseline CT-CAP \nfrom Canada between 2018-2023 (Cohort C). Mitigatio n of GHG emissions \nwas projected for different storagescenarios. \nResults or Findings: Reformatted series were present in cohort A (97%, \n179/183), cohort B (97%, 29/30) and cohort C (100% ,100/100). Of the total file \nsize of cohort B (25590mb), 65% (16685mb) was refor mats and/or duplicate \nseries. On-premise storage of all reformats for cum ulative new UK endometrial \ncancer cases from 2020-2040 would produce 349 metri c tonnes CO2 \nequivalent (MTCO2e). Over 20 years, projected reduc tions in MTCO2e were \n69%(107/349) for storing only acquired axial slices , 80%(70/349) for switching \nto cloud storage, and 36%(222/349) for implemented a data retention policy. \nConclusion: A significant number of studies contained unnecessa ry reformats, \nincreasing average file size. A strategy to revise CT data storage protocols can \nsubstantially lower radiology GHG emissions, withou t compromising patient \ncare. \nLimitations: Small selective patient dataset. \nFunding for this study: None \nEthics committee - additional information: Institutional approval was \nobtained for a quality improvement project. \nAuthor Disclosures:  \nKate Hanneman: Nothing to disclose \nYifan Jia: Nothing to disclose \nMichael Deng: Nothing to disclose \nRebecca Burger: Nothing to disclose \nMoran Drucker Iarovich: Nothing to disclose \nRowland Illing: Nothing to disclose \nSarah Lucy Sheard: Nothing to disclose \nAndrea Grace Rockall: Nothing to disclose \n \n \nUltrasound's Hidden Environmental Cost: Linens and Disposables \nC. L. Thiel¹, J. Leschied², D. Carver², *J. R. Sche el*², R. Omary²,  \nM. Vigil-Garcia, Phd³, P. Gehrels³, C. Meijer³, O. Struk³; ¹Madison, WI/US, \n²Nashville, TN/US, ³Amsterdam/NL \n(john.scheel@vumc.org) \n \nPurpose or Learning Objective: To understand the sources of environmental \nimpact of ultrasound imaging in a US-based, adult d iagnostic radiology service. \nMethods or Background: A life cycle assessment (LCA) was used to evaluate \nthe environmental impact of 2 ultrasound machines a nd their surrounding \nresource needs, including production, use and dispo sal of other capital \nequipment, linens, disposable supplies, pharmaceuti cals, and data storage. A \nsensitivity analysis was performed to assess the im pact of low-carbon \nelectricity sources. \nResults or Findings: Contrary to expectations, linens and disposable sup plies \nemerged as the major contributors to ultrasound's g reenhouse gas (GHG) \nemissions, each accounting for approximately 30% of  its total impact. Energy \nuse from the ultrasound units themselves was compar atively lower, at 7%, \nalong with the production of the ultrasound units ( 7%), and the production and \nuse of workstations (11%). The study also noted tha t ultrasound equipment \nspent 30-45% of the time in non-scanning mode. The sensitivity analysis \nshowed the use of photovoltaics as an electricity s ource would reduce US’s \nGHGs by 9%; however, it would not shift the major s ources of GHGs away \nfrom linens and disposable supplies. \nConclusion: Reducing linen use, adopting reusable alternatives for disposable \nsupplies, and encouraging textile and supply manufa cturers and laundering \nfacilities to develop lower carbon alternatives are  essential for improving the \nsustainability of ultrasound practices. \nLimitations: The study, while providing valuable insights into t he \nenvironmental impact of US, has limitations due to its single-center focus; \nexclusion of mammography, nuclear medicine, and int erventional radiology; a \none-month data collection period; and some data and  modeling limitations. \nFunding for this study: No external funding was received for this study. \nPhilips and VUMC independently contributed to this study through in-kind \nlabor. \nEthics committee - additional information: This study was deemed non-\nhuman subjects research. \nAuthor Disclosures:  \nMarta Vigil-Garcia, Phd: Employee: Philips \nCecilia Meijer: Employee: Philips \nReed Omary: Nothing to disclose \nPien Gehrels: Employee: Philips \nJessiva Leschied: Nothing to disclose \nOlesya Struk: Employee: Philips \nDiana Carver: Nothing to disclose \nJohn R. Scheel: Nothing to disclose \nCassandra L Thiel: Consultant: Philips \nCT's Carbon Footprint: Energy and Disposables \nS. Pruthi¹, C. L. Thiel², D. Carver¹, J. R. Scheel¹ , R. Omary¹, M. Vigil-Garcia³, \nP. Gehrels³, C. Meijer³, *O. Struk*³; ¹Nashville, T N/US, ²Madison, WI/US, \n³Amsterdam/NL \n(olesya.struk@philips.com) \n \nPurpose or Learning Objective: To understand the sources of environmental \nimpact of CT scanning within a US-based, adult diag nostic radiology setting. \nMethods or Background: A life cycle assessment (LCA) was conducted, \nevaluating the production, use, and end-of-life of CT scanners, including \nenergy consumption, production and use of other cap ital equipment, \ndisposable supplies, linens, pharmaceuticals, and d ata storage. A sensitivity \nanalysis was performed assessing the impact of a lo w-carbon electricity \nsource. \nResults or Findings: Energy use and disposable supplies were identified as \nmajor contributors to CT's greenhouse gas (GHG) emi ssions, accounting for \n42% and 20%, respectively. The production of CTs co ntributed 17% to GHG \nemissions. Furthermore, the study revealed a 50% di fference in GHG \nemissions between CT scanners of the same model, su ggesting opportunities \nfor optimization. CT scanners were also found to sp end between 44-72% of the \ntime in a non-scanning mode. Sensitivity analysis s howed that using low-\ncarbon electricity could significantly decrease CT' s energy emissions, shifting \nthe major sources of emissions to the production of  CT imaging equipment and \ndisposable supplies. \nConclusion: Optimizing energy use, minimizing disposable suppli es, and \nensuring efficient equipment utilization are crucia l for reducing CT's \nenvironmental impact. \nLimitations: The study, while providing valuable insights into t he \nenvironmental impact of CT, has limitations due to its single-center focus; \nexclusion of mammography, nuclear medicine, and int erventional radiology; a \none-month data collection period; and some data and  modeling limitations. \nFunding for this study: No external funding was received for this study. \nPhilips and VUMC independently contributed to this study through in-kind \nlabor. \nEthics committee - additional information: This study was deemed non-\nhuman subjects research. \nAuthor Disclosures:  \nSumit Pruthi: Nothing to disclose \nMarta Vigil-Garcia: Employee: Philips \nCecilia Meijer: Employee: Philips \nReed Omary: Nothing to disclose \nPien Gehrels: Employee: Philips \nOlesya Struk: Employee: Philips \nDiana Carver: Nothing to disclose \nJohn R. Scheel: Nothing to disclose \nCassandra L Thiel: Consultant: Philips \n \n \nThe Environmental Cost of MRI: A Life Cycle Assessm ent \nD. Carver¹, C. L. Thiel², J. R. Scheel¹, *R. Omary* ¹, M. Vigil-Garcia³,  \nP. Gehrels³, S. Thornander³, C. Meijer³, O. Struk³;  ¹Nashville, TN/US, \n²Madison, WI/US, ³Amsterdam/NL \n(reed.omary@vumc.org) \n \nPurpose or Learning Objective: To understand the sources of environmental \nimpact of MRI within a US based diagnostic radiolog y department. \nMethods or Background: A life cycle assessment (LCA) was conducted that \nevaluated the production, use, and end-of-life of 3  MRI scanners in an adult \ndiagnostic radiology department. Other model inputs  included the production \nand energy use of other capital equipment, disposab le supplies, linens, \npharmaceuticals, and data storage. A sensitivity an alysis assessed the impact \nof using a low-carbon electricity source. \nResults or Findings: Energy consumption emerged as the dominant source o f \nMRI's greenhouse gas (GHG) emissions, representing 79% of its total impact. \nNotably, the 3T MRI demonstrated 1.4 times higher e nergy use and 1.9 times \nhigher production emissions compared to the 1.5T. A dditionally, MRI scanners \nwere found to be in low-power or ready-to-scan mode  for 72-75% of the time, \nindicating potential for energy optimization. Sensi tivity analysis revealed that \ndecarbonizing the electricity grid could lead to an  87% reduction in energy-\nrelated GHG emissions from MRI. In this scenario, t he production of imaging \nequipment itself would become the largest contribut or to MRI's GHG \nemissions. \nConclusion: Improving energy efficiency through measures such a s optimizing \nscan protocols, developing automation of scanner ef ficiency modes, and \ntransitioning to renewable energy sources are cruci al steps in reducing MRI's \nenvironmental footprint. If changing the grid is no t possible, other opportunities \ninclude reducing scan times via AI (e.g. Smart spee d) or optimized scheduling. \nLimitations: The study, while providing valuable insights into t he \nenvironmental impact of MRI, has limitations due to  its: single-center focus; \nexclusion of mammography, nuclear medicine, and int erventional radiology; \none-month data collection period; and some data and  modeling limitations. \n\n \n \nThursday \nAbstract-based Programme \n \n 129  \nShifting to a low-carbon electricity grid highlight s the additional need to address \nemissions associated with the production of MRI equ ipment itself. \nFunding for this study: No external funding was received for this study. \nPhilips and VUMC independently contributed to this study through in-kind \nlabor. \nEthics committee - additional information: This study was deemed non-\nhuman subjects research. \nAuthor Disclosures:  \nMarta Vigil-Garcia: Employee: Philips \nCecilia Meijer: Employee: Philips \nReed Omary: Nothing to disclose \nPien Gehrels: Employee: Philips \nOlesya Struk: Employee: Philips \nDiana Carver: Nothing to disclose \nJohn R. Scheel: Nothing to disclose \nCassandra L. Thiel: Consultant: Philips \nSophie Thornander: Employee: Philips \n \n \nAI-Powered MRI: Time, Energy, and Emission Savings for a Greener \nFuture \n*T. Polidori*, M. Zerunian, D. De Santis, F. Puccia relli, B. Masci, A. Del Gaudio, \nF. Fanelli, D. Caruso, A. Laghi; Rome/IT \n(tiziano.polidori13@gmail.com) \n \nPurpose or Learning Objective: The study aimed to assess the energy and \ngreenhouse-gas (GHG) emission savings feasible usin g artificial intelligence \n(AI) in multi-district MRI-protocols, including lum bar-spine-MRI, cardiac-MRI, \nand upper abdomen-MRI. We evaluated the impact of A I-algorithms applied on \nMRI acquisition on scan time reduction, energy cons umption, and CO2 \nemissions per patient, providing insights into the potential benefits of AI in \nroutine clinical practice. \nMethods or Background: This retrospective study analyzed 148 patients, \nincluding 45 upper abdomen-MRI, 53 cardiac-MRI, and  50 lumbar spine-\nMRI.MRI scans were acquired both without and with A I assistance applied to \nspecific 2D and 3D sequences.The Air Recon-DL (GE H ealthcare) was used \nfor T2 and DWI sequences in upper abdomen-MRI, as w ell as T1, T2, and \nSTIR sequences in lumbar spine-MRI. The Sonic-DL (G E Healthcare) was \napplied to SSFP sequences specifically for cardiac- MRI.The outcomes \nmeasured were time savings per patient, reduced ene rgy consumption (kW/h), \nand the corresponding reduction in CO2-equivalent e missions. \nResults or Findings: The application of AI across all three districts st udied \nresulted in significant time savings per patient co mpared to non-AI protocol \n(p<0.01): 5’11’’ (58%) for upper abdomen-MRI, 1’30’ ’ (52%) for cardiac-MRI, \nand 6’ (50%) for lumbar spine-MRI.These time reduct ions corresponded to \nsignificant energy savings of 1.39kW/h, 0.40kW/h, a nd 1.68kW/h per patient \n(p<0.05), respectively. The equivalent reduction in  CO2 emissions was 0.57kg \nfor upper abdomen-MRI, 0.16kg for cardiac-MRI, and0 .69 kg for lumbar spine-\nMRI (p<0.05). \nConclusion: The implementation of AI in MRI protocols significa ntly reduces \nscan time, energy consumption, and GHG emissions, h ighlighting its potential \nfor enhancing the sustainability of medical imaging  practices.Integrating AI into \nroutine clinical protocols could offer considerable  environmental benefits, \ncontributing to the reduction of the healthcare sec tor’s carbon footprint. \nLimitations: Limitations include a small patient cohort and the use of a single \nvendor for MRI-protocols. \nFunding for this study: No-funding was received for this study. \nEthics committee - additional information: No \nAuthor Disclosures:  \nDamiano Caruso: Nothing to disclose \nBenedetta Masci: Nothing to disclose \nFrancesco Pucciarelli: Nothing to disclose \nMarta Zerunian: Nothing to disclose \nFederica Fanelli: Nothing to disclose \nDomenico De Santis: Nothing to disclose \nTiziano Polidori: Nothing to disclose \nAndrea Laghi: Nothing to disclose \nAntonella Del Gaudio: Nothing to disclose \n \n \nAutomated scout-image based estimation of contrast agent dosing: a \ndeep learning approach \n*R. T. Schirrmeister*, P. S. Friemel, M. Reisert, F . Bamberg, J. Weiß, A. Rau; \nFreiburg/DE \n(robin.schirrmeister@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: To develop and test a deep learning \nalgorithm for approximation of contrast agent dosag e based on CT scout \nimages. \n \n \nMethods or Background: We prospectively enrolled 817 patients undergoing \nclinically indicated CT imaging, predominantly of t he chest and/or abdomen. \nPatient weight was collected 1) manually and 2) sel f-reported prior to the \nexamination by study staff. Based on the scout imag es, we developed an \nEfficientNet convolutional neural network pipeline to estimate the optimal \ncontrast agent dose based on patient weight and pro vide a browser-based \nuser interface as a versatile open-source tool to a ccount for different contrast \nagent compounds We additionally analyzed the body-w eight-informative CT \nfeatures using a weight-conditional variational aut oencoder. \nResults or Findings: The training cohort consisted of 218 chest, 51 \nabdominal, 511 whole-body, and 37 CT scans of vario us other anatomical \nregions. Self-reported patient weight was statistic ally significantly lower than \nmanual measurements (75.02 kg vs.76.92 kg; p < 10 ⁻⁵, Wilcoxon signed-rank \ntest). Our pipeline predicted patient weight with a  mean absolute error of 4.74 ± \n0.14 kg in 5-fold cross-validation and is publicly available at https://nora-\nimaging.org/ct-scout-weight/. Interpretability anal ysis revealed that both larger \nanatomical shape and higher overall Hounsfield unit s were predictive of body \nweight. \nConclusion: Our open-source deep learning pipeline allows for a utomatic \nestimation of accurate contrast agent dosing based on scout images in routine \nCT imaging studies. This approach has the potential  to streamline contrast \nagent dosing workflows, improve efficiency, and enh ance patient safety by \nproviding quick and accurate weight estimates witho ut additional \nmeasurements or reliance on potentially outdated re cords. \nLimitations: The model's performance may vary depending on patie nt \npositioning and scout image quality and the approac h requires validation on \nlarger patient cohorts and other clinical centers. \nFunding for this study: Funded by an unrestricted research grant from \nSiemens Healthineers. \nEthics committee - additional information: The study was approved by an \nethics committee. Written informed consent was obta ined from each \nparticipant. \nAuthor Disclosures:  \nPaul Simeon Friemel: Nothing to disclose \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nAlexander Rau: Advisory Board: Bayer \nRobin Tibor Schirrmeister: Nothing to disclose \nFabian Bamberg: Nothing to disclose \n \n \nCross-Modality Image Conversion from non-contrast C ardiac Magnetic \nResonance to contrast-enhanced Computed Tomography Angiography \nusing Diffusion Models \n*E. Almar Munoz*, C. G. Colintenorio, C. Kremser, M . Haltmeier, A. Mayr; \nInnsbruck/AT \n(enrique.almar@i-med.ac.at) \n \nPurpose or Learning Objective: Transcatheter Aortic Valve Implantation \n(TAVI) is the preferred treatment for patients with  severe aortic stenosis at high \nto intermediate surgical risk. The gold-standard pr eoperative imaging modality \nis contrast-enhanced CTA; however, non-contrast CMR  is an alternative for \npatients with contraindications to contrast agents despite its limitations in \ndetecting calcifications. We propose diffusion mode ls to improve CMR-to-CTA \nconversion, facilitating comprehensive TAVI plannin g and predicting valve \ncalcifications without contrast. \nMethods or Background: Our pipeline integrates Denoising Diffusion \nProbabilistic Models (DDPMs) and Stochastic Differe ntial Equation (SDE) \nmodels. This pipeline was evaluated using an in-hou se dataset consisting of 39 \npaired CTA and CMR scans. The image pairs were alig ned using rigid \nregistration techniques. To improve the registratio n process, we utilized aorta \nsegmentation masks obtained using nnUNet for CMR sc ans and \nTotalSegmentator for CTA scans. \nResults or Findings: Regarding the aorta segmentation, we obtained Dice \nvalues of 0.987±0.006 for CMR and 0.980±0.005 for CTA. The Dice Score \nobtained in the rigid registration was above 0.87. Regarding the image \nconversion, our results demonstrate that the overal l synthesized CTA images \nexhibit high fidelity to their real counterparts, v alidated by metrics including the \nStructural Similarity Index Measure (SSIM) and Peak  Signal-to-Noise Ratio \n(PSNR), both exceeding 0.80 and 22, respectively. F ocusing on the valve's \ncalcifications, some are accurately converted into CTA-calcified regions but are \nnot always consistent or repeatable. \nConclusion: This study highlights the potential of diffusion mo dels in medical \nimaging, offering a promising solution for patients  unable to receive contrast \nagents, thereby improving the safety and efficacy o f TAVI planning. \nLimitations: Firstly, the model encounters difficulties in repli cating small \ndetails in the CTA, including calcifications. Secon dly, the diffusion models \napplied are very sensitive to image training; both data modalities must present \nlow noise levels or artifacts. \nFunding for this study: Fund provided by FWF-DOC-110 \nEthics committee - additional information: Nothing to declare \n \n\n \n \nThursday \nAbstract-based Programme \n \n 130  \nAuthor Disclosures:  \nMarkus Haltmeier: Nothing to disclose \nEnrique Almar Munoz: Nothing to disclose \nAgnes Mayr: Nothing to disclose \nChristian Kremser: Nothing to disclose \nCarmen Guadalupe Colintenorio: Nothing to disclose \n \n \nDiffusion Model for Non-contrast MR to Aid Diagnosi s of Focal Liver \nLesions: A Multi-Center Study \n*S. Dong*, Z. Shen, F. Yan, R. Li; Shanghai/CN \n(sjdong@sjtu.edu.cn) \n \nPurpose or Learning Objective: To develop a diffusion model for generating \nvirtual dynamic contrast-enhanced MRI (DCE-MRI) ima ges from non-contrast \nT1-weighted scans and assess its efficacy in FLL di agnosis. \nMethods or Background: Gadolinium-based contrast agents (GBCAs) in \nDCE-MRI are crucial for characterizing focal liver lesions (FLLs), but their use \nincreases risks for patients with renal impairment and adds to imaging costs. \nVirtual contrast-enhanced images from non-contrast T1-weighted scans could \nreduce these risks and streamline diagnostics. FLLs  ≥1 cm, identified through \nDCE-MRI, were included, with lesion types such as H CC, ICC, liver \nmetastases, cysts, hemangiomas, and FNH. A diffusio n model was trained on \nnon-contrast T1-weighted and corresponding multipha se DCE-MRI images \n(arterial, portal venous, and delayed phases). Trai ning occurred at Center 1 \n(2018–2023) with a 3:1 split for training and inter nal testing. External validation \nused data from three other centers (2018–2024). A d iagnostic model for FLLs \nwas also trained on synthetic DCE-MRI images. Norma lized mean absolute \nerror (NMAE), peak signal-to-noise ratio (PSNR), an d structural similarity index \nmeasure (SSIM) were used for evaluation. Three radi ologists scored image \nquality on a three-point scale. The human machine c omparison was conducted \nwith six radiologists in different experience. \nResults or Findings: The study included 1187 patients in the training se t \n(mean age, 51 ±12), with 395 internal (52 ±16) and 347, 271, and 219 external \npatients (57 ±11, 56 ±12, 58 ±11). The model showed strong similarity between \nvirtual and real DCE-MRI images, with NMAE 0.021–0. 038, PSNR 28.9–31.8 \ndB, and SSIM 0.881–0.927. Diagnostic accuracy was 9 3% for the internal and \n91% for external sets, outperforming three junior r adiologists (P < .001) and \nmatching three senior radiologists (P = .19). \nConclusion: The diffusion model provides a safe, cost-effective  alternative to \ntraditional DCE-MRI, maintaining high diagnostic ac curacy for FLLs. \nLimitations: None reported. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nRuokun Li: Nothing to disclose \nShunjie Dong: Nothing to disclose  \nFuhua Yan: Nothing to disclose \nZhehan Shen: Nothing to disclose \n \n \n16:00-17:30 Research Stage 4 \nResearch Presentation Session: Neuro \nRPS 1111 \nNeuro interventions and beyond \n \nModerator \nD. Ozretić; Zagreb/HR  \n(david.ozretic@ck.t-com.hr) \n \n \nCerebral metabolic rate of oxygen on admission MRI may predict infarct \ngrowth in hyperacute ischemic stroke patients treat ed successfully with \nthrombectomy: a retrospective observational study \n*A. Bani Sadr*, J. Fournel, M. Hermier, N. Nighogho ssian, Y. Berthezene; \nLyon/FR \n(apbanisadr@gmail.com) \n \nPurpose or Learning Objective: Despite successful thrombectomy, most \nstroke patients experience infarct growth which neg atively affect functional \noutcomes. Advances in oxygen metabolism mapping on admission dynamic-\nsusceptibility contrast MRI have shown promise in a ssessing the viability of \ndiffusion-weighted imaging (DWI) lesions. We aimed to assess the utility of \noxygen metabolism mapping at the voxel scale in det ermining the fate of \ndiffusion-mismatch regions following successful thr ombectomy. \nMethods or Background: We conducted a retrospective analysis of the \nHIBISCUS-STROKE cohort (NCT: 03149705), a single-ce nter, observational \nstudy enrolling patients treated with thrombectomy between 2016 and 2022. \nAdmission DSC-MRI was used to generate time-to-maxi mum (Tmax), cerebral \nblood volume (CBV), cerebral metabolic rate of oxyg en (CMRO2), and oxygen \nextraction fraction (OEF) maps. In patients with su ccessful reperfusion \n(modified Thrombolysis in Cerebral Infarction [mTIC I] score ≥2B), Tmax ≥6s \nvoxels excluding DWI abnormalities were analyzed on  day 6 T2-fluid \nattenuated inversion recovery MRI. Semi-quantitativ e measurements of \nCMRO2, CBV, and OEF were extracted from regions ide ntified as either \nnecrotic or salvaged at follow-up. \nResults or Findings: Among the 321 patients enrolled, 134 (41.7%) met \ninclusion criteria (median age 71.0 years; 58.2% ma le; median NIHSS score \n15.0). In the training cohort, ROC analysis identif ied optimal thresholds for \npredicting necrosis: CBV (0.99), CMRO2 (0.64), and OEF (1.59). In the \nvalidation cohort, CMRO2 achieved an area under the  curve (AUC) of 73.6% \n(95% confidence interval [CI]: 65.0–82.1), signific antly outperforming CBV \n(AUC: 63.5%, 95% CI: 53.8–73.1; P=0.003) and OEF (A UC: 55.0%, 95% CI: \n45.2–64.7; P=0.0005). Multivariable logistic regres sion revealed that CMRO2 \n<0.64 was independently associated with necrosis in  diffusion-perfusion \nmismatch regions (OR: 6.0, 95% CI: 3.2–11.6, P<0.00 01). \nConclusion: In acute stroke patients achieving successful throm bectomy, a \nCMRO2 < 0.64 in regions of diffusion-perfusion mism atch. \nLimitations: DSC-MRI dervied oxygen metabolism mapping is not ye t \nvalidated against PET \nFunding for this study: This work was supported by the RHU MARVELOUS \n(ANR-16-RHUS-0009) of Université de Lyon, within th e program \n“Investissements d'Avenir” operated by the French N ational Research Agency. \nEthics committee - additional information: The local ethics committee \n(Institutional Review Board No: 00009118) approved this study, and all \nparticipants or their relatives gave provided infor med consent. \nAuthor Disclosures:  \nYves Berthezene: Nothing to disclose \nMarc Hermier: Nothing to disclose \nNorbert Nighoghossian: Nothing to disclose \nJulien Fournel: Nothing to disclose \nAlexandre Bani Sadr: Nothing to disclose \n \n \nWhy we fail: Factors leading to unsuccessful mechan ical \nthrombectomies \n*R. Bruen*¹, H. Briody¹, S. Singh¹, N. Healy², M. T . Crockett¹, J. Müller¹,  \nS. O' Reilly¹, J. Thornton¹, P. Nicholson¹; ¹Dublin  9/IE, ²Dublin/IE \n(richardbruen93@gmail.com) \n \nPurpose or Learning Objective: Mechanical thrombectomy (MT) has \nrevolutionized acute ischemic stroke care, demonstr ating superior functional \noutcomes compared to intravenous t-PA alone. A subs tantial minority of \npatients fail to achieve successful recanalization.  This study analyses the \ncauses of unsuccessful MT in a large tertiary refer ral centre over an 11-year \nperiod. \nMethods or Background: We retrospectively reviewed a prospective stroke \nregistry at our institution, identifying all patien ts with acute ischemic stroke who \nunderwent MT between January 2012-2023. Unsuccessfu l MT was defined as \na post-interventional modified Thrombolysis in Cere bral Infarction (mTICI) \nscore < 2b. We collected demographic data, NIHSS, A SPECTS, occlusion \nlocation, pre-MT alteplase administration and speci fic reasons for MT failure, \ncategorizing them into three groups: Target not rea ched, Target reached but \nrecanalization failed, and non-technical reasons. \nResults or Findings: Of 2620 MT procedures performed, 259 (9.9%) were \nunsuccessful. The median patient age was 75 years ( IQR 65-83), 48% (n=124) \nwere female and the median NIHSS score was 16 (IQR 10-19). Occlusions \nwere located in the anterior circulation in 98% of cases, with a median \nASPECTS of 9. In Category 1 (n=48, 19%), non-reachi ng of the target was \nattributed to cervical artery tortuosity (n=23, 48% ), challenging aortic arch \nanatomy (n=7, 14.6%) and inability to traverse a ce rvical occlusion (n=18, \n37.5%). Category 2 failures were the most common (n =146, 56%) and were \ncaused by unsuccessful microcatheter advancement be yond the occlusion \n(n=8, 5.4%), stent retriever and aspiration cathete r failed recanalization (n=39, \n26.7%), and spontaneous/iatrogenic re-occlusion (n= 99, 67.8%). Category 3 \n(non-technical failures; n=62, 24%) were less commo n and were mainly due to \npatient neurological decline. \nConclusion: Unsuccessful mechanical thrombectomy was encountere d in \n9.9% of cases in our cohort. The most common reason  was \nspontaneous/iatrogenic re-occlusion. \nLimitations: Study was performed in a single site tertiary refer ral centre. \nFunding for this study: None \nEthics committee - additional information: Retrospective data. No ethics \nrequired. \n \n \n \n\n \n \nThursday \nAbstract-based Programme \n \n 131  \nAuthor Disclosures:  \nPatrick Nicholson: Nothing to disclose \nRichard Bruen: Nothing to disclose \nJohn Thornton: Nothing to disclose \nHayley Briody: Nothing to disclose \nMatthew Thomas Crockett: Nothing to disclose \nSean O' Reilly: Nothing to disclose \nNuala Healy: Nothing to disclose \nSneha Singh: Nothing to disclose \nJennifer Müller: Nothing to disclose \n \n \nComparison of Endovascular Thrombectomy Outcomes Be tween In-\nHospital and Out-of-Hospital Stroke \n*S. Singh*, P. Rohan, C. Leneghan, R. Bruen, M. T. Crockett, A. O'Hare,  \nS. Power, J. Thornton, P. Nicholson; Dublin/IE \n(snehasingh2412@gmail.com) \n \nPurpose or Learning Objective: To compare patient characteristics and \noutcomes following endovascular thrombectomy (EVT) for ischemic stroke \nbetween patients experiencing in-hospital stroke (I HS) and out-of-hospital \nstroke (OHS). \nMethods or Background: A single-center, retrospective observational cohort  \nstudy was conducted using data from the institution al EVT database at \nBeaumont Hospital. Patients were categorized as IHS  or OHS, and their \nbaseline characteristics and outcomes were compared . \nResults or Findings: Of 2619 patients undergoing EVT, 383 (14.6%) \nexperienced IHS (median age 72 years, 57% male) and  2235 (85.4%) \nexperienced OHS (median age 72 years, 54% male). OH S patients had higher \npre-stroke modified Rankin Scale (mRS) scores (p<0. 0001) and were \nsignificantly more likely to receive intravenous th rombolysis (p<0.0001). IHS \npatients had higher median National Institutes of H ealth Stroke Scale (NIHSS) \nscores at day 1 (11 vs. 8, p=0.14) and day 5 (6 vs.  4, p=0.0586), although \nthese differences were not statistically significan t. IHS patients also had higher \nmRS scores at day 30 (p<0.0001) and day 90 (p<0.000 1). OHS patients \ntended to have better Thrombolysis in Cerebral Infa rction (TICI) scores post-\nrevascularization (p=0.0535). Treatment was faster in the IHS group: onset to \ngroin puncture (210 vs. 308 minutes, p<0.0001), ons et to first reperfusion (243 \nvs. 341 minutes, p<0.0001). There was no significan t difference in groin \npuncture to first reperfusion time (23 vs. 21 minut es, p=0.1367). IHS patients \nhad a longer median EVT procedure time (37 vs. 32 m inutes, p=0.0363). \nConclusion: Despite shorter time intervals to intervention, pat ients with IHS \nexperienced worse functional outcomes after EVT com pared to patients with \nOHS. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: Approved by the local ethics \ncommittee at Beaumont Hospital, Dublin, Ireland \nAuthor Disclosures:  \nPatrick Nicholson: Nothing to disclose \nRichard Bruen: Nothing to disclose \nJohn Thornton: Nothing to disclose \nAlan O'Hare: Nothing to disclose \nMatthew Thomas Crockett: Nothing to disclose \nSarah Power: Nothing to disclose \nCaoimhe Leneghan: Nothing to disclose \nSneha Singh: Nothing to disclose \nPat Rohan: Nothing to disclose \n \n \nHow predictive is CT angiography source image ASPEC TS (CTA-SI \nASPECTS) score on initial CT for futile mechanical thrombectomy?  \nAn ongoing study \n*A. Tsaoulia*, M. Mantatzis, L. Kougias, A. Stofori adi, P. K. Prassopoulos; \nThessaloniki/GR \n(kat_tsaoulia@yahoo.gr) \n \nPurpose or Learning Objective: The standard of care for patients with large \nvessel occlusion (LVO) is mechanical thrombectomy, with or without preceding \nthrombolysis. Patients with a large infarction core  have a higher risk of \nreperfusion edema or hemorrhage. Alberta stroke pro gram early computed \ntomography (ASPECTS) score is widely used to evalua te the extent of acute \nischemic stroke at the middle cerebral artery terri tory. An extention of it is CTA-\nSI ASPECTS (CTAsp), which shows the collateral circ ulation and discriminates \nhypoperfused areas. However, it is not established whether this hypoperfusion \ndepicts core or penumbra and if CTAsp is more accur ate for patient selection. \nWe aim to find a possible correlation between CTAsp  and infarct core \nMethods or Background: We analyzed the initial CT/CTAs of seven patients \nwith LVO that had complete first pass recanalizatio n and the CTAsp scores \nwere calculated. All CT/CTAs were performed on a 16 -slice SIEMENS \nEmotion-16. A non-contrast CT scan was obtained aft er thrombectomy, 24 and \n48 hours later. These scans were performed on a 128 -slice GE OPTIMA. Ten \nROIS were applied in each hemisphere in areas defin ed by ASPECTS, to find \ndensity differences. \nResults or Findings: The patients were divided into two groups based on \nCTAsp scores; 3 patients had a score of 6-7 and 4 w ith a score of 8-10. In two \npatients a discrepancy between ASPECTS and CTAsp wa s found. The final \ninfarct core was associated with CTAsp in 6/7 patie nts, while in one, the \nhypodense area in CTAsp was normalized in follow up . \nConclusion: CTAsp proves to be helpful for attempting to predic t which \npatients will have a good outcome after endovascula r treatment. \nLimitations: Small number of cases-preliminary results \nFunding for this study: None \nEthics committee - additional information: The study is based on an \nimaging score from emergency obtained CT scans. \nAuthor Disclosures:  \nMichalis Mantatzis: Nothing to disclose \nPanos K. Prassopoulos: Nothing to disclose \nAikaterini Tsaoulia: Nothing to disclose \nAnatoli Stoforiadi: Nothing to disclose \nLeonidas Kougias: Nothing to disclose \n \n \nComparison of antithrombogenic coated and uncoated flow-diverters in \nruptured and unruptured cerebral aneurysms \n*D. Weiß*, M. Vach, V. L. Ivan, S. Muhammad, B. Hof mann, M. Neyazi,  \nB. Turowski, M. Kaschner; Düsseldorf/DE \n(DanielArvid.Weiss@med.uni-duesseldorf.de) \n \nPurpose or Learning Objective: Flow-diversion has become a key treatment \noption for complex intracranial aneurysms. Recent a dvancements include \ncoated flow-diverters (FD), designed to potentially  reduce the need for dual \nantiplatelet therapy thereby removing the associate d secondary risks, while \nmaintaining patency and low complication rates. Com paring coated and \nuncoated FDs may offer insights into long-term outc omes and treatment \noptimization. \nMethods or Background: In this retrospective single-center study, we \ninvestigated the data of 21 consecutive patients wi th cerebral aneurysms, \ntreated between 2021 and 2023 with the coated Deriv o 2heal Embolization \nDevice (D2H) and the uncoated Derivo Embolization D evice (DED) (both \nAcandis, Pforzheim, Germany). We described the proc edure and analyzed \nclinical and radiological data, along with long-ter m outcomes after 18 months of \nfollow-up. \nResults or Findings: Nine patients (42.9%) had incidental, while 12 (57. 1%) \nhad symptomatic aneurysms, including ten with WFNS IV subarachnoid \nhemorrhages. Aneurysm locations included mostly the  internal carotid (n=9) \nand the vertebral artery (n=7). All FDs were succes sfully deployed: 11 patients \nreceived the coated device and 10 the uncoated devi ce. After 18 months, \n73.3% patients had favorable outcomes (mRS 0-2). On e coated FD occluded \nasymptomatically after six months, and one uncoated  FD occluded \nimmediately but could be recanalized. \nConclusion: We observed favorable occlusion rates for both coat ed and \nuncoated FDs. The role of dual antiplatelet therapy  remains debated. Large \nmulticenter studies are essential to evaluate the p atency of coated compared \nto uncoated FDs and determine whether they can redu ce thrombogenicity, \npotentially allowing for less or no antiplatelet th erapy in emergencies. \nLimitations: Unequal distribution of emergency and elective trea tments. \nLimited by both the number of patients. The study d oes not address whether \nmonotherapy would be sufficient for drug-coated ste nts. \nFunding for this study: None \nEthics committee - additional information: Local ethics committee of \nmedical faculty \nAuthor Disclosures:  \nMarius Kaschner: Nothing to disclose \nMarius Vach: Nothing to disclose \nMilad Neyazi: Nothing to disclose \nSajjad Muhammad: Nothing to disclose \nDaniel Weiß: Nothing to disclose \nBjörn Hofmann: Nothing to disclose \nVivien Lorena Ivan: Nothing to disclose \nBernd Turowski: Nothing to disclose \n \n \nEvaluation the Effect of CT Black Blood Technique i n Post-treatment \nFollow-up of Intracranial Aneurysms Treated with Fl ow-diverting Stents \n*D. Xie*, Z. Lai, H. Ma, R. Xu, J. Wu, J. Zhao; Gua ngzhou/CN \n(xiedx7@mail.sysu.edu.cn) \n \nPurpose or Learning Objective: To assess the feasibility and effect of CT \nblack blood technique in the post-treatment follow- up of intracranial aneurysms \ntreated with flow-diverting stents. \nMethods or Background: A retrospective analysis was conducted on 18 \npatients who underwent treatment with flow-divertin g stents for intracranial \naneurysms and subsequently underwent digital subtra ction angiography (DSA) \n\n \n \nThursday \nAbstract-based Programme \n \n 132  \nfollow-up. All patients underwent CT angiography (C TA) examination before \nDSA. The contrast-enhancement boost technique was e mployed to process \nCTA images to obtain CT black blood images. Two rad iologists independently \nrated the image quality and diagnostic confidence f or stent-related stenosis on \nboth conventional CTA and CT black blood images usi ng a 4-point scale. The \nperformance of conventional CTA and CT black blood in diagnosing stent-\nrelated stenosis was compared, and the examination time and radiation dose \nof CTA and DSA were recorded and compared. \nResults or Findings: Subjective ratings of image quality and diagnostic \nconfidence for stent-related stenosis were signific antly higher for CT black \nblood images compared to conventional CTA images (3 .94±0.23 vs. 3.06±0.62 \nand 3.89±0.31 vs. 2.83±0.6, respectively; all p＜0.01). DSA detected 2 cases \nof distal stenosis and 5 cases of overall stenosis within the stent. Compared to \nconventional CTA (sensitivity: 100%, specificity: 3 6.4%, accuracy: 61.1%), CT \nblack blood demonstrated significantly improved per formance in detecting \nstent-related stenosis, with diagnostic sensitivity , specificity, and accuracy all \nreaching 100%, and perfect inter-observer agreement  (k=1.0). Regarding \nradiation dose, the average radiation dose for conv entional CTA was \n(67.85±8.31) mGy, whereas the radiation dose required for DSA significantly \nincreased to (516.81±193.83) mGy (p＜0.001). The examination times for CT \nand DSA were (3.53±0.74) minutes and (9.57±8.26) minutes, respectively(p＜\n0.01). \nConclusion: The CT black blood technique shows potential as the  preferred \nmethod for post-treatment follow-up of intracranial  aneurysms treated with flow-\ndiverting stents. \nLimitations: Sample size was relatively small. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nJiale Wu: Author: author \nDingxiang Xie: Nothing to disclose  \nRulin Xu: Nothing to disclose \nZhiman Lai: Nothing to disclose \nHui Ma: Nothing to disclose \nJing Zhao: Author: author \n \n \nImpact of Automatically Assessed Collateral Circula tion and Infarct Core \non Functional Outcome in Acute Ischemic Stroke Pati ents treated with \nEndovascular Thrombectomy \n*I. Požar*¹, F. F. Bajrovi ć², L. Umek², K. Šurlan Popović²; ¹Izola/SI, \n²Ljubljana/SI \n \nPurpose or Learning Objective: This study aimed to evaluate the predictive \nvalue of automatically assessed collateral circulat ion (CC) and infarct core for \nfunctional outcome in acute ischemic stroke (AIS) p atients treated with \nendovascular thrombectomy (EVT). \nMethods or Background: We conducted a retrospective cohort study of 208 \npatients with anterior large vessel occlusion treat ed with EVT. Two AI-powered \nsoftware were used to automatically assess CC and i nfarct core. Comparative \nanalyses included patient demographics, clinical an d imaging data, and \nfunctional outcome. Univariate and multivariable lo gistic regression analyses \nwere conducted to predict the 90-day functional out come. A favorable outcome \nwas defined as a modified Rankin scale (mRS) score ≤2. \nResults or Findings: Among the 208 patients, 114 (54.8%) were women and \n94 were men, with a mean age of 71.4±13.3 years. Patients with higher \ncollateral score (CS) exhibited lower infarct core volumes (p<0.001) and better \nmRS score at 90 days (p=0.008). Among patients with  a favorable outcome, \nthe mean infarct core volume was lower compared to those with poor \noutcomes (5 mL vs. 8.6 mL, p=0.003). In univariate logistic regression, both \ninfarct core (OR 0.94, p=0.005) and CC (OR 1.84, p= 0.014) were predictors of \nfavorable outcome. However, in multivariable models , only infarct core \nremained a significant independent predictor [AORs of 0.95 (p=0.021) and 0.96 \n(p=0.039)]. \nConclusion: Automatically assessed infarct core is a robust pre dictor of \nfunctional outcome in AIS patients post-EVT, while CC's predictive value \ndiminishes when adjusted for infarct core. These fi ndings support the \nintegration of AI-powered evaluations in clinical s ettings to improve prognosis \nand treatment strategies for AIS. \nLimitations: Our study's limitations include using a 40-slice CT  scanner, which \naffects data acquisition speed, particularly in per fusion; non-blinded \nassessments potentially introducing bias; and relia nce on a single CS \nevaluation, which could lead to inaccurate data. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by the \nNational Medical Ethics Committee of the Republic o f Slovenia (No. 0120-\n377/2019/4). \n \n \n \n \nAuthor Disclosures:  \nLan Umek: Nothing to disclose \nFajko F. Bajrović: Nothing to disclose \nKatarina Šurlan Popović: Nothing to disclose \nIngrid Požar: Nothing to disclose \n \n \nHigher Relative Brain Age of stroke patients treate d with mechanical \nthrombectomy is associated with poor outcomes \n*M. Guettier*, H. Biegalski, R. Lopes, J. Dumont, W . Gorwood, H. Henon,  \nN. Bricout, G. Kuchcinski, M. Bretzner; Lille/FR \n(melanie.gttr@hotmail.com) \n \nPurpose or Learning Objective: Relative brain age (RBA) is a novel MRI-\nderived biomarker that quantifies brain health rela tive to other patients within a \ncohort. It has previously been associated with poor er outcomes in untreated \nischemic stroke patients. However, its impact on po ststroke outcomes in \npatients treated with mechanical thrombectomy (MT) remains unclear.We \ninvestigated the clinical determinants of RBA and i ts association with \npoststroke outcomes in patients treated with MT \nMethods or Background: We conducted a retrospective analysis of clinical \nand imaging data from stroke patients admitted to L ille University Hospital for \nanterior circulation MT between 2015 and 2020. Admi ssion axial T2 FLAIR \nMRI images were used to obtain brain parcellation d ata. A modified brain age \nprediction pipeline was applied to estimate brain a ge and derive RBA. Linear \nregression was used to identify cardiovascular risk  factors associated with \nhigher RBA, while logistic regression was employed to assess the impact of \nRBA on post-stroke outcomes. \nResults or Findings: A total of 1,296 patients were included, with a mea n age \nof 70 years; 54% were women. Patients with a histor y of diabetes mellitus and \nsmoking had significantly higher RBA, indicating ol der-appearing brains \n(p=0.001 and p=0.010, respectively). Univariate ana lysis showed that patients \nwith higher RBA, reflecting poorer brain health, we re less likely to achieve \nfavorable functional outcomes after stroke (p=0.027 ). In multivariate analysis, \nseveral factors, including age, RBA, admission NIHS S score, intravenous \nthrombolysis, successful MT, glycemia, time from im aging to MT, and MT \nduration, were significantly associated with post-s troke outcomes (adjusted \nodds ratios: 0.48, 0.80, 0.48, 2.04, 5.72, 0.75, 0. 81, and 0.69, respectively). \nConclusion: Our study highlights the influence of smoking and d iabetes on \nbrain aging and the detrimental effects of poor bra in health on post-stroke \noutcomes, building on decades of clinical knowledge . \nLimitations: Retrospective data \nFunding for this study: This study has been funded by the ESR/EIBIR 2022 \nSeed Grant. \nEthics committee - additional information: The ethical committee (Comité \nde protection des personnes Nord-Ouest IV) classifi ed the study as \nobservational on March 9, 2010, and the committee p rotecting personal \ninformation of the patient approved the study by De cember 21, 2010 \n(n°10.677). Anonymized data supporting the findings  of this study are available \nfrom the corresponding author upon reasonable reque st. \nAuthor Disclosures:  \nWilliam Gorwood: Nothing to disclose \nJulien Dumont: Nothing to disclose \nMartin Bretzner: Nothing to disclose \nGrégory Kuchcinski: Nothing to disclose \nMélanie Guettier: Nothing to disclose \nHugo Biegalski: Nothing to disclose \nHilde Henon: Nothing to disclose \nNicolas Bricout: Nothing to disclose \nRenaud Lopes: Nothing to disclose \n \n \nSilent Brain Infarcts Post-Interventional Cardiac C atheterization: Insights \nfrom High-Resolution DW-MRI in a Randomized Study \n*N. Tan*, X. Zhou; Kunming/CN \n(1041023400@qq.com) \n \nPurpose or Learning Objective: To analyze the incidence and anatomic \ndistribution of acute cerebral embolism and identif y associated risk factors in \npatients undergoing interventional cardiac catheter ization. \nMethods or Background: We conducted a prospective study of patients from \ntwo cohorts between January 2023 and April 2024. Di ffusion-weighted \nmagnetic resonance imaging (DW-MRI) was used to det ect silent brain infarcts \n(SBIs) preoperatively and within 48 hours postopera tively. For transcatheter \naortic valve implantation (TAVI) patients, a cerebr al embolic protection device \n(CEPD, model CEP016F) was randomly employed to mini mize the risk of small \nemboli entering the cerebral vasculature. In atrial  fibrillation ablation \nprocedures, some patients underwent Vein of Marshal l Ethanol Infusion \n(VOMEt) to decrease the recurrence of atrial fibril lation and prevent the \nformation of new embolic events. Independent risk f actors were identified \nthrough multivariate logistic regression analysis. \n\n \n \nThursday \nAbstract-based Programme \n \n 133  \nResults or Findings: Silent brain infarcts were detected in 34 of 48 pat ients \n(70.8%) within 48 hours postoperatively. Among the patients who used CEPDs \n(n = 12), 7 (56.8%) found SBIs. Patients who used C EPDs showed a trend \ntoward smaller infarct volumes ( 5.32 cm³ vs. 8.11 cm³; P = 0.07). In patients \nwho underwent VOMEt procedures (n = 20), 13 (65.0%)  developed SBIs; \nhowever, neither the incidence nor the volume of SB Is showed significant \ndifferences compared to those who did not undergo V OMEt. Multivariate \nlogistic regression analysis identified operation t ime as an independent positive \npredictor of SBIs (odds ratios, 6.190 and 13.564; b oth P < .001). \nConclusion: Silent brain infarcts were detected in 70.8% of pat ients \nundergoing interventional cardiac catheterization, predominantly affecting the \nparietal lobes. These findings highlight the import ance of procedural \noptimization to reduce cerebral embolic risk. \nLimitations: The study was with a relatively small simple size, and conducted \nat a single center. \nFunding for this study: No \nEthics committee - additional information: Kunming Yan’an Hospital Ethics \nCommittee \nAuthor Disclosures:  \nNa Tan: Nothing to disclose \nXinyan Zhou: Nothing to disclose \n \n \nPrognostic Value of CT Contrast Staining after Endo vascular Therapy in \nBasilar Artery Occlusion Stroke \nP. Reidler, O. Öcal, J. Ricke, D. Puhr-Westerheide,  *M. P. Fabritius*; \nMunich/DE \n \nPurpose or Learning Objective: Contrast staining (CS) signifies prolonged \ntissue absorption of iodinated contrast media follo wing endovascular therapy \n(EVT) for large vessel occlusion stroke, indicating  blood-brain barrier \ndisruption. With EVT becoming the standard for trea ting basilar artery \nocclusion (BAO) stroke, our study aimed to determin e the prognostic \nsignificance of post-interventional CS in BAO strok e cases. \nMethods or Background: We included BAO patients who received \npostinterventional noncontrast CT within 24h after EVT. Expert radiologists \nconfirmed the presence of CS on CT and its volume w as quantified. Functional \noutcomes were assessed on the modified Rankin (mRS)  scale at 90 days and \nunfavorable outcome was defined as mRS ≥4. A multivariable LASSO-\npenalized logistic regression analysis was used to determine association of CS \nand other clinical and imaging parameters with func tional outcome. \nResults or Findings: 42 patients fulfilled the inclusion criteria (15 female, \n35,7%). CS on postinterventional CT was present in 18 patients (42.9%) with a \nmedian [interquartile range / IQR] volume of 7.9 mL  [3.7-14.6]. Patients with \nCS had a worse outcome with higher mRS after 90 day s (median [IQR]: 6 [4-6] \nvs. 2 [1-4], p<0.001). Multivariable LASSO analysis  revealed significant and \nstrongest association of CS with clinical outcome. \nConclusion: CS on postinterventional CT after EVT for BAO is an  independent \npredictor of unfavorable functional outcome, outper forming other pre- and post-\ninterventional imaging parameters. \nLimitations: Retrospective, small sample size \nFunding for this study: None \nEthics committee - additional information: LMU Munich \nAuthor Disclosures:  \nMatthias Philipp Fabritius: Nothing to disclose \nDaniel Puhr-Westerheide: Nothing to disclose \nOsman Öcal: Nothing to disclose \nPaul Reidler: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nEVT in young adults with stroke: Outcomes and proce dural \nconsiderations \n*C. Leneghan*¹, D. Leneghan², P. Nicholson¹, S. Sin gh¹, J. Thornton¹,  \nM. T. Crockett¹, A. O'Hare¹, J. Müller¹, P. Fearon¹ ; ¹Dublin/IE, ²Lucerne/CH \n(leneghac@tcd.ie) \n \nPurpose or Learning Objective: This study compares clinical presentations, \noutcomes, and procedural aspects of endovascular th rombectomy (EVT) in \nyoung adults (18-49 years) versus older adults ( ≥50 years) with acute ischemic \nstroke due to intracranial artery occlusion (IAO). \nMethods or Background: We analysed data from a prospectively maintained \nregistry of patients treated with EVT in a large te rtiary referral centre between \n2012 and 2022. Young (18-49) and older (≥50) patients were compared \nregarding baseline characteristics, 30- and 90-day modified Rankin Scale \n(mRS), 24-hour and 5-day NIHSS, reperfusion success  (mTICI), post-EVT \nASPECTS, and 90-day mortality. \n \n \n \n \nResults or Findings: Of 2201 patients, 243 (11%) were young adults. Youn g \npatients presented with lower median pre-treatment ASPECTS (8 vs. 9, \np=0.0008). Older patients had higher NIHSS at 24 ho urs (OR 1.63, 95% CI \n1.07-2.46), but no difference was observed at 5 day s. Functional dependence \n(mRS ≥3) was more frequent in older patients at 30 (OR 1. 92, 95% CI 1.45-\n2.53) and 90 days (OR 2.22, 95% CI 1.66-2.97). Mort ality at 90 days was lower \nin younger patients (OR 0.45, 95% CI 0.28-0.70). \nConclusion: Younger patients undergoing EVT for IAO have lower initial \nASPECTS but demonstrate faster neurological recover y and improved \nfunctional outcomes compared to older patients. Thi s suggests that aggressive \nEVT is warranted even in younger patients with lowe r ASPECTS. \nLimitations: Retrospective design, single-centre study. \nFunding for this study: No specific funding was received for this study. \nEthics committee - additional information: The study was approved by the \nlocal clinical audit committee. \nAuthor Disclosures:  \nPatricia Fearon: Nothing to disclose \nPatrick Nicholson: Nothing to disclose \nJohn Thornton: Nothing to disclose \nDarren Leneghan: Nothing to disclose \nAlan O'Hare: Nothing to disclose \nMatthew Thomas Crockett: Nothing to disclose \nCaoimhe Leneghan: Nothing to disclose \nSneha Singh: Nothing to disclose \nJennifer Müller: Nothing to disclose \n \n \nCould be venous MT safe and effective for dural sin us thrombosis? \n*M. T. Contaldo*, A. Cervo, A. Macera, C. Rollo, A.  Vitiello, G. Pero,  \nG. Schwarz, M. Sessa, M. Piano; Milan/IT \n(maria.contaldo@unimi.it) \n \nPurpose or Learning Objective: Cerebral venous thrombosis (CVT) is a rare \ncause of stroke, that tends to affect young people and the role of endovascular \ntreatment (EVT) remains debated. This study aims to  evaluate the efficacy and \nsafety of mechanical thrombectomy (MT) performed at  our center for dural \nsinus thrombosis. \nMethods or Background: Within a retrospective observational analysis \nconducted over a 6-year period, data from 62 patien ts referred to the stroke \nunit for CVT were analyzed. Among them, 32 patients , classified with severe \nCVT, underwent EVT. We assessed safety by examining  intraprocedural and \nperiprocedural adverse events (asymptomatic, mild, or severe). Clinical \noutcomes were evaluated at baseline and discharge. Efficacy was determined \nby analyzing the recanalization rate (complete, par tial, or absent) at the end of \nthe procedure. \nResults or Findings: A total of 32 patients received EVT, accounting for  36 \nprocedures. EVT with MT was performed within 6 hour s from onset in 21 out of \n32 cases. In most cases, MT was performed as primar y treatment, alongside \nbest medical therapy. An intravenous bolus of hepar in was administered in the \nangio-suite to patients who had not yet received an ticoagulants or \nthrombolytics. Successful recanalization (complete or partial without cortical \nvenous drainage delay) was achieved in 91.3% of pro cedures. CVT recurrence \noccurred in 3 out of 36 procedures. Mortality rate was 3.1%. \nConclusion: This is one of the largest series of patients treat ed with MT in \ncombination with best medical therapy, underscoring  the favorable safety and \nefficacy profile of EVT. MT could be a first-line o ption for multiple dural sinus \nthrombosis, rapid deterioration, venous hypertensio n, or anticoagulant failure. \nLimitations: This study is a single center experience with no co ntrol-group and \na retrospective design, with a heterogeneous case-b y-case patient selection. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nAmedeo Cervo: Nothing to disclose \nMaria Teresa Contaldo: Nothing to disclose  \nAntonio Macera: Nothing to disclose \nAlessio Vitiello: Nothing to disclose \nMaria Sessa: Nothing to disclose \nGuglielmo Pero: Nothing to disclose \nMariangela Piano: Nothing to disclose \nGhil Schwarz: Nothing to disclose \nClaudia Rollo: Nothing to disclose \n \n \n \n \n \n \n\n \n \n 134  \n \n  \nFriday, February 28 \n\n \n \nFriday \nAbstract-based Programme \n \n 135  \n \n08:00-09:00 Research Stage 1 \nResearch Presentation Session: Oncologic \nImaging \nRPS 1216 \nNew perspectives in breast and \ngynaecological cancer \n \nModerator \nG. Ivanac; Zagreb/HR  \n(gordana.augustan@gmail.com) \n \n \nPrognostic role of Whole-body MRI (WB-MRI) in patie nts with metastatic \nbreast cancer receiving systemic anti-cancer therap y \n*C. Pizzi*¹, C. Sattin¹, F. Arnone¹, D. Berloco¹, P . Hoxha¹, P. Summers¹,  \nR. Maggioni¹, A. R. R. Padhani², G. Petralia¹; ¹Mil an/IT, ²Northwood/UK \n(caterina.pizzi@unimi.it) \n \nPurpose or Learning Objective: To investigate the potential of the response \nassessment category (RAC) from MET-RADS-P guideline s as prognostic \nbiomarker in metastatic breast cancer (MBC) patient s. \nMethods or Background: We enrolled MBC patients who underwent whole-\nbody MRI at baseline and at each time point (every 12 weeks disease until \nprogression) after systemic anti-cancer therapy (SA CT). We correlated the \nmaximum RAC at time point 1 (TP1) with overall surv ival (OS). Patients were \ndivided in two groups: those with a maximum RAC 1-2  (highly likely or likely to \nbe responding, respectively) and those with a maxim um RAC 3-4-5 (stable \ndisease, likely or highly likely to be progressing)  at TP1. Survival curves were \ndepicted in Kaplan-Meier plots and compared via a l og-rank test and hazard \nratio (HR) using Cox regression model, with point c omparisons of three-year \nsurvival and median survival duration, using R. \nResults or Findings: Out of 45 MBC patients enrolled, a higher OS was \nobserved in patients with a maximum RAC 1-2 (N=18) than in those with a \nmaximum RAC 3-4-5 (N=27) at TP1 (log-rank test p=0. 007). Because more \nthan 50% of the maximum RAC 1-2 patients are still living, the median survival \nduration could not be determined, median survival i n the maximum RAC 3-4-5 \ngroup was 36 months (upper limit of 95%CI not avail able). The HR for the \nmaximum RAC 3-4-5 patients was 2.28 (95%CI 1.24 – 3 .33). Three years OS \nwas 88.9% for RAC1-2 vs 42.6% for RAC 3-4-5; for a difference of 46.2% \n(95%CI 12.7%-79.8%, p=0.0068). \nConclusion: Our observations support the potential of RAC after  TP1 as a \nprognostic biomarker in MBC patients undergoing SAC T. \nLimitations: Retrospective and monocentric study. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nRoberta Maggioni: Nothing to disclose \nFrancesca Arnone: Nothing to disclose \nGiuseppe Petralia: Nothing to disclose \nCaterina Pizzi: Nothing to disclose \nPaolo Hoxha: Nothing to disclose \nPaul Summers: Nothing to disclose \nAnwar R. R Padhani: Nothing to disclose \nCaterina Sattin: Nothing to disclose \nDonatello Berloco: Nothing to disclose \n \n \nDual-energy CT machine learning model to characteri ze lymph nodes in \npatients with breast cancer \n*P. Morrone*, C. Zampieri, C. Esposito, E. Barone, I. Capitoni, F. Gentili,  \nG. Bagnacci, S. Guerrini, M. A. Mazzei; Siena/IT \n(p.morrone1@student.unisi.it) \n \nPurpose or Learning Objective: To identify a machine learning (ML) model \nwith morphological and dual-energy (DE) data, to ch aracterize lymph node’s \n(LN) status during breast cancer (BC) staging. \nMethods or Background: From a cohort of 636 patients who undergone \nwhole-body DE-CT and subsequent surgery with axilla ry lymphadenectomy \nbetween April 2015 to July 2023, 117 patients were included. Exclusion \ncriteria: previous ipsilateral breast or axillary s urgery, or chemo-radiotherapy; \npoor quality CT; lack of anatomopathological data. For the morphological  \n \n \n \nanalysis, the main diameter of the neoplasm and loc ation, long and short axis \nand morphological features (fat hilum, cortical are a status, extranodal \nextension-ENE) of the ipsilateral axillary LNs were  recorded. For quantitative \nanalysis regions of interest (ROIs) were placed on the neoplasm and axillary \nLNs encompassing an area of post-contrast enhanceme nt as large and \nhomogeneous as possible. An attempt was made to pla ce the ROIs on the \nentire LN excluding the fat hilum and surrounding s tructures, setting a HU \ndisplaying threshold to suppress negative HU values . For each ROI, mean \nattenuation value at 40 and 70keV, iodine concentra tion (IC), water \nconcentration (WC) and effective-Z value were recor ded. \nResults or Findings: 116 BC and 375 LNs were analyzed, 180 pathological \nand 195 non-pathological. On univariate analysis th e attenuation (HU) at 40 \nand 70keV, slope, IC, WC, long and short LNs axis s howed statistically \nsignificant differences between histologically prov en pathological and non-\npathological LNs (p<0.001). There were statisticall y significant differences \n(p<0.001) according to the cortical status and ENE.  The logistic regression-\nbased ML model included IC, short axis, fat hilum, cortical status and ENE; the \nROC curve showed an AUC of 0.881, demonstrating goo d model accuracy. \nConclusion: The ML model provides a good discriminatory ability  to \ndifferentiate pathological from non-pathological ax illary LNs in patients with BC. \nLimitations: Not applicable \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Waived from our etical committe \ndue to the retrospective nature of this study. \nAuthor Disclosures:  \nChiara Esposito: Nothing to disclose \nMaria Antonietta Mazzei: Nothing to disclose \nIacopo Capitoni: Nothing to disclose \nPaola Morrone: Nothing to disclose \nFrancesco Gentili: Nothing to disclose \nCecilia Zampieri: Nothing to disclose \nElisa Barone: Nothing to disclose \nSusanna Guerrini: Nothing to disclose \nGiulio Bagnacci: Nothing to disclose \n \n \nAssociations between ADC histogram analysis values and tumor-micro \nmilieu in uterine cervical cancer \n*H-J. Meyer*¹, A-K. Höhn¹, A. Surov²; ¹Leipzig/DE, ²Minden/DE \n(jonas90.meyer@web.de) \n \nPurpose or Learning Objective: The complex interactions of the tumor \nmicromilieu could be reflected by diffusion-weighte d imaging (DWI) derived \nfrom the magnetic resonance imaging (MRI). The pres ent study investigated \nthe association between apparent diffusion coeffici ent (ADC) values and \nhistopathological features in uterine cervical canc er. \nMethods or Background: This retrospective study used the prebiopsy MRI to \nanalyze histogram ADC-parameters. The biopsy specim ens were stained for Ki \n67, E-cadherine, vimentin and tumor-infiltrating ly mphocytes (TIL, all CD45 \npositive cells). Tumor-stroma ratio (TSR) was calcu lated on routine H&E \nspecimen. Spearman’s correlation analysis and recei ver-operating \ncharacteristics curves were used as statistical ana lyses. \nResults or Findings: The patient sample comprised 70 female patients (ag e \nrange 32-79 years; mean age 55.4 years) with squamo us cell cervical \ncarcinoma. The interreader agreement was high rangi ng from intraclass \ncoefficient (ICC)=0.71 for entropy to ICC=0.96 for ADCmedian. Several ADC-\nhistogram parameters correlated strongly with the T SR. The highest correlation \ncoefficient achieved p10 (r=-0.81, p<0.0001). ADCme an can predict tumors \nwith high TSR, AUC: 0.91, sensitivity: 0.91 (95%CI 0.77;0.96), specificity: 0.91 \n(95%CI 0.78;0.97). Also, several ADC-histogram para meters correlated slightly \nwith the proliferation index Ki 67. No associations  were found with TIL, E-\nCadherin and vimentin. In well and moderately diffe rentiated cancers, ADC \nhistogram values showed stronger correlations with Ki 67 and TSR than in \npoorly differentiated tumors. \nConclusion: ADC values are strongly associated with tumor-strom a ratio. ADC \nmean can be used for prediction of tumors with high  TSR. Associations \nbetween histopathology and ADC values depend on tum or differentiation. ADC \nvalues show only weak associations with Ki 67 and n one with TIL, vimentin \nand E-Cadherin. \nLimitations: First, it is a retrospective study with known inher ent bias. Second, \nthe patient sample is comprised from a single cente r. \nFunding for this study: None \nEthics committee - additional information: Ethics commitee University of \nLeipzig (Ethical code: 012/13–28012013) \nAuthor Disclosures:  \nAlexey Surov: Nothing to disclose \nAnne-Kathrin Höhn: Nothing to disclose \nHans-Jonas Meyer: Nothing to disclose \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 136  \nLow-dose pre-operative CT of ovarian tumor with art ificial intelligence \niterative reconstruction for diagnosing peritoneal invasion \n*X. Cai*¹, J. Han², G. Zhang², F. Yang¹, Y. Wang¹, J. Liu¹, R. Li¹; \n¹Shijiazhuang/CN, ²Shanghai/CN \n \nPurpose or Learning Objective: To test the feasibility of low-dose \nabdominopelvic CT with an artificial intelligence i terative reconstruction (AIIR) \nfor diagnosing peritoneal invasion in pre-operative  imaging of ovarian tumor. \nMethods or Background: In this prospective study, 88 patients with \npathology-confirmed ovarian tumors were enrolled, w here the routine-dose CT \nscan at portal venous phase (120 kV/ref. 200 mAs) w as followed immediately \nwith a low-dose scan (120 kV/ref. 40 mAs). Images a t routine dose were \nreconstructed with hybrid iterative reconstruction (HIR) and images at low dose \nwere reconstructed with AIIR. Two radiologists inde pendently diagnosed the \nperitoneal invasion using a 5-point confidence scal e (1: definitely absent, 5: \ndefinitely present). In case of disagreement, the c onsensus was obtained \nthrough a third radiologist. The diagnostic perform ance was assessed using \nreceiver operating characteristic (ROC) analysis wi th pathological results \nserving as the reference. The inter-observer agreem ent was assessed by \nCohen’s Kappa test. \nResults or Findings: The 88 patients consisted of 37 patients with \nbenign/borderline ovarian tumors and 51 patients wi th ovarian carcinomas. The \neffective dose of low-dose CT at portal venous phas e was 79.8% lower than \nthat of routine-dose scan (2.64 ± 0.46 mSv vs. 13.04 ± 2.25 mSv, p < 0.001). In \ndiagnosing peritoneal invasion, the area under the ROC curve (AUC) of low-\ndose AIIR and routine-dose HIR images was 0.961 and  0.960, respectively (p \n= 0.734). The sensitivity, specificity, and accurac y were 86.1%, 92.3%, and \n89.8%, respectively, for low-dose AIIR images, and 86.1%, 90.4%, and 88.6%, \nrespectively, for routine-dose HIR images. The inte robserver agreement was \ngood for diagnosing peritoneal invasion (κ = 0.694). \nConclusion: In low-dose pre-operative CT of ovarian tumor with 80% dose \nreduction, AIIR delivers similar diagnostic accurac y for peritoneal invasion as \ncompared to routine abdominopelvic CT. \nLimitations: Not applicable. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number 2024KS138. \nAuthor Disclosures:  \nJing Liu: Nothing to disclose \nRuxun Li: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nYaning Wang: Nothing to disclose \nXiaojia Cai: Nothing to disclose \nFan Yang: Nothing to disclose \nJintao Han: Nothing to disclose \n \n \nDeveloping a machine learning model for the differe ntiation of uterine \nleiomyosarcoma from leiomyomas using clinical and M RI radiomics \nfeatures \n*K. Shapriya*, A. Jackson, X. Li, S. Das, N. Bharwa ni, A. G. Rockall; \nLondon/UK \n(kavita.shapriya@nhs.net) \n \nPurpose or Learning Objective: Preoperative differentiation between \nleiomyosarcoma (LMS) and atypical benign leiomyoma (LM) is diagnostically \nchallenging. This study aims to develop and validat e a machine learning (ML) \nmodel using MRI-based clinical and radiomic feature s to distinguish LMS from \nLM. \nMethods or Background: This retrospective study included 214 patients with  \natypical myometrial lesions who underwent surgery b etween 2013 and 2023. \nAll subjects had preoperative full blood count (FBC ) and MR imaging. Among \n214 cases, 193 were LM and 21 were LMS. T2-weighted  sagittal MRI \nsequences were manually segmented then optimized us ing nonuniformity \ncorrection method to ensure image stability. 4114 r adiomic features were \nextracted using TexLab (version 2) and IBSI complia nt MATLAB™ software. \nThese radiomic features and 11 clinical variables ( including age and FBC) \nwere incorporated into several ML models, with the dependent variable being \nLMS (binary). Data was split 70:30 into training an d testing sets. To address \ndata imbalance, an ensemble classification model wa s employed with unequal \nclassification costs, penalizing the misclassificat ion of LMS. The model’s \nperformance was evaluated using area under the curv e (AUC), sensitivity, \nspecificity, accuracy, F1 score, and confusion matr ix. \nResults or Findings: The final ensemble model included four clinical and  six \nradiomic features. The test dataset demonstrated a sensitivity, specificity, \naccuracy and AUC of 1, 0.8, 0.84 and 0.90, and F1 s core of 0.50. \nConclusion: This study presents a promising ML model for preope rative \ndifferentiation of LMS from LM, achieving high accu racy (84%). As sarcoma \nsubjects are uncommon, the ML model was developed t o take data imbalance \ninto account. High sensitivity was achieved, but wi th some loss of specificity. \nFuture research will focus on validating this model  using larger datasets to \nenhance its reliability and clinical application. \nLimitations: None \nFunding for this study: No funding \nEthics committee - additional information: HRA and Health and Care \nResearch Wales (HCRW) \nReference: 20/HRA/4925 \nAuthor Disclosures:  \nXingfeng Li: Nothing to disclose \nKavita Shapriya: Nothing to disclose \nAndrea Grace Rockall: Nothing to disclose \nSaranya Das: Nothing to disclose \nAlastair Jackson: Nothing to disclose \nNishat Bharwani: Nothing to disclose \n \n \n08:00-09:00 Research Stage 2 \nResearch Presentation Session: \nMusculoskeletal \nRPS 1210 \nImaging of body composition \n \nModerator \nG. Guglielmi; Foggia/IT  \n(giuseppe.guglielmi@unifg.it) \n \n \nClinical validation of a deep learning based automa ted HUAC analysis for \nimproved sarcopenia assessment \nV. K. Venugopal¹, V. Rengan², *S. Ingole*¹; ¹New De lhi/IN, ²Chennai/IN \n(sarang.ingole@carpl.ai) \n \nPurpose or Learning Objective: To assess the validity and reliability of an \nautomated sarcopenia estimation approach using a de ep-learning based \nensemble psoas segmentation and Hounsfield Unit Ave rage Calculation \n(HUAC) model in comparison to traditional manual me asurements. \nMethods or Background: This study retrospectively analyzed 149 CT scans, \ncomparing sarcopenia assessments between manual HUA C measurements \nand those derived from an automated TransUNet-based  system. The AI model \ncombined convolutional neural networks with Transfo rmer blocks to enhance \nfeature extraction and contextual understanding of muscle tissue, crucial for \nprecise sarcopenia evaluation. The HUAC was calcula ted by measuring the \narea and mean Hounsfield Units (HU) of the left and  right psoas muscles at the \nL3 vertebra level. Statistical analysis included me an, standard deviation, \ncorrelation, paired t-tests, Bland-Altman plots, an d advanced validation metrics \nsuch as Intersection over Union (IoU) and Dice coef ficient to evaluate the \nmodel's segmentation accuracy. \nResults or Findings: The AI model produced a mean HUAC of 19.66, slightl y \nhigher than the 18.03 from manual assessments, with  corresponding standard \ndeviations of 4.27 and 4.54, respectively. The corr elation coefficient of 0.78 \nindicated strong agreement between the two methods.  The model achieved an \nIoU of 90% and a Dice coefficient of 0.90, demonstr ating high precision in \nmuscle segmentation. The systematic bias observed ( mean difference of -1.63 \nHUAC) highlights areas for further calibration of t he AI model. \nConclusion: The integration of AI in sarcopenia assessment thro ugh HUAC \ncalculations offers a promising alternative to manu al measurements, providing \nspeed, reproducibility, and precision. Despite some  variance, the AI method \naligns closely with traditional approaches, suggest ing that with further \nrefinement, it could become a standard tool in clin ical settings. \nLimitations: Small sample set \nFunding for this study: Nil \nEthics committee - additional information: IRB Waiver \nAuthor Disclosures:  \nSarang Ingole: Nothing to disclose \nVasantha Kumar Venugopal: Consultant: Carpl.ai \nVinayak Rengan: Founder: Curium life technologies \n \n \nQualitative and quantitative CT evaluation of abdom inal fat and muscle \ntissue in patients with ankylosing spondylitis and investigation of their \npossible effects on biological agent treatment resp onse \nN. Kaştan, *I. Erdem Toslak*, S. Bakırcı, A. Yavuz; Antal ya/TR \n(driclalerdem@yahoo.com) \n \nPurpose or Learning Objective: Ankylosing spondylitis (AS) is a chronic \ninflammatory disease affecting the axial skeleton. Inflammatory cytokines like \nTNFalpha and interleukins increase in AS, along wit h adipose tissue and \n\n \n \nFriday \nAbstract-based Programme \n \n 137  \nmuscle catabolism. Sarcopenia in AS correlates with  higher inflammation, \ngreater disease activity, and reduced muscle perfor mance. We hypothesize \nthat the sarcopenia index and quantitative measures  of muscle and fatty \ntissues may relate to the response to biological ag ent treatment \nMethods or Background: This retrospective study involved 62 adults \ndiagnosed with AS who underwent CT at the L2 verteb ra level before any \ntreatment and received biological agent therapy. CT  measurements included \nvisceral and subcutaneous abdominal adipose tissue cross-sectional area \n(VAT cm², SAT cm²), total abdominal muscle area (TA MA), psoas muscle \nvolume (PsoA), sarcopenia index (SMI), visceral and  subcutaneous abdominal \nadipose tissue attenuation (VAT HU, SAT HU), and ps oas muscle attenuation \n(Pso HU). BASDAI score changes were assessed at the  first post-treatment \nvisit to evaluate disease activity. Comparisons and  correlations were \nperformed between SMI, adipose and muscle tissue me asurements, and \nclinical parameters. \nResults or Findings: TAMA, PsoA, and SMI values were significantly highe r \nin patients with full recovery compared to those wi th partial recovery (p<0.05). \nHowever, VAT and SAT (cm²), VAT and SAT (HU), and P so HU did not \nsignificantly affect recovery (p>0.05). This study demonstrates that SMI, \nTAMA, and psoas muscle volume can serve as prognost ic markers for \nbiological treatment response in AS patients. \nConclusion: Our study is an example of opportunistic-quantitati ve imaging \nmethods and it has been shown that SMI value, TAMA and psoas muscle \nvolume value can be used as prognostic markers in r esponse to biological \ntreatment in AS patients. \nLimitations: Retrospective design and small sample size were the  limitations \nof our study. \nFunding for this study: No external funding. \nEthics committee - additional information: Local IRB \nAuthor Disclosures:  \nIclal Erdem Toslak: Nothing to disclose \nAlpaslan Yavuz: Nothing to disclose \nNazmi Kaştan: Nothing to disclose \nSibel Bakırcı: Nothing to disclose \n \n \nOpportunistic Osteoporosis Assessment from Routine CT - Effect of \nIntravenous Contrast Agents on Absolute Values, T-S cores, and Derived \nClassifications in Single- and Dual-Energy CT \nL. D. Grünewald, V. Koch, S. Mahmoudi, J-E. Scholtz , *J. Gotta*, S. Martin,  \nC. Booz, I. Yel, T. Vogl; Frankfurt/DE \n \nPurpose or Learning Objective: To evaluate the impact of intravenous \ncontrast agents on osteoporosis assessment via rout ine CT in arterial and \nvenous phases and identify mitigation strategies us ing dual-energy CT \n(DECT). \nMethods or Background: 288 patients (154 men, 134 women; median age 62 \nyears) who underwent abdominal DECT scans in non-co ntrast, late-arterial, \nand portal venous phases between January 2018 and D ecember 2023 were \nretrospectively analyzed. Trabecular HU values were  measured in all phases, \nincluding 90kV and 150kV DECT series, using automat ic segmentation. T-\nscores were calculated to classify patients as oste oporotic, osteopenic, or \nnormal. Changes in HU values, T-scores, and classif ications due to contrast \nwere compared to non-contrast images, with effects quantified using  \nCohen’s d. \nResults or Findings: Median trabecular HU at L1 was 147 (IQR 116–185). \nContrast in late arterial and portal venous phases increased HU values by \n+14.4 (+11.2%) and +25.7 (+20.7%), respectively. Us ing 150kV DECT reduced \nthese changes to -20.5 (-12.2%) for arterial and -2 3.15 (-12.6%) for venous \nphases. Cohen’s d was lowest for normal arterial ph ase (+0.55) and highest for \n90kV arterial phase (+1.9). Based on T-scores, 120 patients were classified as \nhealthy, 108 as osteopenic, and 60 as osteoporotic.  The lowest number of \nreclassifications occurred in arterial (n=92) and v enous (n=104) phases. For \narterial phase, 44 patients shifted from osteoporos is to osteopenia; for venous \nphase, 52 shifted similarly. High-kV acquisition re duced these reclassifications \n(n=24 arterial, n=32 venous) but increased shifts f rom healthy to osteopenia. \nConclusion: Intravenous contrast significantly affects HU-based  osteoporosis \nassessment, leading to reclassifications, especiall y from osteopenia to healthy. \nUsing 150kV DECT can partially reduce these reclass ifications, though it may \nincorrectly shift healthy cases toward osteopenia. \nLimitations: Modifying kV settings is not immediately possible w ithout \ndedicated equipment \nFunding for this study: No funding was received for this study \nEthics committee - additional information: Consent waived due to \nretrospective nature \n \n \n \n \n \n \n \nAuthor Disclosures:  \nSimon Martin: Nothing to disclose \nChristian Booz: Nothing to disclose \nIbrahim Yel: Nothing to disclose \nThomas Vogl: Nothing to disclose \nJan-Erik Scholtz: Nothing to disclose \nVitali Koch: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \nJennifer Gotta: Nothing to disclose \n \n \nFemoral osteoporosis prediction model using autoseg mentation and \nmachine learning analysis with PyRadiomics on abdom en-pelvic \ncomputed tomography (CT) \n*H. Ha*¹, H. Lim², M. Park¹; ¹Anyang-Si/KR, ²Seoul/ KR \n(ha.hongil@gmail.com) \n \nPurpose or Learning Objective: This study aimed to assess the diagnostic \nperformance of osteoporosis prediction by the combi nation of \nautosegmentation of the proximal femur and machine learning analysis with a \nreference standard of dual-energy X-ray absorptiome try (DXA) \nMethods or Background: Abdomen-pelvic CT scans were retrospectively \nanalyzed from 1,122 patients who received both DXA and abdomen-pelvic \ncomputed tomography (APCT) scan from January 2018 t o December 2020. \nThe study cohort consisted of a training cohort and  a temporal validation \ncohort. The left proximal femur was automatically s egmented, and a prediction \nmodel was built by machine-learning analysis using a random forest (RF) \nanalysis and 854 PyRadiomics features. The technica l success rate of \nautosegmentation, diagnostic test, area under the r eceiver operator \ncharacteristics curve (AUC), and precision recall c urve (AUC-PR) analysis \nwere used to analyze the training and validation co horts. \nResults or Findings: The osteoporosis prevalence of the training and \nvalidation cohorts was 24.5%, and 10.3%, respective ly. The technical success \nrate of autosegmentation of the proximal femur was 99.7%. In the diagnostic \ntest, the training and validation cohorts showed 78 .4% vs. 63.3% sensitivity, \n89.4% vs. 98.1% specificity. The prediction perform ance to identify \nosteoporosis within the groups used for training an d validation cohort was high \nand the AUC and AUC-PR to forecast the occurrence o f osteoporosis within \nthe training and validation cohorts were 90.8% [95%  confidence interval (CI), \n88.4–93.2%] vs. 78.0% (95% CI, 76.0–79.9%) and 94.6 % (95% CI, 89.3–\n99.8%) vs. 88.8% (95% CI, 86.2–91.5%), respectively . \nConclusion: The osteoporosis prediction model using autosegment ation of \nproximal femur and machine-learning analysis with P yRadiomics features on \nAPCT showed excellent diagnostic feasibility and te chnical success. \nLimitations: The limitation of this study was that there was an imbalance in the \nsex ratio of osteoporosis patients and that this wa s a single-center study. \nFunding for this study: This work was supported by the Central Medical \nService (CMS) Research Fund. The specific grant num ber was not assigned by \nthe company or funder (Central Medical Service Comp any, Ltd., Seoul, Korea). \nEthics committee - additional information: The study was approved by the \ninstitutional review board of Hallym University Sac red Heart Hospital (No. \nHALLYM 2020-12-015), and the need for informed cons ent was waived due to \nthe nature of the retrospective analysis. \nAuthor Disclosures:  \nHongil Ha: Nothing to disclose \nMinsu Park: Nothing to disclose \nHyunkyung Lim: Nothing to disclose \n \n \nRadiomic Analysis of Thigh Fat Fraction Maps to Ide ntify Patterns in \nNeuromuscular Disorders \n*G. Vignati*, M. Moscatelli, R. Fabrizio, R. Pascuz zo, C. Foschini, F. Doniselli, \nD. Aquino, F. Mazzi, L. M. Sconfienza; Milan/IT \n(giacomo.vignati@unimi.it) \n \nPurpose or Learning Objective: To analyze radiomics features extracted \nfrom thigh fat fraction (FF) maps in order to ident ify common patterns in \nneuromuscular disorders across different patients. \nMethods or Background: Radiomics features of the classes “shape”, “first-\norder”, and “gray-level co-occurrence matrix” (GLCM ) were extracted from the \nthigh FF maps of all patients for each of the 13 VO Is using PyRadiomics, with \na fixed bin size of 32. A preliminary feature selec tion step was necessary due \nto the large number of features extracted (n=1305) relative to the limited \nnumber of patients (n=25). After feature selection,  the sparse K-means \nclustering algorithm was applied: it is a clusterin g approach that identifies \nrelevant features while performing clustering. Fina lly, Uniform Manifold \nApproximation and Projection (UMAP) was used to vis ualize the selected \nfeatures and statistical analyses were done with R (version 4.3.1) using the \ncaret, sparcl, and umap packages. \n \n\n \n \nFriday \nAbstract-based Programme \n \n 138  \nResults or Findings: The sparse K-means algorithm identified two cluster s of \n14 and 11 patients, respectively, based on 60 selec ted radiomic features from \n8 muscles. Clinical diagnosis of patients affected by neuromuscolar disorders \nis compared with cluster assignment and distinctive  features were observed \nbetween genetic and inflammatory/autoimmune myopath ies. \nConclusion: This study successfully utilized radiomics features  from thigh fat \nfraction (FF) maps to identify distinct patterns in  neuromuscolar disorders \nacross different patients suggesting that radiomic analysis could be a valuable \ntool for understanding and classifying muscle disor ders in clinical settings. \nLimitations: Limitations of this study are the small sample size  (only 25 \npatients), the large number of initial radiomic fea tures (n=1305) and the feature \nselection process, which could introduce bias. The study also relied on \nmanually segmented regions of interest (VOIs), whic h may introduce variability \nin the analysis \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nDomenico Aquino: Nothing to disclose \nMarco Moscatelli: Nothing to disclose \nFabio Doniselli: Nothing to disclose \nFederica Mazzi: Nothing to disclose \nGiacomo Vignati: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nChiara Foschini: Nothing to disclose \nRiccardo Pascuzzo: Nothing to disclose \nRenato Fabrizio: Nothing to disclose \n \n \nMultiparametric MRI at 3 and 7 T for characterizati on of skeletal muscle \npathology in patients with filamin-C, desmin and LB D3 mutations \n*C. S. Mathy*, L. V. Gast, T. Gerhalter, M. Türk, T . Bäuerle, A. Doerfler,  \nM. Uder, A. M. Nagel, R. Schröder; Erlangen/DE \n \nPurpose or Learning Objective: To characterize patterns of skeletal muscle \nchanges in myofibrillar myopathy (MFM), a group of rare neuromuscular \ndiseases with desmin-positive aggregates and myofib rillar degeneration, using \nmultiparametric MRI. \nMethods or Background: Less affected lower leg of nine patients with \ngenetically confirmed MFM due to FLNC (n=5), DES (n =2) and LDB3 (n=2) \nmutations (50.9±8.6 years, 6m, 3f) and 10 healthy controls (50.0±11.0 years, \n6m, 4f) were examined. 1H-MRI at 3 T included T1-we ighted and T2-weigthed \nSTIR for qualitative assessment of fatty replacemen t/edema, Dixon-type \nsequence for proton-density fat fraction (PDFF) qua ntification and diffusion-\ntensor imaging (DTI) for characterization of (micro -)structural changes. \n39K/23Na-MRI acquisition-weighted Stack-of-Stars se quences at 7 T allowed \nafter partial-volume and relaxation correction quan tification of apparent tissue \nsodium/potassium concentrations (aTSC/aTPC). \nResults or Findings: Muscular fatty replacement and edema-like alteratio ns \nwere highly variant intermuscular and interindividu al. 35/63 of elevated muscle \ncompartments of patients with MFM were highly fatty  replaced (PDFF >50%). \nCalculated apparent diffusion coefficients (ADCs) f rom DTI were reduced in \ngastrocnemius lateralis (GL), peroneus (PER) and ex tensor digitorum longus \n(EDL) muscles (p = 0.04 – 0.03) when excluding high ly fatty replaced muscles, \nbut simulations showed that this behavior could pri marily be attributed to \nincreasing PDFF. Fat-corrected aTSC were increased in all muscle regions \n(mean all muscles: 55.6±16.3 mM vs 23.2±5.5 mM, p <0.001), aTPC \ndecreased in all regions but GL and PER (mean all m uscles: 75.4±13.3 mM vs \n108.9±9.9 mM, p <0.001). \nConclusion: Alterations of 39K/23Na ion homeostasis that go bey ond \nchanges caused by fatty-replacement in irreversible  disease stages could be \nproved in patients with MFM. This could form the ba sis for a novel biomarker \nfor determining early disease extent and disease re sponse to therapies. \n(Micro-)structural changes were indistinguishable f rom mere fatty replacement \nchanges. \nLimitations: Low number of participants (MFM prevalence <1:100.0 00). \nFunding for this study: C.S.M. and T.B. were founded by the Deutsche \nForschungsgemeinschaft (DFG, German Research Founda tion) – 493624887 \n(Clinician Scientist Program NOTICE). Funding by th e DFG is gratefully \nacknowledged (project 500888779 / RU5534 MR biosign atures at UHF). \nEthics committee - additional information: Approved by local ethic comittee \nof Friedrich-Alexander University Erlangen-Nurember g \nAuthor Disclosures:  \nTobias Bäuerle: Nothing to disclose \nTeresa Gerhalter: Nothing to disclose \nClaudius Sebastian Mathy: Nothing to disclose \nMatthias Türk: Nothing to disclose \nMichael Uder: Nothing to disclose \nRolf Schröder: Nothing to disclose \nLena V. Gast: Employee: Siemens Healthineers \nArmin M. Nagel: Nothing to disclose \nArnd Doerfler: Nothing to disclose \n \nThe Top 100 Most Cited Articles on Musculoskeletal Radiology:  \nA Bibliometric Analysis \n*L. Moore*; Dublin/IE \n(lucymooreart@gmail.com) \n \nPurpose or Learning Objective: To identify and characterize the most \ninfluential publications relating to the musculoske letal system and radiology. \nThe number of citations an article receives is refl ective of its impact in the \nscientific community. \nMethods or Background: The top 100 most cited articles were identified \nusing the Web of Science database. Data pertaining to the year of publication, \npublishing journal, journal impact factor, authorsh ip, country of origin and \ninstitution were collected. \nResults or Findings: The number of citations per article for the top 100  list \nranged from 149 to 709 (median 208; mean 240). The average number of \ncitations per year, per article, ranged from 5 to 6 0 (median 12, mean 26). The \nUnited States was the most common country of origin  (n=74). The Journal with \nthe greatest number of articles was Radiology (n=34 ). The University of \nCalifornia contributed the most articles (n=11). \nConclusion: This study presents a detailed analysis of the top 100 most-cited \narticles published in musculoskeletal radiology. It  provides clinicians and \nresearchers with insight into the current influenti al research papers in this field \nand the characteristics of those studies. It also h ighlights research trends and \nareas that may benefit from further research. \nLimitations: The use of citation count is a source of bias; the more time that \nhas passed since the publication of an article, the  more likely it is to be cited \nover time. In order to mitigate this source of bias , the average citation count \nper year was also used. Some articles may have been  inadvertently excluded \nas a result of search criteria used. In addition, using journal IF from one \nparticular year (2024) does not allow for temporal fluctuations in IF. \nFurthermore, the potential bias of self-citation wa s not accounted for in this \nstudy. \nFunding for this study: None. \nEthics committee - additional information: This article does not require \nethics committee approval. \nAuthor Disclosures:  \nLucia Moore: Nothing to disclose \n \n \n08:00-09:00 Research Stage 3 \nResearch Presentation Session: Physics \nin Medical Imaging \nRPS 1213 \nEvolution of CT: a key to its sustainability \n \nModerator \nN. Saltybaeva; Lucerne/CH  \n(natalia.saltybaeva@luks.ch) \n \n \nValidation of a novel CBCT reconstruction algorithm  for treatment \nplanning and IGRT in neoadjuvant radiotherapy of lo cally advanced rectal \ncancer patients \n*M. C. Daniotti*¹, S. Trivellato², L. De Sanctis¹, V. Pisoni², J. Stancanello³,  \nJ. Mason³, R. Pellegrini³, S. Arcangeli², E. De Pon ti²; ¹Milan/IT, ²Monza/IT, \n³Stockholm/SE \n(martina.daniotti@gmail.com) \n \nPurpose or Learning Objective: A new CBCT reconstruction algorithm based \non poli-energetic quantitative (Polyquant) method e mpowered with a \nconvolutional neural network scatter correction has  been recently proposed. \nThis study aimed to validate the use of the Polyqua nt CBCTs (pCBCTs) for \nimage-guided radiotherapy (IGRT) and planning for l ocally advanced rectal \ncancer (LARC). \nMethods or Background: Translational shifts obtained on all axes with the \nregistration of pCBCTs to CT were compared to the c linical version of Elekta \nXVI CBCT ones and statistical significance was inve stigated with the t-test and \nANOVA-test. pCBCTs were calibrated with a populatio n-based curve (pop-CC) \nelaborated coupling pCBCT gray levels to the CT rel ative electron density \n(RED) for ten pelvic patients. pop-CC was validated  by comparing dose \ncalculations on pCBCT and bulk density pseudo-CT us ing 1%/3mm local \ngamma-analysis. Five LARC patients treated on Elekt a VersaHD were \nselected. The RED difference between CT and first s ession-pCBCT were \nassessed on a voxel-to-voxel basis, on a contour ba sis, and on a dosimetric \nbasis using 1%/3mm local gamma-analysis to compare pCBCT and pseudo-\nCT calculations. \n\n \n \nFriday \nAbstract-based Programme \n \n 139  \nResults or Findings: The translational differences between IGRT results of \nCBCT and pCBCT were always <1mm and not statistical ly significant. The \npop-CC pCBCT calibration resulted in dose calculati ons comparable with the \npseudo-CT ones, with gamma passing rates > 95%. For  LARC patients, voxel-\nto-voxel and structure-based analysis showed no rel evant RED discrepancies \nbetween pCBCT and CT. Residual RED differences resu lted dosimetrically \nnegligible compared with dose distributions calcula ted on pseudo-CT, with \ngamma passing rates > 95%. \nConclusion: The optimized pCBCTs were successfully RED calibrat ed and \nvalidated for IGRT and planning for LARC radiothera py. The results suggest \nthat pCBCTs could be exploited in the clinical work flow for adaptive \nradiotherapy in LARC patients, \nLimitations: Further investigation for their extended use might still be \nnecessary. \nFunding for this study: No funding \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nJonathan Mason: Employee: Elekta AB \nSara Trivellato: Nothing to disclose \nMartina Camilla Daniotti: Nothing to disclose \nStefano Arcangeli: Nothing to disclose \nElena De Ponti: Nothing to disclose \nValerio Pisoni: Nothing to disclose \nJoseph Stancanello: Employee: Elekta AB \nLorenzo De Sanctis: Nothing to disclose \nRoberto Pellegrini: Employee: Elekta AB \n \n \nThe impact of detector coverage on motion artefacts  in photon-counting \nCT imaging using a motion phantom \n*E. Verelst*¹, G. Van Gompel¹, D. Crotty², H. Linde r³, P. D. Deak⁴, J. De Mey¹, \nN. Buls¹; ¹Brussels/BE, ²Cork/IE, ³Stockholm/SE, ⁴Münsingen/CH \n \nPurpose or Learning Objective: Reducing motion-induced artefacts is an \nincreasingly important aspect of ultra high-resolut ion (UHR) photon-counting \nCT (PCCT). Using a motion-controlled phantom, this study evaluates the \nbenefit of using wider detector coverage in PCCT to  reduce motion artefacts. \nMethods or Background: A tissue-simulating cuboid phantom, 80-mm in \nlength, was mounted onto a custom-built motion-cont rolled phantom. A 3-mm \ndiameter stent (Superia, Nano-Therapeutics, India) was inserted into a \ncommensurate hole inside the phantom. Programmed to  simulate pulsatile \nmotion, helical images were acquired during motion by a prototype Silicon-\nbased PCCT (Si-PCCT) using 40- and 80-mm detector c overage, representing \ntable speeds of 128.6-mm/s and 257.1-mm/s, respecti vely. Motion-induced \nimage artefacts were evaluated by measuring the vol umetric error relative to \nthe nominal phantom volume. A 5-point Likert rating  evaluated stent \nappearance for both detector coverages against a re ference static image (1-no \nsimilarity, 5-similar). Differences were assessed u sing a paired sample t-test \nand Wilcoxon signed-rank test, respectively. P-valu es < 0.05 indicated \nstatistical significance. \nResults or Findings: Relative to the nominal phantom volume (1571-mm3), \nthe 80-mm detector coverage statistically significa ntly reduced volumetric error \n(mean 81-mm3, SD 18-mm3) compared to the error gene rated with 40-mm \ncoverage (254-mm3, 48-mm3), p=0.006. Stent appearan ce under wider \ndetector coverage was likewise assessed to better m atch the reference image, \nwith average Likert scores for 40 mm and 80mm of 1 [1–1.5] and 4 [3.5–3.75], \nrespectively, p=0.043. \nConclusion: In high-resolution PCCT imaging, to combat motion-i nduced \nartefacts, it is important to combine wide-detector  UHR CT acquisitions with \nhigher table speeds. This study demonstrates the po tential of using a prototype \nwide-coverage Si-PCCT system with fast tables speed s to reduce such \nartefacts. While this study specifically evaluated pulsatile motion, an 80-mm \ndetector coverage may reduce additional body motion -induced artefacts, such \nas peristalsis. \nLimitations: This is an ex-vivo phantom study. \nFunding for this study: Flemish Research Foundation (FWO), personal grant, \nnr: 1SH1Z24N. \nEthics committee - additional information: Ethical approval was not required \nfor this study \nAuthor Disclosures:  \nJohan De Mey: Nothing to disclose \nEmma Verelst: Nothing to disclose \nDominic Crotty: Employee: GE Healthcare \nGert Van Gompel: Nothing to disclose \nPaul D. Deak: Employee: GE Healthcare \nNico Buls: Nothing to disclose \nHugo Linder: Employee: GE Healthcare \n \n \n \n \n3D-printed Anthropomorphic Head Phantom Featuring W hite and Gray \nMatter Structures for Evaluating CT Imaging \nK. Mei¹, L. Roshkovan¹, S. Sharma², S. Ross², J. Wo o¹, S. S. Halliburton³,  \n*L. Liu*¹, R. Thompson³, P. Noël¹; ¹Philadelphia, P A/US, ²Vernon Hills, IL/US, \n³Mayfield Village, OH/US \n \nPurpose or Learning Objective: To develop a 3D-printed, patient-specific \nbrain phantom for assessing performance of non-cont rast CT head imaging. \nMethods or Background: Unenhanced T1 MRI scan (best available gold \nstandard) of a healthy brain (21y/o,F) was retrospe ctively collected and \nconverted into CT Hounsfield unit image to generate  a realistic phantom using \nPixelPrint technique. The brain phantom was created  as a 30 mm thickness \nsection including both left and right hemispheres ( approximately 157 x 120 \nmm) at 1:1 scale. Additionally, a separate skull ph antom was printed from the \nsame patient images using calcium-doped filament. T he brain phantom, with \nsurrounding skull phantom, was scanned with CT at 1 20 kVp and 24.4 mGy. \nImages were reconstructed with and without iterativ e denoising at 0.5 mm pixel \nspacing. Attenuation values were measured in gray a nd white matter. \nResults or Findings: Gray and white matter were clearly distinguishable in an \nappropriate CT examination window, with and without  denoising. Line profile \nwas plotted along the center of the phantom. Realis tic CT values of \napproximately 45 HU for gray matter and 25 HU for w hite matter were \nobserved. Maximum density observed in the skull rea ched approximately 750 \nHU, which was limited by the density of the 3D-prin ting filament used. Image \nnoise, estimated by standard deviation, ranged betw een 3 and 4.5 HU across \nboth sets of denoised images. \nConclusion: The PixelPrint 3D-printed brain phantom successfull y depicts \nrealistic tissue attenuationfor white and gray matt er, demonstrating potential as \na valuable tool for evaluating CT head imaging perf ormance. \nLimitations: This study converts MR images to CT numbers and sim ulates \nnon-contrast CT scans. \nFunding for this study: This work was partly supported by Canon Medical \nSystems Corporation (Otawara, Japan). \nEthics committee - additional information: University of Pennsylvania \nAuthor Disclosures:  \nSteven Ross: Employee: Canon Medical Research USA \nKai Mei: Nothing to disclose \nPeter Noël: Nothing to disclose \nRichard Thompson: Employee: Canon Healthcare USA \nLeening Liu: Nothing to disclose \nSandra Simon Halliburton: Employee: Canon Healthcar e USA \nLeonid Roshkovan: Nothing to disclose \nShobhit Sharma: Employee: Canon Medical Research US A \nJohn Woo: Nothing to disclose \n \n \nTowards functional lung color K-edge imaging enable d by spectral \nphoton-counting CT in combination with dedicated co ntrast agents:  \na phantom study \n*A. J. Gutwinska*¹, D. Rosario², A. Pang², C. A. He rnandez-Fajardo¹,  \nR. Coulibaly¹, A. Robert¹, S. Rit¹, D. P. Cormode²,  S. A. Si-Mohamed³; \n¹Lyon/FR, ²Philadelphia, PA/US, ³Bron/FR \n(agnieszka.gutwinska@creatis.insa-lyon.fr) \n \nPurpose or Learning Objective: To evaluate the image quality of color K-\nedge imaging for contrast agents based on 8 differe nt elements using a clinical \nprototype spectral photon-counting CT (SPCCT). \nMethods or Background: A SPCCT with a field-of-view of 500mm was used \n(Philips; Israel). An anthropomorphic thoracic phan tom (QRM GmbH) with \ntwelve 1.5mL K-edge solutions (gadolinium-Gd, holmi um-Ho, ytterbium-Yb, \nhafnium-Hf, tantalum-Ta, tungsten-W, gold-Au, bismu th-Bi) ranging from 0 to \n2mg/mL was scanned at 120kVp and 50/75/150mAs. Five  acquisitions per \nagent were performed using dedicated energy thresho lds. Conventional \nimages in Hounsfield units and color K-edge images in mg/mL were \nreconstructed with isotropic voxels of 0.7mm3. Nois e, mean relative error \n(MRE) between prepared and measured concentrations,  signal-to-noise ratio \n(SNR) were measured on color K-edge images. Contras t-to-noise ratio (CNR) \non conventional and color K-edge images were measur ed and compared. \nResults or Findings: Mean noise ranged from 0.04 to 0.13mg/mL among all \nsamples with a lowest value for Gd at 75mAs (0.04±0 .01 mg/mL) and highest \nfor Ta at 50mAs (0.13±0.03mg/mL). Overall MRE was 2 9.7% with higher \naccuracy for Yb (e.g., 7.0% at 150mAs), and lower f or Ta (55.7% at 150mAs). \nSNR increased as function of concentrations with a factor per mg of 16.7, 14.6, \n10.9, 9.5, 9.3, 9.3, 8.8, 4.7 for Gd, Yb, Ho, Hf, A u, W, Ta and Bi, respectively, \nat 75mAs. CNR in color K-edge images increased as f unction of \nconcentrations, and were higher in comparison to CN R in conventional images \n(e.g., 1075%, 957%, 677%, 658%, 601%, 590%, 536%, 4 11%, for Gd, Au, W, \nTa, Yb, Ho, Bi and Hf, respectively, at 75mAs). \nConclusion: Image quality of color K-edge imaging in an anthrop omorphic \nthoracic phantom demonstrated high performances for  8 color K-edge agents \nusing SPCCT whilst outperforming sensitivity in com parison to conventional \nimaging. \n\n \n \nFriday \nAbstract-based Programme \n \n 140  \nLimitations: Phantom study. \nFunding for this study: The ERC starting Grant \"KOLOR SPCCT Imaging\" \n(N°101118079). \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nAntoine Robert: Nothing to disclose \nSalim Aymeric Si-Mohamed: Nothing to disclose \nDavid Peter Cormode: Nothing to disclose \nChristian Alejandro Hernandez-Fajardo: Nothing to d isclose \nSimon Rit: Nothing to disclose \nAgnieszka Joanna Gutwinska: Nothing to disclose \nDerick Rosario: Nothing to disclose \nRamata Coulibaly: Nothing to disclose \nAmanda Pang: Nothing to disclose \n \n \nTask-based image quality evaluation of ultra-high r esolution color K-edge \nimaging enabled by spectral photon-counting CT: a p hantom study \n*R. Coulibaly*¹, A. Robert¹, A. Houmeau¹, M. N. Ant onuccio², P. C. Douek¹,  \nS. Rit¹, J. Greffier³, S. A. Si-Mohamed¹; ¹Lyon/FR,  ²Paris/FR, ³Nimes/FR \n(ramata.coulibaly@creatis.insa-lyon.fr) \n \nPurpose or Learning Objective: To evaluate the image quality of color K-\nedge imaging with a gadolinium agent using spectral  photon-counting CT \n(SPCCT) in a phantom with a mixture of contrast age nts. \nMethods or Background: A clinical prototype SPCCT system (FOV 500mm, \nPhilips; Israel) was used to scan custom-made cylin drical phantom of 27cm \n(Color iQCT). Three inserts of the phantom were fil led up with agents as \nfollows: iodine only, mixture of iodine and gadolin ium, gadolinium only. Two \nconfigurations were considered, one with 0.5mg/mL o f each contrast agent and \nanother with 2mg/mL. For each configuration, two se ries of nine helical scans \n(120kVp) were acquired at 75mAs and 150mAs. Spectra l K-edge images of \ngadolinium were obtained by doing material decompos ition using 3 basis \n(water/iodine/gadolinium), using an iterative recon struction algorithm at 3 levels \n(iDose 0, 6, 11) were compared between inserts with  gadolinium using \niQMetrix-CT software. \nResults or Findings: Despite the presence of iodine, color K-edge imagin g \nenable specific differentiation of the gadolinium, showing a concentration \ndifference of 1.1% between inserts of gadolinium on ly and mixture (150mAs, \ni11, 2mg/mL). NPS peak was observed at the same spa tial frequency in all \nconfigurations (i.e., 0.053±0.018 mm⁻¹), whereas the noise magnitude \ndecreased when the dose and the iDose4 levels incre ased (-19.65±0.01% \nbetween iDose0 and iDose11). TTF values at 50% (f50 ) were similar between \ninserts with gadolinium only and the mixture, regar dless of the dose or iDose \nlevels (e.g., 0.213±0.090 mm⁻¹ vs 0.237±0.061 mm⁻¹, at 150mAs, i11, \n0.5mg/mL). f50 values were improved with increasing  dose, iDose levels, and \nwith higher concentrations. \nConclusion: Color K-edge imaging of a gadolinium contrast agent  using \nSPCCT demonstrated high spatial resolution and low noise magnitude at low \nconcentrations, even in the presence of iodine. \nLimitations: Phantom study. \nFunding for this study: ERC starting Grant \"KOLOR SPCCT Imaging\" \n(N°101118079). \nEthics committee - additional information: No ethics was required for this \nphantom study. \nAuthor Disclosures:  \nAntoine Robert: Nothing to disclose  \nSalim Aymeric Si-Mohamed: Nothing to disclose \nAngele Houmeau: Nothing to disclose \nPhilippe Charles Douek: Nothing to disclose \nSimon Rit: Nothing to disclose \nJoel Greffier: Nothing to disclose \nRamata Coulibaly: Nothing to disclose \nMaria Nicole Antonuccio: Employee: Philips \n \n \nX-ray phase contrast imaging, moving beyond traditi onal X-ray imaging \nmethods: a first pilot in intra-operative specimen imaging \n*G. Havariyoun*; London/UK \n(g.havariyoun@nhs.net) \n \nPurpose or Learning Objective: Several surgical procedures benefit from the \nability to image resected tissue samples in real ti me, e.g. to ensure no margin \ninvolvement. Micro-CT or tomosynthesis have great p otential, but suffer from \nlimited soft-tissue sensitivity of X-rays. X-ray ph ase contrast imaging (XPCI) \nprovides soft tissue sensitivity and increased cont rast through exploitation of \nphase effects. This work is presented on behalf of the UCL AXIm team. \n \n \n \n \nMethods or Background: XPCI was initially restricted to specialized facili ties \nsuch as synchrotrons. Our group has developed a met hod that has enabled \ncreation of a pre-commercial prototype compatible w ith surgical and radiology \nworkflows. This has been used to image >100 breast tissue samples from \nbreast conserving surgery both in vitro and in real  time. Images were \ncompared to standard specimen radiography and histo pathology. The system \nalso allows higher resolution (~10 micrometre) imag ing in slower scans for e.g. \ndigital histology. \nResults or Findings: System optimization (which also included size reduc tion) \nled to clinically acceptable scan times, which were  verified by trialling the \nsystem in a real intra-operative context. XPCI imag ing resulted to sensitivity \nand specificity values of 83% (95% CI 69-92%) and 8 3% (95% CI 70-92%), \nrespectively. Standard specimen radiography resulte d to sensitivity and \nspecificity values of 32% (95% CI 20-49%) and 86% ( 95% CI 73-93%), \nrespectively. \nConclusion: XPCI has a specificity comparable to standard speci men \nradiography but a significantly higher sensitivity.  This would lead to significant \nreduction in re-excision rates and in turn a reduct ion in patient stress, surgical \ntimes, healthcare costs and improved cosmetic outco mes. \nLimitations: Comparisons with standard specimen radiography were  made as \nthis is the most commonly used tool in the clinical  setting. Comparison with \nmore advanced techniques will be made in the future . \nFunding for this study: This work is funded by the Wellcome Trust (Grant \n200137/Z/15/Z). Alessandro Olivo (AXIm lead) is fun ded by the Royal \nAcademy of Engineering under their “Chairs in Emerg ing Technologies” \nscheme (CiET1819/2/78). \nEthics committee - additional information: The Breast Cancer Now Tissue \nBank (Approval No. 15/EE/0192) provided the ethical ly approved samples, the \nauthors thank the patients who have generously cons ented to donate their \ntissues which have been utilised in this work. \nAuthor Disclosures:  \nGlafkos Havariyoun: Nothing to disclose \n \n \nA workflow to harmonize CT abdomen protocols beyond  dose \nequalization \n*J. Vignero*, B. Miseur, J. Binst, H. Bosmans; Leuv en/BE \n(janne.vignero@uzleuven.be) \n \nPurpose or Learning Objective: A radiologist had raised concerns about \nexcessive noise in CT abdomen images of scanner A, while another scanner \n(B) of the same model, using similar protocol setti ngs, produced images of \nacceptable quality. This discrepancy led to a study  aimed at harmonizing \nprotocol settings to ensure consistent image qualit y across all patient sizes, \nmoving beyond standard dose equalization methods. C urrently, protocol \nadjustments—based on dose level, scan task, and tub e current modulation \nstrength—are often made intuitively. \nMethods or Background: Clinical CT scan data, collected through a dose \nmonitoring platform (DOSE, Qaelum), was analysed us ing water equivalent \ndiameter (WED), global noise level (GNL), kVp and C TDI. Three phantoms \nwith ellipsoid cross-sections (WED: 22, 33 and 43cm ) and iodine/calcium \ninserts were scanned to map protocol settings to sc an parameters (kVp, mAs, \nCTDI) and GNL for each scanner and phantom. The rec onstruction settings \nwere kept fixed. \nResults or Findings: Initially, scanner A and B produced similar radiati on \ndoses for the same WED groups, but scanner A had on  average a 13% higher \nGNL, with a maximum of 18% difference for the small est WED group. The \nradiologist could define GNL upper limits for each kVp. Using the phantom \nscan maps, protocol settings were chosen to achieve  the desired image \nquality. Post-optimization, 87% of scans met the GN L criteria, compared to \n70% before. To achieve this, radiation doses were i ncreased with 17%. \nConclusion: Identical scanner models and protocol settings do n ot guarantee \nconsistent image quality. Quality measures should b e included in optimization \nefforts. We present a procedure using new phantoms and dedicated metrics to \nimprove protocol harmonization. \nLimitations: The pipeline has only be verified on two scanners. \nFunding for this study: In part funded by the i-Violin project that is co-f unded \nunder the EU4Health Programme 2021-2027, grant agre ement no. 101056832. \nEthics committee - additional information: Retrospective, technical study \nAuthor Disclosures:  \nHilde Bosmans: Shareholder: Qaelum \nJanne Vignero: Nothing to disclose \nJoke Binst: Nothing to disclose \nBram Miseur: Nothing to disclose \n \n \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 141  \nImpact of Acquisition Parameters on Quantitative Im aging Using Rapid \nkVp-Switching Spectral CT \n*O. Sandvold*¹, A. Perkins², H. Daerr³, T. Koehler³ , R. Proksa¹,  \nR. Manjeshwar², P. Noël¹; ¹Philadelphia, PA/US, ²Cl eveland, OH/US, \n³Hamburg/DE \n \nPurpose or Learning Objective: To investigate the quantitative effects of \nvarying the ratio of high and low kVp tube voltage durations on spectral CT \nresults with a rapid kVp-switching X-ray tube. \nMethods or Background: Rapid kVp-switching offers excellent spectral \nseparation, but experimental research on acquisitio n parameters is limited. \nThis study addresses the gap. A rapid kVp-switching  X-ray tube (Philips \nHealthcare) on a spectral CT bench system was opera ted at 500 mAs \nalternating between 140 and 80 kVp. The total integ ration period (IP) \ncontaining high and low kVp IPs was 1 ms. We varied  the ratio of 140 kVp \nduration to total IP time from 0.15 to 0.85 and cor respondingly adjusted the low \nkVp IP time. A 3D-printed plastic phantom (20 cm di ameter) was rotated at 1 \nHz and contained four tissue-mimicking inserts: iod ine (2.0, 5.0 mg/ml), iodine \n4.0 mg/ml + human equivalent (HE) blood, and HE blo od (Sun Nuclear). We \nperformed reconstruction and two-material decomposi tion without applying \ndenoising. The noise and contrast-to-noise ratio (C NR) of the known \nconcentration inserts were measured in photoelectri c material images. Dose \nwas estimated from the reference detector. \nResults or Findings: Dose comparisons showed 0.15, 0.25, 0.33, and 0.67 \nratio scans used 47%, 51%, 63%, and 91% of the high est dose scan (0.85 \nratio). Measured noise values appeared to follow a quadratic trend as the ratio \nincreased, with 0.33 containing the lowest average noise. The 0.33 ratio image \ndose normalized CNR was approximately 1.24x, 1.14x,  1.11x, and 1.68x \ngreater than the dose normalized CNR in 0.15, 0.25,  0.67, and 0.85 ratio \nimages respectively. \nConclusion: Dose normalized CNR depends strongly on the ratio b etween \nhigh and low kV durations. This ratio should be con sidered for optimizing the \nspectral acquisition. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nHeiner Daerr: Employee: Philips Innovative Technolo gies \nThomas Koehler: Employee: Philips Innovative Techno logies \nOlivia Sandvold: Nothing to disclose \nPeter Noël: Nothing to disclose \nRoland Proksa: Nothing to disclose \nRavindra Manjeshwar: Employee: Philips Healthcare \nAmy Perkins: Employee: Philips Healthcare \n \n \n08:00-09:00 Research Stage 4 \nResearch Presentation Session: Breast \nRPS 1202 \nImproving the clinical impact of contrast-\nenhanced mammography \n \nModerator \nR. Alcántara; Barcelona/ES  \nAuthor Disclosures:  \nRodrigo Alcántara: Advisory Board: GE Healthcare; R esearch Grant/Support: \nGE Healthcare; Speaker: GE Healthcare, BD, Bayer \n \n \nCEM for the assessment of screening recalls: diagno stic performance at \nthree-year follow-up \n*S. Marziali*¹, A. Cozzi¹, M. Fanizza¹, V. Magni¹, L. Menicagli¹, A. Benedek¹,  \nG. Di Giulio², F. Sardanelli¹; ¹Milan/IT, ²Pavia/IT  \n(Sara.Marziali@unimi.it) \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of \ncontrast-enhanced mammography (CEM) for the assessm ent of screening \nrecalls. \nMethods or Background: Recalled women were prospectively enrolled at two \ncentres to undergo CEM alongside standard assessmen t (SA) through \nadditional views, tomosynthesis, and/or ultrasound between January 2019 and \nJuly 2021. Exclusion criteria were symptoms, implan ts, allergy to contrast \nagents, renal failure, and pregnancy. SA and CEM we re independently \nevaluated by one of six radiologists, who recommend ed biopsy or ≥3-year \nfollow-up. Diagnostic performance of CEM (low-energ y plus recombined \nimages) was calculated considering histopathology f or lesions biopsied and/or \nsurgically removed as well as ≥3-year follow-up for both breasts. \nResults or Findings: Between January 2019 and July 2021, 220 women were \nenrolled, 207 of them (median age 56.6 years) with 225 suspicious findings, \n135 of them referred for biopsy (4 by rCEM alone, 2 /4 being one DCIS and one \ninvasive carcinoma). During ≥3-year follow-up, 4 interval cancers were \nreported: one mucinous invasive carcinoma and one D CIS at the site of \nprevious assessment with biopsy; one DCIS in a diff erent quadrant of the same \nbreast; and one invasive carcinoma NST at the contr alateral breast (found at \nthe third year after CEM). No woman was lost at fol low-up. The overall CEM \nperformance was: sensitivity 80/84 (95.2%, 95% CI 8 8.3−98.7%); s pe c ific ity \n108/123 (87.8%, 95% CI 80.7−90.3%9). \nConclusion: The role of CEM in the assessment of recalls is con firmed at over \n3-year follow-up in terms of both sensitivity and s pecificity. The two cases of \ninterval cancers at the site of previous biopsy hig hlight the need of radiologic-\npathologic correlation. \nLimitations: Limited sample size. \nFunding for this study: GE Healthcare \nEthics committee - additional information: San Raffaele Hospital, Milan, \nItaly \nAuthor Disclosures:  \nAndrea Cozzi: Nothing to disclose \nSara Marziali: Nothing to disclose \nFrancesco Sardanelli: Speaker: Bayer AG, Siemens He althineers Advisory \nBoard: Bayer AG, Bracco imaging, GE healthcare Rese arch/Grant Support: \nBayer AG, Bracco imaging, GE healthcare, \nMarianna Fanizza: Nothing to disclose \nGiuseppe Di Giulio: Nothing to disclose \nVeronica Magni: Nothing to disclose \nAdrienn Benedek: Nothing to disclose \nLaura Menicagli: Nothing to disclose \n \n \nCOntrast enhanced Mammography in women with previou s Breast \ncancer Operated with conserving surgery (COMBO TRIA L): interim \nresults of a prospective intraindividual study \n*G. Vatteroni*¹, N. Turri¹, F. Fici¹, M. Filippini² , N. Basla³, G. Pinna¹,  \nG. Pruneddu¹, R. M. Trimboli¹, D. Bernardi¹; ¹Milan /IT, ²Brescia/IT, ³Pavia/IT \n(giulia.vatteroni@gmail.com) \n \nPurpose or Learning Objective: To present interim results from the 'COMBO \nTRIAL, a prospective intraindividual study evaluati ng the performance of \nContrast-Enhanced Mammography(CEM) vs Digital Mammo graphy(DM) for the \nsurveillance of women with a personal history of br east cancer(BC). \nMethods or Background: Between January 2023 and April 2024, women who \nunderwent breast-conserving surgery for BC were inv ited to undergo CEM for \nroutine surveillance. Exclusion criteria included: suspicious symptoms of BC, \nallergy to iodinated contrast agents, renal failure , breast implants. For each \npatient, one reader reported CEM while a second rea der, independent and \nblinded, evaluated only LE images equivalent to DM.  The reference standard \nwas 1-year follow-up for negative cases and biopsy/ surgery for BI-RADS 4/5. \nCancer Detection(CD) rate for both DM and CEM and i ncremental CD rate for \nCEM, sensitivity, specificity, PPV, NPV and accurac y were calculated. We \nevaluated differences in diagnostic performance bet ween DM and CEM using \nMcNemar test(p<0.05 significant). \nResults or Findings: Overall, 600 women were included in the analysis wi th a \nrecall rate of 9.8%. Among them, 14 cases of BC wer e detected: 8 (5 DCIS +3 \ninvasive) were identified by both DM and CEM, while  CEM detected 6 \nadditional cases (1 DCIS+5 invasive). Three cases w ere missed by both DM \nand CEM but subsequently detected by US, resulting in a global recurrence \nrate of 2.8%. The CD rate for CEM was 23 per 1000, compared to 13 per 1000 \nfor DM, indicating an incremental CD for CEM of 10 per 1000 (p=0.014). \nCompared to DM, CEM demonstrated significantly high er sensitivity(82.4% vs. \n47.1%), slightly lower specificity(96.4% vs. 97.6%) , slightly higher PPV(40.0% \nvs. 36.4%), slightly higher NPV(99.5% vs. 98.4%), a nd similar accuracy(96.0% \nvs. 96.2%). \nConclusion: Implementation of CEM in BC surveillance was associ ated with a \nsignificant increased detection of invasive cancers . \nLimitations: n/a \nFunding for this study: this study received research support by Siemens \nHelthineers \nEthics committee - additional information: The ethics commitee approved \nthis study \nAuthor Disclosures:  \nRubina Manuela Trimboli: Nothing to disclose \nFederica Fici: Nothing to disclose \nGiulia Pruneddu: Nothing to disclose \nDaniela Bernardi: Nothing to disclose \nNicolò Turri: Nothing to disclose \nGiulia Vatteroni: Nothing to disclose \nMarco Filippini: Nothing to disclose \nNicoletta Basla: Nothing to disclose \nGiulia Pinna: Nothing to disclose \n \n\n \n \nFriday \nAbstract-based Programme \n \n 142  \nEvaluation of lesion conspicuity on contrast-enhanc ed mammography \nimproves the performance in the assessment of malig nancy \n*M. Conti*, R. Rella, S. Palma, S. Amodeo, N. Di Ca taldo, D. Moretti,  \nM. Costantini, O. Tommasini, P. Belli; Rome/IT \n(conti.marco87@gmail.com) \n \nPurpose or Learning Objective: Aim of the study is to assess the \nperformance of Lesion Conspicuity (LC) in Contrast Enhanced Mammography \n(CEM) in the prediction of malignancy. \nMethods or Background: This is an observational retrospective study \ninvolving 153 women (median age, 44.1 years, IQR: 3 6-52) who underwent \nCEM and subsequent histological assessment at Polic linico Universitario A. \nGemelli IRCCS (April 2021-October 2023). Two radiol ogists (with 2 and 7 \nyears of experience in breast imaging, independentl y) evaluated low-energy \n(LE) images and LC (categorizing it as absent, low,  moderate or high) and \nassigned a BIRADS category of suspicion to the lesi on basing on both. \nDiagnostic performance of LE images and LC interpre tation together was \ncalculated using histological results of the biopsy  as gold standard. Subgroup \nanalyses based on mammographic breast density, back ground parenchymal \nenhancement (BPE) on CEM and type of lesions were a lso performed. \nResults or Findings: The interpretation of LE images together with the v alue \nof LC showed a sensitivity (SE) of 96.8% (95%CI: 92 .1%-99.1%) and a \nspecificity (SP) of 66.7% (95%CI: 46.0%-83.5%) vs. a SE=90.7% (95%CI: \n82.5%-95.9%) and a SP=52.6% (95%CI: 28.8%-75.5%) fo r LE images \nevaluation alone. Diagnostic performance of LE imag es + LC evaluation was \nbetter than LE images alone both in dense (SE=96.4%  and SP=89.8% vs. \nSE=93.3% and SP=60.0%) and adipose (SE=97.3% and SP =66.7% vs. \nSE=89.3% and SP=44.4%) breasts. Diagnostic performa nce of LE images + \nLC evaluation is better than LE images alone both i n minimal/mild and \nmoderate/marked BPE and in all types of lesions. \nConclusion: The evaluation of LE images joint with the value of  LC \ndemonstrated a better performance than LE images al one in predicting the \nmalignancy of lesions. \nLimitations: Limitations of the present study include its retros pective and \nmonocentric design. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by the \nlocal Institutional Review Board (ID: 6476) \nAuthor Disclosures:  \nNicola Di Cataldo: Nothing to disclose \nRossella Rella: Nothing to disclose \nOscar Tommasini: Nothing to disclose \nSimone Palma: Nothing to disclose \nMarco Conti: Nothing to disclose \nMelania Costantini: Nothing to disclose \nSilvia Amodeo: Nothing to disclose \nDelia Moretti: Nothing to disclose \nPaolo Belli: Nothing to disclose \n \n \nDiagnostic Accuracy of Contrast-Enhanced Mammograph y (CEM) in \nPreoperative Staging of Breast Tumors: A Comparativ e Study with \nHistology and Mammography \n*M. Balbino*¹, F. Masino², M. Montatore³, S. Surian o⁴, F. A. Carpagnano⁴,  \nG. Capuano³, G. Guglielmi⁵; ¹Triggiano/IT, ²Bari/IT, ³Barletta/IT, ⁴Foggia/IT, \n⁵Andria/IT \n(marinabalbino93@gmail.com) \n \nPurpose or Learning Objective: To evaluate the diagnostic accuracy of CEM \nin detecting and measuring breast tumors, comparing  it with histological \nfindings. \nMethods or Background: This study evaluated the diagnostic accuracy of \nContrast-Enhanced Spectral Mammography (CEM) in 69 breast cancer \npatients treated at the “Santa Maria” Hospital (Bar i, Italy) from January 2018 to \nSeptember 2023. Patients ranged from 33 to 86 years  old, with an average age \nof 55.8 years. All underwent CEM and subsequent bio psy or cytology for \nconfirmation. Exclusion criteria included deep lesi ons, post-biopsy hematomas, \nor renal insufficiency. CEM was performed using a G E Healthcare system with \ntwo post-injection projections: Cranio-Caudal and M edio-Lateral-Oblique. \nThree experienced radiologists analyzed the scans, comparing lesion sizes \nfrom mammography, CEM, and histology. \nResults or Findings: Results showed that CEM underestimated lesion size in \n28 cases, while mammography did so in 21. CEM ident ified 92.3% of multifocal \nmasses, compared to 58.3% with mammography. In hist ology, 49.27% of \npatients had mass-like neoformations, with 33.3% sh owing homogeneous \ncontrast enhancement. High contrast enhancement cor related with higher Ki67 \nproliferation indices (50-65%). The majority of neo plasms were Infiltrating \nDuctal Carcinomas. Statistical analysis revealed th at CEM was more accurate \nthan mammography, with the Wilcoxon signed-rank tes t showing no significant \ndifference between CEM and histological measurement s (p=0.9928). CEM  \n \nshowed high diagnostic accuracy in preoperative sta ging, supporting its use \nover MRI, particularly due to its lower cost, faste r acquisition time, and better \npatient tolerance. The study concludes CEM is a val uable, cost-effective \nalternative to MRI for assessing tumors before surg ery, particularly in \npreoperative staging and the identification of mult ifocal lesions. \nConclusion: CEM demonstrated high accuracy in assessing breast tumor size \nand extent, proving to be a valid alternative to MR I due to its lower cost, faster \nacquisition time, and better patient tolerability. \nLimitations: No \nFunding for this study: No \nEthics committee - additional information: No \nAuthor Disclosures:  \nMarina Balbino: Nothing to disclose \nGiuseppe Guglielmi: Nothing to disclose \nFrancesca Anna Carpagnano: Nothing to disclose \nSilvia Suriano: Nothing to disclose \nManuela Montatore: Nothing to disclose \nFederica Masino: Nothing to disclose \nGiulia Capuano: Nothing to disclose \n \n \nOptimizing microcalcification assessment: the role of contrast-enhanced \nmammography on reducing unnecessary biopsies \n*A. Santonocito*, T. H. Helbich, P. Clauser, P. A. Baltzer; Vienna/AT \n(santonocitoambra@gmail.com) \n \nPurpose or Learning Objective: Microcalcifications are frequently observed \nin screening mammography, with malignancy rates ran ging from 6% to 82%. \nTheir characterization and work-up are a major chal lenge for the radiologist. \nThis study aimed to assess the role of contrast-enh anced mammography \n(CEM) in managing microcalcifications. \nMethods or Background: This retrospective, single-centre, IRB-approved, \nstudy included consecutive patients underwent CEM f or BI-RADS 4 \nmicrocalcifications between October 2018 and Septem ber 2024. Patients \nwithout a standard of reference were excluded. The standard of reference was \nhistology from biopsy or surgery, or a one-year fol low-up for non-suspicious \ncases. Morphology, distribution, density, and densi ty heterogeneity of \nmicrocalcifications were assessed on LE images; enh ancement type and \nlesion conspicuity were assessed on RC images accor ding to the CEM lexicon. \nMicrocalcification and enhancement characteristics were analysed through \nSpearman correlation and chi-square tests to evalua te associations with \nhistological outcomes. \nResults or Findings: A total of 210 lesions in 197 patients (mean age 55  ±11 \nyears old) were analysed. Of these, 124 (63%) were benign and 72 (37%) \nwere malignant. Lesion conspicuity correlated with morphology (r=0.429), \npleomorphism (r=0.514), distribution (r=0.235), and  density heterogeneity \n(r=0.204), while no correlation was found with dens ity (r=-0.111). Chi-square \ntest showed significant differences between benign and malignant \ncalcifications for lesion type (p<0.001), morpholog y (p<0.001), grade of \npleomorphism (p<0.001), distribution (p<0.001), den sity (p<0.001), \nenhancement type (p<0.001), lesion conspicuity (p<0 .001) and grade of \ndensity heterogeneity (p=0.02). No significant diff erences were found for size \n(p>0.05). \nConclusion: Our findings suggest that CEM may help differentiat e between \nbenign and malignant microcalcifications, potential ly reducing unnecessary \nbiopsies. \nLimitations: Retrospective study; small cohort \nFunding for this study: None \nEthics committee - additional information: Number: 1391/2022 \nAuthor Disclosures:  \nPascal A.T. Baltzer: Nothing to disclose \nThomas H. Helbich: Nothing to disclose  \nAmbra Santonocito: Nothing to disclose \nPaola Clauser: Nothing to disclose \n \n \nContrast Enhanced Mammography (CEM) in the manageme nt of locally \nadvanced breast cancer receiving neoadjuvant therap y (NAT) \n*S. Vidali*, F. Di Naro, D. De Benedetto, G. Bicchi erai, E. Vanzi, C. Bellini,  \nC. Boeri, V. Miele, J. Nori; Florence/IT \n(sofia.vidali@yahoo.it) \n \nPurpose or Learning Objective: Among contrast-enhanced mammography \n(CEM)'s indication, is the evaluation of locally ad vanced breast cancer(LABC)'s \nresponse to neoadjuvant therapy (NAT). CEM has pote ntial not only in the \nassessment of radiologic complete response (rCR) bu t also in predicting \ntumour response based on biological features and en hancement patterns. \nMethods or Background: We retrospectively analysed post-NAT CEMs of \n141 patients diagnosed with LABC between 2016-2024 and correlated \nenhancement patterns (rCR, residual enhancement, RE ) with diagnostic biopsy \nbiologic features and post-surgical pathology data (pCR, residual disease RD). \n\n \n \nFriday \nAbstract-based Programme \n \n 143  \nResults or Findings: CEM showed RE in 92 patients (68 with RD and 24 wit h \npCR) was negative in 49 (17 with RD and 32 with pCR ), with resulting mean \nsensitivity of 80% and specificity of 57%. Reclassi fying DCISs as RD instead of \npCR, specificity raised to 65%: these performance d ata are adequate and \ncomparable to those reached by CE-MRI, the gold sta ndard. Based on \nbiological subtypes, distinct enhancement patterns resulted coherent with the \nexpected therapeutic responses: luminal A tumours ( 46 patients) were less \nresponsive to NAT (11% pCR, 5 patients, 4 of them w ith rCR, 80%), while \nHER2+ (37 patients) and triple negative(TN)BCs (20 patients), the more \naggressive forms were more likely to achieve pCR an d rCR: respectively 59% \npCR (22 patients) in the HER2+ group of which 64% ( 14 patients) with rCR, \n50% pCR in the TNBCs (10 patients) of which 60% (6 patients) with rCR. \nLuminal B lesions (38 patients) achieved pCR in 45%  of cases (17 patients) of \nwhich 13 with rCR (76%). \nConclusion: CEM is adequate and reliable in assessing disease r esponse to \nNAT: enhancement patterns demonstrated correlation with biological features, \nforecasting CEM’s potential as a prognostic and man agement tool for \nincreasing conservative therapies and diagnostic fo llow-up. \nLimitations: Sample's numerosity and heterogeneity. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Regione Toscana, Comitato \nEtico Area Vasta Centro, reference number: SPE_16.2 51 \nAuthor Disclosures:  \nGiulia Bicchierai: Nothing to disclose \nCecilia Boeri: Nothing to disclose \nJacopo Nori: Nothing to disclose \nSofia Vidali: Nothing to disclose \nVittorio Miele: Nothing to disclose \nChiara Bellini: Nothing to disclose \nErmanno Vanzi: Nothing to disclose \nDiego De Benedetto: Nothing to disclose \nFederica Di Naro: Nothing to disclose \n \n \nContrast Enhanced Mammography Screening in Women wi th Dense \nBreasts \n*J. M. Net*, J. Spoont, S. Stamler, C. Pluguez-Turu ll, N. Brofman,  \nA. Hamedi-Sangsari, M. Yepes; Miami, FL/US \n(josenet@me.com) \n \nPurpose or Learning Objective: To determine outcomes of screening \nContrast Enhanced Mammography (CEM) in women with B I-RADS density 3 or \n4, more specifically to determine outcomes of CEM i n this population who did \nnot undergo screening breast ultrasound. \nMethods or Background: HIPAA compliant and IRB approved retrospective \nstudy evaluated all medical records of patients who  underwent CEM at our \ninstitution between 8/2019-8/2024. Reports were rev iewed and scored for \nbreast density, additional work up (additional imag ing, US, MRI), final BI-\nRADS, biopsy results and whether target represented  CEM finding, if high risk \nlesion or malignant case was scored based on final surgical pathology. Final \nresults were tallied to include overall CDR, PPV3, and interval cancer rate. \nResults or Findings: 1215 CEM studies were performed between 2020-2024, \n1079 were performed for screening of which 740 were  assigned BIRADS \ndensity 3 or 4. 87 patients (11.7%) were referred f or additional work up \nincluding US in 75, and MRI in 12. Of these, 46 wer e referred for biopsy which \nconfirmed malignancy in 16 patients translating int o a CDR of 21.6/1000 and \nPPV3 of 35.5%. Of the remaining 653 patients (88.2% ) without additional work \nup, 46 underwent annual MRI within 12 months of CEM  confirming the only 2 \ninterval cancers in this study (0.3%), one stage 0 DCIS/linear 1.3 cm NME on \nstaggered 6 month MRI and the other a 1.2 cm TNBC ( T1Nitc) on staggered 6 \nmonth MRI - neither was seen in retrospect. \nConclusion: The low interval cancer rate combined with high CDR  suggest \nthat CEM screening in women with dense breasts can potentially replace \nsupplemental breast US for adjunct screening. \nLimitations: Limitations include modest sample size, retrospecti ve design, and \ninclusion of cases from a single institution which limits generalizability. \nFunding for this study: None \nEthics committee - additional information: IRB approved study. \nAuthor Disclosures:  \nAntoine Hamedi-Sangsari: Nothing to disclose \nJose Miguel Net: Nothing to disclose \nMonica Yepes: Nothing to disclose \nNicole Brofman: Nothing to disclose \nSarah Stamler: Nothing to disclose \nCedric Pluguez-Turull: Nothing to disclose \nJamie Spoont: Nothing to disclose \n \n \n \n09:30-11:00 Research Stage 1 \nResearch Presentation Session: Vascular \nRPS 1315 \nImaging of the aorta, pulmonary, and \ncoronary arteries \n \nModerator \nM. Cejna; Feldkirch/AT  \n(manfred.cejna@lkhf.at) \n \n \nLow-energy virtual monochromatic CT with deep-learn ing image \nreconstruction to improve detection of endoleak \n*T. Higashigawa*¹, Y. Ichikawa¹, K. Nakajima², T. K obayashi¹, K. Domae¹,  \nA. Yamazaki¹, N. Kato¹, H. Sakuma¹; ¹Tsu/JP, ²Ise/J P \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of \nlow-energy virtual monochromatic CT imaging (VMI) c ombined with deep-\nlearning image reconstruction (DLIR) for the detect ion of endoleaks. \nMethods or Background: A cohort of 71 patients after endovascular aortic \nrepair who underwent dynamic contrast-enhanced CT b etween March 2022 \nand August 2023 were studied. Raw data were reconst ructed using three \ndifferent methods: 70-keV VMI using conventional hy brid iterative \nreconstruction (HIR [ASiR-V50%]), and 40- and 70-ke V VMI using DLIR \n(TrueFidelity-H). Contrast-to-noise ratio (CNR) of the endoleaks on venous \nphase CT was calculated. Three observers assessed t he presence or absence \nof endoleak on a 5-point scale, taking into account  the confidence level: score-\n1, endoleaks are definitely not present; score-2, p robably not present; score-3, \nmay be present; score-4, probably present; score-5,  definitely present. A score \nof 3 or higher was considered positive for endoleak . \nResults or Findings: Endoleaks were observed in 41 (58%) of 71 subjects.  \nThe CNRs of endoleaks were significantly higher in 40-keV DLIR (17.1±9.8) \ncompared to 70-keV HIR (6.4±3.8; P<0.001) and 70-ke V DLIR (10.2±6.2; \nP<0.001). ROC analysis for endoleak detection showe d that AUC for 40-keV \nDLIR (0.92-0.99) was the largest for all observers (70-keV DLIR, 0.91-0.97; 70-\nkeV HIR, 0.88-0.96). The percentage of patients wit h endoleaks who were \ncorrectly identified with a confidence level of ≥ score-4 in 40-keV VMI with \nDLIR was significantly higher compared to those in 70-keV VMI with HIR in one \nobserver (Observer1, 85%(35/41) vs 73%(30/41), P=0. 02; Observer2, \n85%(35/41) vs 78%(32/41), P=0.20; Observer3, 98%(40 /41) vs 90%(37/41), \nP=0.10, respectively). \nConclusion: The 40-keV VMI combined with DLIR reconstruction me thod \nimproves the CNR of endoleaks and may help to corre ctly identify endoleaks \nwith higher confidence compared to 70-keV VMI with HIR. \nLimitations: The limitations of the study is the relatively smal l study \npopulation. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by \ninstitutional review board (approval number; H2019- 207). \nAuthor Disclosures:  \nKensuke Domae: Nothing to disclose \nTakatoshi Higashigawa: Nothing to disclose \nHajime Sakuma: Nothing to disclose \nYasutaka Ichikawa: Nothing to disclose \nNoriyuki Kato: Nothing to disclose \nTatsuhiro Kobayashi: Nothing to disclose \nAkio Yamazaki: Nothing to disclose \nKen Nakajima: Nothing to disclose \n \n \nImage Quality Improvement of Ultra-low Dose CT Pulm onary \nAngiography Using Deep Learning Reconstruction Algo rithm: Two-center \nProspective Study \n*J. Lu*, L. Shen, Z. Zhao, Z. Bi, M. Zeng, M. M. Wa ng; Shanghai/CN \n(lujinjuan1016@163.com) \n \nPurpose or Learning Objective: To investigate the effects of deep learning \nreconstruction (DLR) on the image quality in ultra- low dose CT pulmonary \nangiography (CTPA), compared to hybrid iterative re construction (HIR) at \nroutine dose. \nMethods or Background: This study prospectively included 130 patients with  \nsuspected pulmonary embolism (PE) who underwent CTP A examination in two \nhospitals from April to July 2024. The noise index of routine dose (RD) group \nand ultra-low dose (ULD) group was set to 10 and 20 , respectively. The CT \nimages of RD group were reconstructed using HIR, wh ile ULD group were \nreconstructed with HIR and DLR. Pulmonary CT value,  signal-to-noise ratio \n\n \n \nFriday \nAbstract-based Programme \n \n 144  \n(SNR), and contrast-to-noise ratio (CNR) were asses sed as quantitative criteria \nof image quality. Two senior radiologists independe ntly evaluated the overall \nimage noise, pulmonary artery visibility, and diagn ostic confidence based on a \n5‑point Likert scale (5, best; 1, worst). The Mann-Wh itney U test and the \nWilcoxon signed rank test was used for statistical analysis. \nResults or Findings: No statistically significant difference was found i n the \nclinical data between the two groups (p>0.05). The ULD-DLR group exhibited \nhigher SNRs and CNRs in all seven pulmonary arterie s compared to the RD-\nHIR group (both p<0.05). The overall image noise an d diagnostic confidence of \nthe ULD-DLR images were significantly better than t hat in the RD-HIR images \nand ULD-HIR images (both p<0.001). The effective do se in the RD group and \nULD group were 2.84±0.49mSv and 0.70±0.21mSv, respectively, representing \na reduction of approximately 75% in the ULD group ( p<0.001). \nConclusion: DLR can significantly reduce the radiation dose of CTPA \nexamination without compromising the diagnosis of P E. Even at ultra-low \nradiation dose, its image quality is still better t han HIR at routine dose. \nLimitations: Not applicable. \nFunding for this study: No. \nEthics committee - additional information: Shanghai Geriatrics Medical \nCenter Ethics Committee (B2024-009) \nAuthor Disclosures:  \nMengsu Zeng: Nothing to disclose \nZicheng Zhao: Nothing to disclose \nMingliang Mingliang Wang: Nothing to disclose \nLeilei Shen: Nothing to disclose \nZhenghong Bi: Nothing to disclose \nJinjuan Lu: Nothing to disclose \n \n \nPatient tailored contrast volume for preoperative C T angiography of the \naorta: a prospective study based on patient heart r ate and body surface \narea \n*M. Dewilde*, W. Coudyzer, A. Laenen, H. Bosmans, G . Maleux; Leuven/BE \n(miloud.dewilde@gmail.com) \n \nPurpose or Learning Objective: To prospectively compare aortic image \nquality by adapting contrast volume and kiloVoltage  (kV) in patients referred for \npreoperative aortic computed tomography (aCT). \nMethods or Background: Eighty prospectively included patients were \nassigned into 3 groups: 50% of the contrast dose ca lculated on body surface \narea (BSA) and heart rate (HR) (group 1, n=56); 50%  of the contrast dose \ncalculated on BSA and HR and additional kV reductio n (group 2, n=11); 50% of \ncontrast dose calculated on BSA and HR and addition al contrast dilution 80% \ncontrast & 20% NaCl (group 3, n=13). Quantitative ( measurement of \nHounsfield units) analysis at different anatomical aortic levels and qualitative \nimage analysis by 2 radiologists using a visual sco re (1 = inadequate; 5 = \nexcellent) was performed. \nResults or Findings: Mean contrast dose injected was 46.1 ml, 28.3 ml an d \n35.0 ml for group 1, 2 and 3 respectively, with a s ignificant difference between \ngroup 1 and 2 (P=<0.001) and between group 1 and 3 (P=<0.001); no \ndifference between group 2 and 3 (P=0.072). Mean qu alitative scores were \n4.35/5, 2.82/5 and 3.46/5 for group 1, 2 and 3 resp ectively. No patient needed \nrepeat imaging for inadequate aortic CT-imaging. In terobserver agreement was \nmoderate for group 1 and 3 (0.577 and 0.576 resp.) and fair for group 2 (0.282) \nwith consistent difference in scoring. \nConclusion: Meaningful contrast dose reduction in preoperative aCT while \nmaintaining diagnostic efficacy is feasible through  utilization of a contrast \ninjection algorithm incorporating patient’s HR and BSA, coupled with adjusting \nkV values. \nLimitations: Limited number of patients (n=80). Only patients wi th a pre \noperative angio CT were included. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The institutional ethics committee \napproved this study (S58042). \nAuthor Disclosures:  \nWalter Coudyzer: Nothing to disclose \nAnnouschka Laenen: Nothing to disclose \nHilde Bosmans: Nothing to disclose \nMiloud Dewilde: Nothing to disclose \nGeert Maleux: Nothing to disclose \n \n \n \n \n \n \n \n \n \nAI-based reconstruction algorithm applied to low-kV  and low contrast \nmedium volume CT for TAVI planning compared with lo w dose CT with \nModel-Based algorithm: image quality and radiation dose exposure \n*C. R. G. L. O. M. Talei Franzesi*, D. Ippolito, C.  Maino, P. N. Franco,  \nD. G. Gandola, R. Corso; Milan/IT \n(ctfdoc@hotmail.com) \n \nPurpose or Learning Objective: To evaluate image quality and radiation \ndose reduction of deep learning reconstruction algo rithm in CT angiography \n(CTA) studies performed for TAVI planning, compared  with low dose CTA \nreconstructed with hybrid iterative algorithm \nMethods or Background: Fifty six patients candidates for TAVI were enrolle d \nin this study and 26 patients (study-group) were ex amined with 128 MDCT \nscanner, with 80 kV, automated mAs dose-modulation and 50 mL of contrast \nmedia (CM), combined with a new deep learning recon struction algorithm \n(Precise Image); while a control group of 32 patien ts were evaluated with 256 \nMDCT (100 KV; automated mAs; 50 mL of CM) reconstru cted with hybrid \niterative reconstruction algorithm (iDose4). Subjec tive (using a 4-point Likert \nscale) and objective image quality (vascular enhanc ement, SNR and CNR in \ndifferent aortic levels and in the iliac arteries) were evaluated and the radiation \ndose exposure of both groups (CTDIvol and DLP) was calculated \nResults or Findings: Study group with deep learning algorithm demonstrat ed \nsignificantly higher mean attenuation values (p<.05 ) in all the measurements \ncompared to the control group with model based algo rithm (aortic root 621HU \nvs 314 HU; external iliac arteries 537HU vs 335HU).  Mean DLP and CTDI of \nstudy group was significantly lower than in control  group (DLP: 395 mGy*cm vs \n1600 mGy*cm, p<0.001; CTDI: 8.03 mGy vs 23.5 mGy, p <0.001), with an \noverall radiation dose reduction of about 75%. Furt hermore, study group \nshowed a significant decrease of image noise with a n increase of image quality \nConclusion: Deep learning based CT reconstruction algorithm com bined with \nlow Kv setting allows to significantly reduce radia tion dose exposure and \nincrease the image quality in CTA protocol for TAVI  planning, in comparison \nwith low dose CTA reconstructed with hybrid iterati ve algorithm \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nDavide Giacomo Gandola: Nothing to disclose \nCesare Maino: Nothing to disclose \nCammillo Roberto Giovanni Leopoldo Oreste Massimili ano Talei Franzesi: \nNothing to disclose \nRocco Corso: Nothing to disclose \nPaolo Niccolò Franco: Nothing to disclose \nDavide Ippolito: Nothing to disclose \n \n \nDetermining elasticity of the thoracic aorta in pat ients with giant cell \narteritis using non-contrast-enhanced magnetic reso nance imaging at 1.5 \nT \n*M. Both*, C. Jochum, J. H. Schirmer, E. A. Strathm ann, P. Langguth,  \nS. Sandra Freitag-Wolf, C. Von Der Burchard, O. Jan sen, M. Salehi Ravesh; \nKiel/DE \n(mboth@rad.uni-kiel.de) \n \nPurpose or Learning Objective: The application of imaging techniques for \nearly detection of thoracic aortic aneurysms in pat ients with giant cell arteritis \n(GCA), including the identification and monitoring of subgroups at high risk for \nthis condition, is still the subject of debate. We investigated whether aortic \nstiffness could be quantified based on MRI and used  as a potential biomarker \nfor post-inflammatory damage. \nMethods or Background: Ten GCA patients in clinical remission and 36 \nhealthy volunteers (HVs) were examined using non-co ntrast-enhanced cine-\nbalanced steady-state free precession (bSSFP) MRI t echnique to determine \nthe distensibility and diameter of the ascending (A Ao), descending (DAo), and \narch (AArch) segments of the thoracic aorta. In add ition, changes in aortic \ndiameters during follow-up in GCA patients and the impact of demographic and \nclinical characteristics on the aortic elasticity w ere investigated. \nResults or Findings: Distensibility was significantly higher in the AArc h \n(p=0.039) and in the DAo (p=0.004) than in the AAo in HVs, but not in GCA \npatients. Aortic distensibility was significantly l ower in patients than in HVs in \nthe AArch (0.89 vs. 2.15, p=0.035). Age was an addi tional predictor of aortic \nstiffening in the AAo (p=0.029) and DAo (p=0.001) o f HVs. In patients with \nGCA, the diameter increased at an above-average rat e in all aortic segments \n(AAo 1.04 mm/year, AArch 1.12 mm/year, DAo 0.95 mm/ year) compared to \nbaseline MRI. \nConclusion: The bSSFP MRI technique revealed functional and str uctural \ndifferences in the thoracic aorta of patients with GCA as a potential marker for \nweakness of the thoracic aortic wall. \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 145  \nLimitations: The small size of our patient group is the main lim itation due to its \nsingle-center design. Another limitation relates to  the fact that most of our \nstudy patients presented with predominantly cranial  symptoms, some without \nproof of aortitis on MRI. \nFunding for this study: None \nEthics committee - additional information: The Ethics Committee at the \nFaculty of Medicine of Kiel University approved thi s study (No. D577/18). \nAuthor Disclosures:  \nClaus Von Der Burchard: Nothing to disclose \nMona Salehi Ravesh: Nothing to disclose \nPatrick Langguth: Nothing to disclose \nMarcus Both: Nothing to disclose \nChiara Jochum: Nothing to disclose \nJan Henrik Schirmer: Nothing to disclose \nSandra Sandra Freitag-Wolf: Nothing to disclose \nOlav Jansen: Nothing to disclose \nEike Andreas Strathmann: Nothing to disclose \n \n \nAccelerating Coronay CT Angiographies via Improved Patient \nPreparation \n*A. M. C. Boehner*, B. Salam, A. Jacob, A. Isaak, C . C. Pieper, D. Kütting; \nBonn/DE \n(boehner.amc@gmail.com) \n \nPurpose or Learning Objective: Coronary Computed Tomography \nAngiography (CCTA) often requires extensive prepara tion, contributing to \nprolonged in-room time. This study aims to assess t he impact of pre-\nexamination preparation, including the administrati on of IV beta-blockers and \nECG lead placement outside the examination room, on  reducing in-room time \nfor coronary CT scans. \nMethods or Background: A prospective study with 139 patients was \nconducted, comparing standard in-room preparation ( control cohort) with \nreceiving preparation outside (intervention cohort)  and mostly omitted \npreparation, scanned via a spiral acquisition proto col (spiral cohort). Patients' \nheart rates were regulated in the intervention grou p before entering the \nexamination room. Key measures included: patient en tering the examination-\nroom, installation of patient monitoring, heart rat e adjustment, first scan, heart \nrate during scanning and image quality. \nResults or Findings: The intervention cohort demonstrated significantly \n(P<0.0001) reduced in-room time compared to the con trol cohort (984±347s \nvs. 704±308s). The spiral cohort performed best and  displayed the lowest \nvariability (583±103s, P<0.0001). The heart rates of the spiral cohort was \nhighest with 69±21bpm (P<0.04), but the interventio n cohort did not differ from \nthe control cohort (60±7bpm vs. 59±6bpm, P=0.88). The rate of non-diagnostic \nsegments remained low across all groups (control: 3 .5%, intervention: 1.5%, \nspiral: 2.0%). \nConclusion: Pre-examination preparation outside of the examinat ion room, \nincluding installation of patient monitoring and ad ministration of IV beta-\nblockers, significantly reduces in-room time for CC TA without compromising \nimage quality. Alternatively, a spiral image acquis ition protocol allows for the \nomission of most preparatory steps. Both approaches  offer a feasible strategy \nto streamline workflow and enhance efficiency in ca rdiac imaging departments. \nLimitations: The limitation regarding the spiral cohort is the n eed to preselect \npatients with a coronary-calcium-score <400. Some r adiological departments \nmay lack the premises to implement our approach. \nFunding for this study: The study was conducted in collaboration with \nSiemens Healthineers. \nEthics committee - additional information: Administrational \nAuthor Disclosures:  \nAlexander Isaak: Nothing to disclose \nAlexander Marc Christian Boehner: Nothing to disclo se \nAlice Jacob: Nothing to disclose \nDaniel Kütting: Nothing to disclose \nClaus Christian Pieper: Nothing to disclose \nBabak Salam: Nothing to disclose \n \n \nOptimizing HU Thresholds for Accurate Calcium Scori ng in Contrast-\nEnhanced CT: Robust Alternatives to the Agatston Sc ore \n*L. D. Grünewald*, V. Koch, S. Mahmoudi, J. Gotta, P. Reschke, J-E. Scholtz, \nS. Martin, C. Booz, T. Vogl; Frankfurt/DE \n \nPurpose or Learning Objective: To approximate the Agatston score in \ncontrast-enhanced CTs using a volumetric approach w ithout distortions from \ncontrast agents. \nMethods or Background: The aorta of 1276 patients (886 men, 390 women; \nmedian age 67 years; interquartile range 57-76) wit hout prior surgical \ninterventions who underwent contrast-enhanced multi -phase CT between \nJanuary 2018 and December 2023 were retrospectively  analyzed. For all  \n \npatients, the Agatston score was derived from unenh anced CT scans for the \nthoracic and abdominal aorta. The number and volume  of plaques were \nassessed for the thoracic and abdominal aorta in ar terial and venous contrast \nphases using thresholds ranging from 100 to 1000 to  assess the influence of \ncontrast agents. Correlations with the Agatston sco re were calculated, and \nlinear regression was used to identify the optimal threshold. \nResults or Findings: Median aortic enhancement was 46 HU (unenhanced), \n323 HU (arterial), and 120 HU (venous). In venous p hases, a threshold of 300 \nHU yielded the highest correlations with the Agatst on score (thoracic: r=0.91; \nabdominal: r=0.93; p<0.001). In arterial phases, a threshold of 900 HU \nprovided the best correlation (thoracic: r=0.72; ab dominal: r=0.65; p<0.001). \nLinear regression confirmed these thresholds, but s howed only moderate \npredictive power (R²=0.66 for venous, R²=0.52 for a rterial phases). Dynamic \nthresholding resulted in poor correlation (r=0.26; p<0.001) and low predictive \nvalue (R²=0.07). \nConclusion: Plaque volume assessment using optimized HU thresho lds can \nreliably approximate the Agatston score in contrast -enhanced CTs, offering a \nrobust assessment without contrast-induced distorti on. This approach is \nparticularly valuable when non-contrast images are unavailable, such as in \nstaging or pre-TAVR evaluations. \nLimitations: Patient collective with high plaque burden. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Waiver due to retrospective \nnature \nAuthor Disclosures:  \nSimon Martin: Nothing to disclose \nChristian Booz: Nothing to disclose \nThomas Vogl: Nothing to disclose \nJan-Erik Scholtz: Nothing to disclose \nVitali Koch: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nPhilipp Reschke: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \nJennifer Gotta: Nothing to disclose \n \n \nPerformances of spectral CT for the detection and c haracterization of \ncommunications between the true and the false lumen  in aortic \ndissections \nA. Janin-Manificat¹, M. Sigovan¹, L. Boussel¹, A. M illon¹, P. C. Douek¹,  \n*S. Boccalini*²; ¹Lyon/FR, ²Villeurbanne/FR \n(sara.boccalini@yahoo.com) \n \nPurpose or Learning Objective: To assess the performance of conventional \nCT (conv-CT) and spectral CT (spectral-CT) for the detection and \ncharacterization of communications between the true  (TL) and false lumen (FL) \nin aortic dissections, using 4D-flow MRI as the ref erence. \nMethods or Background: 18 patients with type A and B aortic dissection who  \nunderwent 4D-flow MRI, conv-CT and spectral-CT were  included. For each \npatient, the exams closest in time, without any sur gical or endovascular \nintervention in-between were retrieved and subjecti vely analysed by two \nobservers, independently for conv-CT and in consens us for MRI and spectral-\nCT. Communications between the two lumens were iden tified as: intimal tears \non conv-CT; focal alterations in velocities corresp onding to jet flows on 4D-\nflow; both intimal tears and focal changes of contr ast concentration \ncorresponding to jet flows on spectral-CT. The numb er, size, and location of \ncommunications were noted. Additionally, the direct ion of the flow was \nassessed for spectral-CT and MRI. \nResults or Findings: Of the 176 communications detected with 4D-flow, \nspectral-CT allowed visualisation of 122 (69%) comp ared to 58 (33%) for Obs1 \nand 38 (22%) for Obs2 for conv-CT, yielding an accu racy twice as high (63% \nvs. 29-30%). On spectral CT, in only 45 cases (26%)  the size of the \ncommunications could be assessed, in all other case s only jet flows were \ndetected without visible intimal tears. The flow wa s unidirectional TL-FL in 2 \ncases for both MRI and spectral-CT and bidirectiona l in 5 and 3 cases for the \ntwo modalities. In all other cases the flow was in the direction TL-FL. \nConclusion: Spectral-CT outperformed conv-CT for the detection of \ncommunications between TL and FL in aortic dissecti ons. Spectral-CT allows \ndirect visualization of flow jets, and their direction, through intimal tears. \nLimitations: Low number of patients; time in-between different e xams \nFunding for this study: No \nEthics committee - additional information: Approved \nAuthor Disclosures:  \nAntoine Janin-Manificat: Nothing to disclose \nPhilippe Charles Douek: Speaker: Philips \nLoïc Boussel: Speaker: Philips \nSara Boccalini: Speaker: Philips \nMonica Sigovan: Nothing to disclose \nAntoine Millon: Nothing to disclose \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 146  \nComparison of artificial intelligence and inexperie nced physicians in \npulmonary embolism detection at deep learning recon struction-based \nultra-low radiation dose CT pulmonary angiography \n*J. Lu*, L. Shen, Z. Zhao, Z. Bi, M. Zeng, M. M. Wa ng; Shanghai/CN \n(lujinjuan1016@163.com) \n \nPurpose or Learning Objective: To assess the performance of artificial \nintelligence (AI) software and inexperienced physic ians in diagnosing \npulmonary embolism (PE) at deep learning reconstruc tion-based ultra-low \ndose (ULD) CT pulmonary angiography (CTPA). \nMethods or Background: This prospective two-center study contained 210 \npatients with suspected pulmonary embolism (PE) who  underwent CTPA \nexamination, randomizing into two groups with equal  proportion of patients. \nImages in the routine-dose (RD) group were reconstr ucted using hybrid \niterative reconstruction (HIR, AIDR 3D, FC08), whil e ULD images were \nreconstructed using HIR and deep learning reconstru ction (DLR, AiCE), \nrespectively. A subset of 74 participants (1:1 PE t o non-PE ratio) was randomly \nselected and evaluated by two inexperienced physici ans and AI software \n(Discover PE, uAI). Reference standard was establis hed by expert consensus. \nThe diagnostic accuracy (sensitivity and specificit y) of the AI or reader \ninterpretations were compared between methods by bo otstrapping. \nResults or Findings: There was no statistically significant difference i n the \npatient demographics between two groups. ULD-DLR im ages exhibited \nsignificantly higher objective and subjective image  quality compared to both \nRD-HIR and ULD-HIR images. The AI software exhibite d near-perfect accuracy \nin both ULD-HIR and ULD-DLR sets (sensitivity: 97.3 0 %, specificity: 100 %). \nIn comparison, two physicians showed a mean sensiti vity of 75.68% and \nspecificity of 93.75% in ULD-HIR sets, and a mean s ensitivity of 94.59% and \nspecificity of 100.00% in ULD-DLR sets. Inter-obser ver agreement was \nmoderate for HIR (κ = 0.75) and good for DLR (κ = 0.81). The effective dose of \nULD group was significantly lower than the RD group  (2.74±0.47 mSv vs. \n0.73±0.25 mSv, p<0.001). \nConclusion: DLR can significantly reduce the radiation dose of CTPA \nexamination without compromising the diagnosis of p ulmonary embolism even \nat ultra-low radiation dose. AI software outperform s inexperienced physicians \nin interpreting ULD images. \nLimitations: Not applicable. \nFunding for this study: No. \nEthics committee - additional information: Shanghai Geriatrics Medical \nCenter Ethics Committee (B2024-009) \nAuthor Disclosures:  \nMengsu Zeng: Nothing to disclose \nZicheng Zhao: Nothing to disclose \nMingliang Mingliang Wang: Nothing to disclose \nLeilei Shen: Nothing to disclose \nZhenghong Bi: Nothing to disclose \nJinjuan Lu: Nothing to disclose \n \n \n09:30-11:00 Research Stage 2 \nResearch Presentation Session: \nInterventional Radiology \nRPS 1309 \nInterventions in malignant liver disease \n \nModerator \nL. Novosel; Zagreb/HR  \n(novosel_luka@hotmail.com) \n \n \nOncologic Ablation in Germany: 2018-2023 data from the German Society \nof Interventional Radiology Registry \n*J. Uhlig*¹, L. Biggemann¹, J. Nadjiri², T. Kroenck e³; ¹Göttingen/DE, \n²Munich/DE, ³Augsburg/DE \n(johannes.uhlig@med.uni-goettingen.de) \n \nPurpose or Learning Objective: To assess the current utilization, technical \napproaches and complications of oncologic ablation in Germany. \nMethods or Background: The German Society of Interventional Radiology \n(“DeGIR”) registry was queried for patients receivi ng ablation treatments \nbetween 2018-2023. Patient demographics, indication s, and technical ablation \nparameters were descriptively assessed. \nResults or Findings: N=9157 patients receiving oncologic ablation were \nincluded (34.3% female; median age 67yo). Between 2 018-2023, annual \nablation number remained approximately constant at 1000 cases/year. \nAblations were performed in the liver (71.4%), kidn ey (11.3%), musculoskeletal \nsystem (7.3%) and lung (5.4%), mainly with curative  intent (64%) or for \nsymptomatic treatment / palliation (30.6%). N=7371 patients were imaged with \nCT before ablation (80.5%), 4176 with MRI (45.6%), and 173 with PET (1.9%; \nnot mutually exclusive). Ablation guidance was achi eved using CT (89.6%), \nultrasound (5.7%), MRI (2.8%), cone-beam CT or fluo roscopy (0.9%, each). \nAblation procedures were mainly performed under gen eral anesthesia (74.6%) \nor analgosedation (14.5%). Microwave ablation was p erformed in most cases \n(69.4%), followed by radiofrequency (23.6%) and cry oablation (2.8%), often \ncombined with tract ablation (51.2%). Only 48 proce dures (0.5%) were \npreemptively terminated, mainly due to anatomical d ifficulties (n=16) or \nuncooperative patients (n=11). During or within the  first 24h after ablation, \n4.8% of patients experienced any complications, the  majority being low-grade. \nAnother 45 patients (0.5%) experienced delayed comp lications 24h or later \nafter ablation, mainly infections/abscesses (n=21).  \nConclusion: Oncologic ablations are routinely performed in Germ any with low \nprocedural complication rates, mostly using CT-guid ed microwave or \nradiofrequency ablation for hepatic or renal tumors . \nLimitations: Since participation in the DeGIR registry is not ma ndatory, there \ncould be selection bias of included cases and parti cipating sites, limiting the \ngeneralizability of results. \nFunding for this study: Not applicable. \nEthics committee - additional information: Not applicable - retrospective \nanonymised registry data. \nAuthor Disclosures:  \nThomas Kroencke: Nothing to disclose \nJohannes Uhlig: Nothing to disclose \nLorenz Biggemann: Nothing to disclose \nJonathan Nadjiri: Nothing to disclose \n \n \nEnhancing Neoadjuvant Immunotherapy Efficacy throug h Partial \nCryoablation in a Hepatocellular Carcinoma Mouse Mo del \n*T. Kao*¹, E. Meister¹, J. Santana², J. Israel², A. Shewarega², J. Tefera²,  \nD. C. Madoff², L. J. Savic¹, J. Chapiro²; ¹Berlin/D E, ²New Haven, CT/US \n(Tabea.kao@charite.de) \n \nPurpose or Learning Objective: Hepatocellular carcinoma (HCC) exhibits an \nimmunosuppressive microenvironment which can be agg ravated by incomplete \ntumor ablation. Immune checkpoint inhibitors (ICIs)  such as anti-PD-1 are \nguideline-approved therapies for advanced HCC. Comb ining ablation with ICIs \ncould potentially strengthen anti-cancer immunity, but supporting evidence is \nlimited. We aim to evaluate the effect of neoadjuva nt systemic anti-PD-1 on the \nlocal immune response in residual tumors following partial cryoablation in a \nTIB-75 murine HCC model. \nMethods or Background: Forty-eight male and female BALB/c mice aged 6-\n12 weeks underwent orthotopic inoculation of TIB-75  cells to induce a solitary \nHCC lesion. After 7 days, mice were randomized into  4 treatment groups: (a) \ncontrol, (b) anti-PD-1, (c) partial cryoablation, a nd (d) anti-PD-1 followed by \npartial cryoablation. The percentage of positively stained T-cell subsets and \ntumor-associated macrophages within the tumor was a ssessed in paraffinized \nliver tissue samples using immunohistochemistry (CD 3+, CD4+, CD8+, CD68+, \nCD206+, FOXP3+) and quantified on digitized slides.  Treatment groups were \ncompared using unpaired Mann-Whitney U and Kruskal- Wallis test with Dunn \ncorrection. \nResults or Findings: Mice treated with anti-PD-1 (n=12, group b) showed \ngreater tumoral infiltration of CD3+, CD4+ and CD8+  T-cells than control \n(CD3+: mean 21.4% vs. 6.7%; P=<0.0001, CD4+: mean 2 1.3% vs. 6.0%; \nP=<0.0001, CD8+: mean 7.5% vs. 3.8%; P=0.005). Part ial cryoablation alone \n(n=12) had greater infiltration of CD206+ M2-like m acrophages than control \n(mean 32.4% vs. 14.6%; P=0.007). Anti-PD-1 combined  with partial \ncryoablation (n=12) showed significantly more infil tration of CD3+ T-cells \n(mean 13.7% vs. 6.1%; P=0.002) and fewer CD206+ M2- like macrophages \n(mean 26.1% vs. 32.4%; P=0.3474) than partial cryoa blation alone (n=12). \nConclusion: Immune evasion following partial cryoablation can b e \ncounteracted with neoadjuvant anti-PD-1, suggesting  effective combination \ntherapy to treat both early-stage and advanced-stag e HCC. \nLimitations: The model was inoculated in healthy, non-cirrhotic mouse liver. \nFunding for this study: NIH grant 2R01CA206180 \nEthics committee - additional information: All experimental procedures \nwere approved by the Yale University Institutional Animal Care and Use \nCommittee (IACUC protocol number: 2022-20262). \nAuthor Disclosures:  \nJoshua Israel: Nothing to disclose \nJulius Chapiro: Nothing to disclose \nJonathan Tefera: Nothing to disclose \nJessica Santana: Nothing to disclose \nAnnabella Shewarega: Nothing to disclose \nDavid Craig Madoff: Nothing to disclose \nEllen Meister: Nothing to disclose \nTabea Kao: Nothing to disclose \nLynn Jeanette Savic: Nothing to disclose \n\n \n \nFriday \nAbstract-based Programme \n \n 147  \nMRI-based risk stratification for viable Hepatocell ular Carcinomas post-\nTransarterial Chemoembolization: Correlation with p athological \noutcomes and prognostic implications \n*W. Wang*, Y-C. Wang; Nanjing/CN \n(weilang_wang@163.com) \n \nPurpose or Learning Objective: Accurate risk stratification of viable \nhepatocellular carcinomas (HCC) following transarte rial chemoembolization \n(TACE) is essential for the development of individu alized treatment strategies \nand enhancing the accuracy of prognosis predictions . \nMethods or Background: This multi-center, retrospective study includes HCC  \npatients who received TACE as their initial and sol e treatment from February \n2015 to October 2022 as training set (203 viable tu mors). Additionally, a \ndataset from a multicenter clinical trial (NCT03113 955) was subject to \nsecondary analysis as test set (102 viable tumors).  The final pathological \nvalidation set consists of a separate center, inclu ding individuals who had liver \nresection post-first TACE (120 viable tumors). All participants in both the \ntraining and test cohorts underwent contrast-enhanc ed MRI scans at baseline, \nand at one and six months after TACE. In the traini ng set, univariate and \nmultivariate logistic regression analysis was perfo rmed to identify clinical, \nlaboratory and imaging variables to include in the predictive model. \nResults or Findings: The predictive model incorporated five key imaging \nfeatures: Mild-moderate T2 hyperintensity, T2-weigh ted peritumoral \nhyperintensity, Diffusion restriction, Irregular sh ape, and Heterogeneity. The \nmodel achieved areas under the curve (AUCs) of 0.85  (95% confidence \ninterval [CI] 0.79 to 0.90) for the training cohort  and 0.88 (95% CI 0.81 to 0.95) \nfor the external test cohort. The risk model effect ively distinguished high-risk \nfrom low-risk groups in the test cohort, with signi ficant differences in \nprogression-free survival (PFS) (P = 0.004) and two -year overall survival (OS) \n(P = 0.028). In the pathology cohort, the model cor related with microvascular \ninvasion (MVI) grades (P = 0.003) and liver capsule  invasion (P = 0.007). \nConclusion: This risk model based on imaging features for viabl e HCCs post-\nTACE exhibits robust predictive power for tumor via bility at six months and for \nlong-term survival outcomes. \nLimitations: Not applicable \nFunding for this study: This study has received funding by National Natural  \nScience Foundation of China (NSFC, No. 82271978, 92 359304, 82330060) \nand Zhongda Hospital Affiliated to Southeast Univer sity, Jiangsu Province \nHigh-Level Hospital Pairing Assistance Construction  Funds (No. zdyyxy09). \nEthics committee - additional information: This multicenter, retrospective \nstudy was reviewed and approved by IEC for clinical  research of the Zhongda \nHospital, Southeast University, approval number [20 22ZDSYLL410-P01], and \nconducted following the ethical principles outlined  in the Helsinki Declaration of \n1964 and its subsequent amendments, or other ethica l standards with \nequivalent requirements. All patients and their fam ilies signed informed \nconsent forms prior to surgery. \nAuthor Disclosures:  \nYuan-Cheng Wang: Nothing to disclose \nWeilang Wang: Nothing to disclose \n \n \nTransarterial Embolization Alone Versus Drug-Elutin g Beads \nChemoembolization for HepatocellularCarcinoma (RAD- 18-TAcE):  \na Randomized Clinical Trial \n*M. Taninokuchi Tomassoni*¹, M. Renzulli¹, S. Zanel la¹, A. Doriguzzi Breatta², \nP. Marra³, F. De Cobelli⁴, C. Mosconi¹; ¹Bologna/IT, ²Turin/IT, ³Bergamo/IT,  \n⁴Milan/IT \n(makoto.taninokuchi@studio.unibo.it) \n \nPurpose or Learning Objective: This randomized clinical trial aims to \ncompare transarterial embolization (TAE)and drug-el uted beads transarterial \nchemoembolization (DEB-TACE) in the treatment of he patocellular carcinoma \n(HCC). \nMethods or Background: Patients diagnosed with unresectable HCC were \nrandomly assigned to either theTAE or DEB-TACE grou p. The primary \nendpoint was time to progression (TTP), and seconda ry endpoints included \noverall survival, cost-effectiveness, tumor respons e rates, and adverse events. \nResults or Findings: A total of 111 patients were enrolled, with 56 in t he TAE \ngroup and55 in the DEB-TACE group. Baseline charact eristics were balanced \nbetween the two groups. The primary endpoint analys is showed that TAE was \nnot different from DEB-TACE in termsof TTP (average  of 12.13 and 10.87 \nmonths respectively, p=0.432). Overall survival, tu morresponse rates, and \nadverse events were also similar between the two gr oups. The cost-\neffectiveness ratio of DEB-TACE vs. TAE was evaluat ed considering that, with \nequaleffectiveness of the two treatments highlighte d by the previous points, \nthere being nostatistically significant difference in terms of days of \nhospitalization between DEB-TACEand TAE (average of  4.62 days and 5.20 \ndays respectively, p=0.638). \nConclusion: In this randomized clinical trial, TAE showed compa rable \noutcomes to DEB-TACE in the treatment of unresectab le hepatocellular \ncarcinoma. These findings suggest that TAE could be  considered as an \nalternative for treating HCC with no differences in  terms of safety and efficacy. \nLimitations: The relatively small sample size and short-term fol low-up period \nmay limit the generalizability of the findings and the ability to detect subtle \ndifferences in outcomes between TAE and DEB-TACE. \nFunding for this study: This study was funded by the Italian Ministry of \nHealth. \nEthics committee - additional information: The trial was conducted in \naccordance with ethical standards and received appr oval from the institutional \nreview boards. The study was registered at www.clin icaltrials.gov \n(NCT04803019). \nAuthor Disclosures:  \nAndrea Doriguzzi Breatta: Nothing to disclose \nPaolo Marra: Nothing to disclose \nMatteo Renzulli: Nothing to disclose \nCristina Mosconi: Nothing to disclose \nSara Zanella: Nothing to disclose \nMakoto Taninokuchi Tomassoni: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \n \n \nMR guided catheter-based radiotherapy/brachytherapy  of liver tumours – \nfirst experience and feasibility \n*M. P. Fabritius*, A. Haghpanah, O. Dietrich, D. Pu hr-Westerheide,  \nV. F. Schmidt, S. Corradini, J. Ricke, O. Öcal, M. Seidensticker; Munich/DE \n \nPurpose or Learning Objective: To show feasibility and safety of MR guided \ncatheter-based radiotherapy/brachytherapy of primar y or secondary liver \ntumours \nMethods or Background: Between June 2023 and April 2024, 27 patients \nwith 54 liver lesions were treated within a prospec itve single-center trial on MR-\nguided catheter-based radiotherapy (MR BRIGHT trial ). Treatments were \nperformed under conscious sedation and local anesth esia using a 1.5T MRI \nsystem (Magnetom Solafit, Siemens) with a 15 cm loo p coil. Gadoxetic acid \n(0.1 mmol/kg) was administered for contrast enhance ment, followed by \ninsertion of an 18G coaxial needle and navigation t o the lesion via real-time \ngradient-echo fluoroscopy sequences (iMRI UI Extens ion, Research Software \nPackage). The needle was exchanged for a 6F hydroph ilic angiography sheath \nwith a brachytherapy catheter. 3D T1-weighted seque nces were sent to the \nradiation department for brachytherapy with an IR19 2 high-dose-rate (HDR) \nafterloading unit. Target doses ranged from 15 to 2 5 Gy, depending on tumor \ntype (HCC, CRC, GIST, NET, and other metastases). C atheters were removed \nafter BT, and the radiation tract sealed with gelat in sponge. \nResults or Findings: The average lesion diameter was 13 ± 6 mm, whereas \nthe average clinical target volume (CTV) was 3.0 ± 2.9 cm3. The average room \ntime was 74 ± 35 minutes, the average time for catheter placement was 19 ± \n11 minutes. The mean dose administered per lesion ( D100) was 18.9 ± 3.6 Gy. \nComplications during and after BT were generally ra re with only 2 patients (7.4 \n%.) having a minor bleeding without need for blood transfusion or intervention. \nConclusion: Overall, MR-guided catheter-based radiotherapy for liver tumours \nis feasible and safe, particularly for small lesion s. With low complication rates \nand precise dosimetry achieved through advanced ima ging, this approach \nholds promise for effective tumour management. \nLimitations: n/a \nFunding for this study: None \nEthics committee - additional information: LMU Munich \nAuthor Disclosures:  \nStefanie Corradini: Nothing to disclose \nOlaf Dietrich: Nothing to disclose \nMatthias Philipp Fabritius: Nothing to disclose \nAlireza Haghpanah: Nothing to disclose \nVanessa Franziska Schmidt: Nothing to disclose \nMax Seidensticker: Nothing to disclose \nDaniel Puhr-Westerheide: Nothing to disclose \nOsman Öcal: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nEvaluation of Pain and Satisfaction in Patients wit h Liver Tumor treated \nwith CT guided High Dose Rate Brachytherapy under A nalgosedation– \nPreliminary Results \n*M. Z. Erforth*, L. K. Segger, U. Fehrenbach, F. Co llettini, B. Gebauer,  \nT. A. Auer; Berlin/DE \n(mo-zelda.erforth@charite.de) \n \nPurpose or Learning Objective: To evaluate feasibility for CT guided high \ndose rate (HDR) brachytherapy under analgosedation performed by \ninterventional radiologists in patients with liver tumors. \nMethods or Background: In this prospective single-center study (EA/122/23) , \n97 patients who received CT-guided HDR brachytherap y along with \nanalgosedation using fentanyl and midazolam were en rolled and 77 were \nincluded in the final analysis. At the outset, a pe rsonality profile (from the \nEORTC catalog) related to their pain experience was  also recorded for each \npatient. Structured questionnaires were employed to  assess the patients' pain \n\n \n \nFriday \nAbstract-based Programme \n \n 148  \nlevels and satisfaction both at and after the inter vention. Three months later, \nthe patients were recontacted, and a follow-up surv ey was conducted. The \nresults were recorded by means of a numeric analog scale and presented as \ncategorical variables. \nResults or Findings: First, pain was measured (0: no pain; to 10: maximu m \npain) 1. at catheter placement; 2. at the radiation ; 3. after the radiation \n(catheter removal). At catheter placement, 75.5% (5 8/77) rated the pain as low \n(0-2), 18.0% (14/77) as moderate (3-6), and 6.5% (5 /77) as severe (7-10). At \nthe radiation and afterwards pain levels were rated  as low in 83.0% (64/77) \nand 79.0% (61/77), as moderate in 11.5% (9/77) and 17.0% (13/77) and as \nsevere in 5.5% (4/77) and 4.0% (3/77), respectively . Second, patient \nsatisfaction was measured (1: completely dissatisfi ed; to 10: completely \nsatisfied). In 1.3% (1/77) the lowest score was rec orded while in 98.7% (76/77) \na score ≥7 was recorded. \nConclusion: CT guided HDR brachytherapy under analgosedation is  feasible \nand can be performed by interventional radiologists  themselves without \ngeneral anesthesia. \nLimitations: Limitations include the short observation period an d the small \ncohort of patients with a heterogeneous clinical hi story, neoplasm histology, \nand location, as well as inhomogeneity regarding pr evious treatments. \nFunding for this study: None \nEthics committee - additional information: Institutional Review Board \napproved prospective study (EA1/122/23). \nAuthor Disclosures:  \nLaura Katharina Segger: Nothing to disclose \nMo Zelda Erforth: Nothing to disclose \nUli Fehrenbach: Nothing to disclose \nBernhard Gebauer: Nothing to disclose \nTimo Alexander Auer: Nothing to disclose \nFederico Collettini: Nothing to disclose \n \n \nDual-phase Cone-Beam CT (DP-CBCT) role as imaging n avigation \nguidance in HCC lesions treatment with trans-arteri al chemoembolization \n(TACE): a single centre experience \n*N. Rossini*¹, C. Floridi¹, M. Macchini¹, L. M. Cac ioppa¹, A. Felicioli¹,  \nC. Mincarelli², R. Candelari¹, A. Giovagnoni¹; ¹Anc ona/IT, ²Macerata/IT \n(nicolorossini44@gmail.com) \n \nPurpose or Learning Objective: To evaluate how intra-procedural DP-CBCT \nnavigation guidance influences TACE success rate in  terms of residual disease \nin follow-up imaging. \nMethods or Background: This retrospective analysis includes all patients w ith \nHCC treated with TACE (cTACE or DEB-TACE) in our ce ntre between January \n2017 and January 2024 with at least 1 month of imag ing (CT or MRI) follow-up \navailable. All patients had a recent baseline CT or  MRI before TACE. Patients \nwere divided in two groups, the first one included patients with DP-CBCT \nperformed intra-procedurally during TACE (DP-CBCT g roup), the second \nincluded patients with no CBCT performed during tre atment (no-DP-CBCT \ngroup). The two groups were similar in vascular ana tomy, lesions number, \nmorphology and localization. Response to treatment was evaluated in imaging \nfollow-up with mRECIST criteria. The two groups wer e compared for treatment \nresponse after TACE in terms of residual disease in  follow-up imaging. \nResults or Findings: 152 patients were included in the study (M:F 112:40 ). 82 \npatients were included in the DP-CBCT group whereas  70 in the no-DP-CBCT \ngroup. Residual disease was of 26.9% in DP-CBCT gro up and of 63.0% in no-\nDP-CBCT group. A significant difference in terms of  residual disease was \nobserved between the two groups (p<0.05). Significa nt lower cases of residual \ndisease were present in DP-CBCT group. \nConclusion: DP-CBCT imaging navigation guidance improves signif icantly \nsuccess rate in TACE, ensuring a better visualizati on of HCC lesions feeding \nvessels and a consequent higher possibility of comp lete treatment of the \nnodules without residual disease. \nLimitations: The main limitation of this study is the brief foll ow up imaging that \nshould be extended in future studies. \nFunding for this study: This research received no external funding. \nEthics committee - additional information: All the procedures performed in \nstudies involving human participants were in accord ance with the ethical \nstandards of the institutional and/or national rese arch committee and with the \n1964 Declaration of Helsinki and its later amendmen ts or comparable ethical \nstandards. This study obtained the approval of the Internal Review Board (IRB) \nof University Politecnica Delle Marche. \nAuthor Disclosures:  \nCinzia Mincarelli: Nothing to disclose \nAlessandro Felicioli: Nothing to disclose \nChiara Floridi: Nothing to disclose \nLaura Maria Cacioppa: Nothing to disclose \nMarco Macchini: Nothing to disclose \nRoberto Candelari: Nothing to disclose \nAndrea Giovagnoni: Nothing to disclose \nNicolo' Rossini: Nothing to disclose \n \nIdentifying Key Predictors of Mortality and Liver D ecompensation in \nHepatocellular Carcinoma Patients Treated with Yttr ium-90 \nRadioembolization \nM. Arabi, H. Alghamdi, *A. A. F. Almesned*, O. Alan azi, M. Alghamdi,  \nM. Bukhaytan, M. Alkhalaf, M. Almaimoni, N. Alagraf y; Riyadh/SA \n(azizmesned@gmail.com) \n \nPurpose or Learning Objective: This study aimed to identify the predictors of \nmortality and liver decompensation in patients with  HCC treated with Y-90 \nradioembolization. \nMethods or Background: A retrospective analysis of 140 patients with HCC \nwho underwent Y-90 radioembolization was conducted.  Kaplan‒Meier and \nmultivariate Cox regression analyses were performed  to identify the significant \npredictors of mortality. \nResults or Findings: The cohort comprised 69.3% males with a mean age of  \n71.3 ±11.9 years. Most patients (73.6%) had Child-P ugh class A cirrhosis and \n34.3% had BCLC stage B disease. Among the 140 patie nts, 57.1% died after \ntreatment and liver decompensation was recorded in 39.2%. The median \nsurvival was significantly longer in those without liver decompensation (3.2 vs \n0.7 years, p<0.001). Multivariate analysis revealed  that male sex (adjusted \nodds ratio [aOR] 5.889, p=0.009), cirrhosis (aOR 6. 82, p=0.047), and \ninternational normalized ratio (INR) (aOR 316.664, p=0.013) were independent \npredictors of liver decompensation. Cox regression analysis revealed several \nsignificant predictors of mortality. Ascites (HR 2. 012, 95% CI, 1.122–3.61; \np=0.019), portal vein invasion (HR 1.695, 95% CI, 1 .057–2.718; p=0.029), and \ndiabetes mellitus (HR 1.823, 95% CI, 1.017–3.265; p =0.044) were associated \nwith increased mortality risk. Conversely, non-mult ifocal HCC (HR 0.593, 95% \nCI, 0.369–0.955; p=0.031), treatment of the liver l obe other than the right lobe \n(HR, 0.482; 95% CI 0.236–0.986, p=0.046), and age ≥60 years (HR 0.288, \n95% CI, 0.139–0.597; p=0.001) were associated with a reduced risk of \nmortality. \nConclusion: This study identified the key predictors of mortali ty in patients \nwith HCC undergoing Y-90 radioembolization, potenti ally improving patient \nselection and management strategies. \nLimitations: While this study provides valuable insights, severa l limitations \nshould be acknowledged. The retrospective nature of  the study introduces \npotential biases in patient selection and data coll ection. The lack of post-\ninfusion dosimetry limits the precision of dose-res ponse analyses. \nFunding for this study: The study was not supported by funding. \nEthics committee - additional information: The study was approved by the \ninstitutional review board, and the need for inform ed consent was waived. This \nstudy was conducted in accordance with the 2010 gui delines of the Declaration \nof Helsinki. \nAuthor Disclosures:  \nOmar Alanazi: Nothing to disclose \nMuath Almaimoni: Nothing to disclose \nMohammed Bukhaytan: Nothing to disclose \nMohammed Alkhalaf: Nothing to disclose \nMeshari Alghamdi: Nothing to disclose \nHamdan Alghamdi: Nothing to disclose \nAbdulaziz Abdullah F Almesned: Nothing to disclose \nNawaf Alagrafy: Nothing to disclose \nMohammad Arabi: Nothing to disclose \n \n \nApplication of cross-modality image registration sy stem for localising \nintraoperative colorectal cancer liver metastases d uring ablation \n*X. Wu*; Hangzhou/CN \n(Wuxia1981@zju.edu.cn) \n \nPurpose or Learning Objective: Localisation of target tumours under CT \nguidance can be challenging due to insufficient sof t tissue resolution and metal \nartifacts. This study aims to validate the accuracy  of the automatic image \nregistration system (AIRS) in localising target les ions throughout the CT-guided \nCRLM ablation procedure, thereby exploring a novel guidance method for \ninterventional procedures. \nMethods or Background: This retrospective, single-center study included \npatients with CRLM who underwent CT-guided liver ab lation between January \n2021 and August 2023. Three experienced physicians collectively annotated \nthe visibility and lesion centre positions of CRLMs  on both the preprocedural \ncontrast-enhanced MRI and intraoperative CT image s eries, which served as \nthe ground truth. The AIRS and two junior physician s delineated the lesion \ncentre positions in the same CT sequences. The loca lisation errors of the AIRS \nand junior physicians were analysed using the non-p arametric Kruskal–Wallis \ntest for one-way analysis. \nResults or Findings: One hundred and twenty consecutive patients with 22 4 \nCRLMs treated across 128 sessions were enrolled. Th ere were 128 pairs of \nMR-pCT (pre-procedural CT) multi-modal registration s and 1,008 pairs of pCT-\niCT (intra-procedural) mono-modal registrations. AI RS demonstrated superior \nlocalisation error than the physician group in loca lising lesions suboptimal \nvisible on pCT (5.94±2.61 mm vs 8.04±5.32 mm, p=0.006), lesions excellently \n\n \n \nFriday \nAbstract-based Programme \n \n 149  \nvisible on iCT (5.14±2.65 mm vs 6.15±3.84 mm, p=0.01) and lesions \nsuboptimal visible on iCT (6.13±2.80 mm vs 8.94±4.60 mm, p＜0.001). \nConclusion: Compared with less experienced physicians, an AIRS can quickly \nand accurately locate target lesions in CT-guided c olorectal cancer liver \nmetastasis ablation procedures, especially for lesi ons with poor visibility, thus \npaving the way for a new navigation method in color ectal cancer liver \nmetastasis ablations. \nLimitations: This was a retrospective investigation with a relat ively small \nsample size, which may have restricted the generali sability of our findings. \nFunding for this study: None \nEthics committee - additional information: This single-centre retrospective \nstudy was approved by the local ethics committee an d was exempted from \ninformed consent. \nAuthor Disclosures:  \nXia Wu: Nothing to disclose \n \n \nDeep Learning-Based Reconstruction and Superresolut ion for MR-guided \nThermoablation \n*M. T. Winkelmann*¹, J. Kuebler¹, S. Gassenmaier¹, D. Nickel², K. Nikolaou¹,  \nS. Afat¹, R. Hoffmann¹; ¹Tuebingen/DE, ²Erlangen/DE  \n(moritz.winkelmann@med.uni-tuebingen.de) \n \nPurpose or Learning Objective: This study explores the impact of deep \nlearning-enhanced image generation for T1-weighted volume-interpolated \nbreath-hold examinations (DL-VIBE) on image quality  and procedural \nparameters during MR-guided thermoablation of liver  malignancies, compared \nto standard VIBE images (SD-VIBE). \nMethods or Background: 34 consecutive patients (mean age: 65.4 ± 11.5 \nyears, women: n=13) with liver malignancies underwe nt MR-guided microwave \nablation using a 1.5 T MR scanner. Intraprocedural VIBE sequences (SD-\nVIBE) were used to monitor needle position and asse ss the ablation zone. The \nraw T1-weighted VIBE data were retrospectively proc essed with a deep \nlearning algorithm (DL-VIBE) to reduce noise and im prove sharpness. Two \ninterventional radiologists independently evaluated  the image sets in a blinded \nmanner, comparing DL-VIBE with unprocessed SD-VIBE images. Criteria \nassessed included diagnostic confidence, image qual ity, noise, artifacts, and \nsharpness. Interrater agreement was analyzed, and n oise maps were created \nto evaluate signal-to-noise ratio improvements. \nResults or Findings: DL-VIBE significantly improved overall image qualit y, \nreduced noise and artifacts, and enhanced the sharp ness of liver contours and \nportal vein branches compared to SD-VIBE (P<0.001).  Additionally, DL-VIBE \nimproved imaging of the interventional path, needle  tip detectability, and \ndiagnostic confidence in needle positioning and the  ablation zone (P<0.001), \nwith high interrater agreement (κ = 0.86). Quantitative noise maps \ndemonstrated a higher signal-to-noise ratio, and th e reconstruction process \ntook approximately 4 seconds, reducing breath-hold time by 2 seconds. \nConclusion: DL-VIBE significantly enhances image quality and di agnostic \nconfidence during MR-guided thermal ablation proced ures, offering time \nsavings and potential improvements in patient outco mes. \nLimitations: - Small number of patients. - Sequences were retros pectively \nprocessed, not used during actual interventions. - Image quality assessments \nwere retrospective and may differ from real-time ev aluations. - Time savings \nare theoretical due to the retrospective study desi gn, with uncertainty about \nfeasibility during actual interventions. \nFunding for this study: No funding was received for this study \nEthics committee - additional information: This retrospective study was \napproved by the institutional review board (Eberhar d Karls University of \nTübingen, project number: 055/2017BO2) \nAuthor Disclosures:  \nKonstantin Nikolaou: Nothing to disclose \nMoritz T. Winkelmann: Nothing to disclose \nSaif Afat: Nothing to disclose \nRüdiger Hoffmann: Nothing to disclose \nDominik Nickel: Nothing to disclose \nJens Kuebler: Nothing to disclose \nSebastian Gassenmaier: Nothing to disclose \n \n \nMachine learning using MR imaging radiomics and cli nical features can \npredict the response of large hepatocellular carcin oma to transarterial \nradioembolization \nO. Sarioğlu, A. Canturk, *R. C. Yarol*, H. Gulmez, E. Derebe k, A. Gülcü; \nIzmir/TR \n(raifyarol@gmail.com) \n \nPurpose or Learning Objective: To evaluate the potential of machine \nlearning-based models for predicting the response o f large hepatocellular \ncarcinoma to transarterial radioembolization \nMethods or Background: A total of 49 patients (38 responder and 11 non-\nresponder) were included in the study. Laboratory r esults and clinical \nconditions were collected. Treatment response was a ssessed according to \nmRECIST criteria from the 3-month follow-up MR exam inations. Complete or \npartial response was categorized as the responder g roup, while stable or \nprogressive disease was classified as the non-respo nder group. Radiomics \nfeatures were extracted from contrast-enhanced T1-w eighted images (CE-T1) \nand T2-weighted images (T2WI). 141 radiomics featur es were obtained from \neach lesion. Classification learning models were us ed to create prediction \nmodels for TARE response. 5-fold cross-validation t echnique was utilized to \nidentify the prediction rates of treatment response . \nResults or Findings: Number of radiomics features demonstrated statistic ally \nsignificant differences between the groups are 9 an d 12 on T2W and CE-T1 \nimages, respectively. The model based on radiomics features obtained from \nCE-T1 images demonstrated an accuracy rate of %79.6  to predict response \nwith an AUC of 0.92. The sensitivity and specificit y rates were %79 and %100, \nrespectively. The accuracy and AUC rates of the mod el using radiomics \nfeatures extracted from T2W images were %79.6 and 0 .77, respectively. \nSensitivity and specificity rates of the model were  %80 and %67, respectively. \nWhen only clinical and laboratory parameters were u sed, the model showed an \naccuracy rate of %77.6 and an AUC of 0.65. The sens itivity and specificity \nvalues of the clinical and laboratory model were %7 9 and %50, respectively. \nAnother model using both clinical and CE-T1 radiomi cs features showed an \naccuracy rate of %73.5 \nConclusion: Machine learning-based radiomics models based on MR I can \npredict the response of large hepatocellular carcin oma to transarterial \nradioembolization \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Ethics committee approval \nobtained \nAuthor Disclosures:  \nAytaç Gülcü: Nothing to disclose \nAli Canturk: Nothing to disclose  \nRaif Can Yarol: Nothing to disclose  \nOrkun Sarioğlu: Nothing to disclose \nHakan Gulmez: Nothing to disclose \nErkan Derebek: Nothing to disclose \n \n \n09:30-11:00 Research Stage 3 \nResearch Presentation Session: Cardiac \nRPS 1303 \nThe evolving impact of artificial \nintelligence (AI) in cardiac imaging \n \nModerator \nA. Isaak; Bonn/DE  \n \n \nFormulation of a predictive model for total cardiac  volume (TCV) \nestimation: Optimizing donor-recipient size matchin g and outcomes \n*S. Gowda*, V. Raj, R. Kothari; Bengaluru/IN \n(srjgwd@gmail.com) \n \nPurpose or Learning Objective: Accurate donor heart size measurement is \ncrucial for successful heart transplantation (HT). Traditional weight-based \ndonor-to-recipient (D-R) size matching in paediatri c HT has poor correlation \nwith cardiac size and significantly restricts the d onor pool. The aim of the study \nis to develop a novel predictive model to accuratel y calculate Total Cardiac \nVolume (TCV) tailored to the Indian population, aim ing to expand the donor \npool and reduce size mismatches. \nMethods or Background: This multi-centre study incorporated paediatric and  \nyoung adults (ages 0-30) with normal CT chest angio grams. TCV was \npredicted using common variables such as weight, he ight, gender and cardiac \nwidth on chest radiograph (CXR) with CT derived TCV  (3D segmentation) as \nthe gold standard. Three predictive models were ana lysed, and subjects were \nsplit into training and testing data. Model A- weig ht only Model B- weight, \nheight, gender and age Model C- Model B plus horizo ntal cardiac width from \nCXR. \nResults or Findings: Model C showed highest accuracy in predicting TCV \nwith an R² of 0.94 for training data and 0.91 for t esting data, with mean \nabsolute percentage error (MAPE) of 3%. Model A was  weakest with an R² of \n0.82 for training data, 0.68 for testing data, and a MAPE of 6.3%. \nConclusion: TCV can be accurately predicted using readily avail able donor \nmetrics. The proposed D-R TCV matching model can si gnificantly expand the \ndonor pool and improve size matching in paediatric heart transplantation in \nIndia. \n\n \n \nFriday \nAbstract-based Programme \n \n 150  \nLimitations: Single centre study, which may also have an in-buil t case \nselection bias. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Owing to the retrospective nature \nof the study, ethical committee approval was waived  off by the institutional \nethics committee. \nAuthor Disclosures:  \nRicha Kothari: Nothing to disclose \nVimal Raj: Nothing to disclose \nSuraj Gowda: Nothing to disclose \n \n \nProspective Comparison of Automated vs. Human-Guide d Cardiac MRI \nPlanning \n*C. G. Glessgen*¹, L. A. Crowe¹, J. Wetzl², M. Schm idt², S. S. Yoon²,  \nJ-P. Vallee¹, J-F. Deux¹; ¹Geneva/CH, ²Erlangen/DE \n \nPurpose or Learning Objective: Cardiac MRI (CMR) is demanding due to the \nnumber of planning steps and parameters requiring c ontinuous monitoring. The \nhigh mistake risk can impact procedure quality, sca n times, and data \nhomogeneity. The impact of an AI-based automated CM R planning software \non procedure errors and scan times compared to huma n-guided examinations \nis evaluated. \nMethods or Background: Consecutive patients undergoing non-stress CMR \nwere prospectively enrolled into two acquisition mo des: manual or automated \nutilizing prototype software (Siemens Healthineers,  Erlangen, Germany). \nPatients with pacemakers or targeted indications we re excluded. All underwent \nthe same CMR protocol with contrast administration,  in breath-hold (BH) or free \nbreathing (FB). Supervising radiologists recorded p rocedure errors (plane \nprescription, forgotten views, incorrect propagatio n of a cardiac plane, field-of-\nview mismanagement). Scan times and Dead Phase (non -acquisition portion) \nwere computed from scanner logs. Most data were non -normally distributed \nand compared using nonparametric tests. \nResults or Findings: Eighty-two patients (mean age, 51.6 years; 56 male)  \nwere included. Forty-four patients underwent automa ted CMR and 38 manual \nCMR. The rate of procedure errors per CMR was lower  (p=0.01) in automated \n(0.45) than in manual (1.13). The ratio of error-fr ee examinations was higher \n(p=0.03) in automated (31/44; 70.5%) than in manual  (17/38; 44.7%). \nAutomated studies were shorter than manual studies in FB (30.3 vs. 36.5 \nminutes, p<.001) but had similar durations in BH (4 2.0 vs. 43.5 minutes, \np=0.42). Dead Phase was lower in automated studies for both FB and BH \nstrategies (p<.001). \nConclusion: AI-based automation performed cardiac MRI studies a t a clinical \nlevel with fewer planning errors and improved effic iency compared to human \nplanning. \nLimitations: No reproductibility analysis of plane adjustments b etween \nradiologists was performed. Radiologists could not realistically be blinded to \nthe study arm as automatic acquisitions were perfor med with almost no visible \nhuman interaction. \nFunding for this study: None \nEthics committee - additional information: All patients gave informed \nconsent \nAuthor Disclosures:  \nLindsey A. Crowe: Nothing to disclose \nJean-Paul Vallee: Nothing to disclose \nCarl Guillaume Glessgen: Nothing to disclose \nSeung Su Yoon: Employee: Siemens Healthineers \nMichaela Schmidt: Employee: Siemens Healthineers \nJean-François Deux: Nothing to disclose \nJens Wetzl: Employee: Siemens Healthineers \n \n \nPredicting Mortality After Transcatheter Aortic Val ve Replacement Using \nAI- Based Fully Automated Left Atrioventricular Cou pling Index \n*E. Zsarnóczay*¹, A. Varga-Szemes², U. J. Schoepf²,  S. Rapaka³, N. Fink²,  \nM. Vecsey-Nagy², P. Sharma³, P. Maurovich-Horvat¹, T. S. Emrich²; \n¹Budapest/HU, ²Charleston, SC/US, ³Princeton, NJ/US  \n \nPurpose or Learning Objective: To determine whether artificial intelligence \n(AI)–based fully automated assessment of left atrio ventricular coupling index \n(LACI) can provide incremental value above other tr aditional risk factors for \npredicting mortality among patients with severe aor tic stenosis (AS) \nundergoing coronary CT angiography (CCTA) before tr anscatheter aortic valve \nreplacement (TAVR). \nMethods or Background: This retrospective study evaluated patients with \nsevere AS who underwent CCTA examination before TAV R between 2014 and \n2019. An AI-prototype software fully automatically calculated left atrial (LA) and \nleft ventricular (LV) end-diastolic volumes and LAC I was defined as the ratio \nbetween them. Clinical parameters, the Society of T horacic Surgeons \nPredicted Risk of Mortality (STS-PROM) risk score, and all-cause mortality \nafter TAVR were recorded. Uni- and multivariate Cox  proportional hazard  \nmethods were used to identify the predictors of mor tality in models adjusting \nfor relevant significant parameters, STS-PROM score , and patients with \npreserved LV ejection fraction (EF). \nResults or Findings: A total of 656 patients (77 years [IQR, 71-84 years ]; 387 \n[59.0%] male) were included. The all-cause mortalit y rate was 21.6% over a \nmedian follow-up time of 24 (10–40) months. When ad justing for clinical \nconfounders, LACI≥43.7% was found to independently predict mortality \n(adjusted HR, 1.52, [95CI: 1.03,2.22]; p=0.032). Af ter adjusting for the STS-\nPROM score in a separate model, LACI ≥43.7% remained an independent \nprognostic parameter (adjusted HR, 1.47, [95CI: 1.0 3,2.08]; p=0.031). In a sub-\nanalysis of patients with preserved LVEF, LACI rema ined a significant predictor \n(adjusted HR, 1.72 [95CI: 1.02,2.89]; p=0.042). \nConclusion: AI-based fully automated assessment of LACI can be used \nindependently to predict mortality in patients unde rgoing TAVR, including those \nwith preserved LVEF. \nLimitations: This study was performed in a single-center and sin gle-vendor \nsetting, using an AI-powered software prototype, se lection bias may exist \nbecause only patients with available outcomes data were included. \nFunding for this study: Not applicable. \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nEmese Zsarnóczay: Nothing to disclose \nPuneet Sharma: Employee: Siemens Healthineers \nPál Maurovich-Horvat: Nothing to disclose \nMilán Vecsey-Nagy: Nothing to disclose \nUwe Joseph Schoepf: Research/Grant Support: Bayer, Bracco, Elucid \nBioimaging, Guerbet, HeartFlow \nSaikiran Rapaka: Employee: Siemens Healthineers \nTilman Stephan Emrich: Consultant: Siemens Medical Solutions USA Inc \nNicola Fink: Nothing to disclose \nAkos Varga-Szemes: Research/Grant Support: Siemens \n \n \nDeep Learning Denoising Algorithm for Improved Asse ssment of \nCoronary Arteries in Transcatheter Aortic Valve Imp lantation CT Imaging \n*L. R. M. Lanzafame*¹, T. D'Angelo¹, A. Othman², C.  Booz³; ¹Messina/IT, \n²Mainz/DE, ³Frankfurt/DE \n(ludovicalanzafame@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the impact of a \ndeep learning-based denoising (DLD) technique on im age quality and \ndiagnostic accuracy for the assessment of coronary arteries in pre-procedural \ntranscatheter aortic valve implantation (TAVI) CT p lanning. \nMethods or Background: A retrospective analysis was conducted on 200 \npatients with severe aortic stenosis who underwent CT scans for TAVI \nplanning between October 2022 and April 2024. Conve ntional images were \nreconstructed, and denoised images were generated u sing DLD model. \nObjective image quality was assessed by measuring t he mean Hounsfield unit \n(HU) and standard deviation (SD) in the aortic root , coronary arteries, and \nsubcutaneous fat to calculate noise, signal-to-nois e ratio (SNR), and contrast-\nto-noise ratio (CNR). Two independent readers subje ctively evaluated \nsharpness, noise, vascular contrast, and overall im age quality using a 5-point \nLikert scale. Diagnostic performance was compared b etween original and \ndenoised images by assessing accuracy, sensitivity,  specificity, positive \npredictive value (PPV), and negative predictive val ue (NPV), using invasive \ncoronary angiography as the reference standard. \nResults or Findings: Denoised images demonstrated significantly improved  \nSNR (37.5 ± 12.8 vs. 12.3 ± 4.1) and CNR (45.3 ± 15.4 vs. 14.7 ± 4.4), along \nwith reduced noise levels (16.9 ± 7.9 vs. 47.9 ± 11.6 HU) (all p<0.001). \nSubjective evaluations also favored denoised images  in terms of sharpness, \nnoise reduction, contrast, and overall quality (all  p<0.001). DLD reconstructions \nrevealed higher diagnostic performance, showing a o n a per-segment basis \nsensitivity of 95.9%, specificity of 94.3%, PPV of 86.5%, NPV of 98.4%, and \naccuracy of 94.8%. \nConclusion: The DLD algorithm significantly improves image qual ity and \ndiagnostic accuracy in pre-TAVI CT imaging for coro nary artery evaluation. \nLimitations: The retrospective design prevented evaluation of im age quality at \nreduced radiation doses. Furthermore, the results a re specific to our \nacquisition protocol. \nFunding for this study: This research did not receive external funding. \nEthics committee - additional information: The study was approved by the \nEthics Committee of Johannes Gutenberg University o f Mainz (Ref. Nr. 2022-\n16477_1) \nAuthor Disclosures:  \nChristian Booz: Speaker: Siemens Healthineers \nLudovica Rosa Maria Lanzafame: Nothing to disclose \nTommaso D'Angelo: Speaker: Philips Speaker: Bracco \nAhmed Othman: Nothing to disclose \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 151  \nA recommendation: test-retest reliability of radiom ic features in \nmyocardial T1 and T2 mapping \n*M. Manzke*¹, F. C. Laqua², B. Böttcher¹, A-C. Klem enz¹, M-A. Weber¹,  \nB. Baeßler², F. G. Meinel¹; ¹Rostock/DE, ²Würzburg/ DE \n(Mathias.Manzke@med.uni-rostock.de) \n \nPurpose or Learning Objective: To investigate the reproducibility of radiomic \nfeatures in myocardial native T1 and T2 mapping. \nMethods or Background: Cardiac MRI T1 maps from 50 healthy volunteers \n(29 women and 21 men, mean age 39.4 ± 13.7 years) u nderwent two identical \ncardiac MRI examinations at 1.5T. The protocol incl uded native T1 and T2 \nmapping in both short-axis and long-axis orientatio n. For T1 mapping, we \ninvestigated standard (1.9 x 1.9 mm) and high (1.4 x 1.4 mm) spatial \nresolution. After manual segmentation of the left v entricular myocardium, 100 \nradiomic features from seven feature classes were e xtracted and analyzed. \nTest–retest repeatability of radiomic features was assessed using the \nintraclass correlation coefficient (ICC) and classi fied as poor (ICC <0.50), \nmoderate (0.50–0.75), good (0.75–0.90) and excellen t (>0.90). \nResults or Findings: For T1 maps acquired in short-axis orientation at \nstandard resolution, repeatability was excellent fo r 6 features, good for 29 \nfeatures, moderate for 19 features and poor for 46 features. We identified 15 \nfeatures from 6 classes which showed good to excell ent reproducibility for T1 \nmapping in all resolutions and all orientations. Fo r short-axis T2 maps, \nrepeatability was excellent for 6 features, good fo r 25 features, moderate for 23 \nfeatures and poor for 46 features. 12 features from  5 classes were found to \nhave good to excellent repeatability in T2 mapping independent of slice \norientation. \nConclusion: We have identified a subset of radiomic features wi th good to \nexcellent repeatability independent of slice orient ation and spatial resolution. \nWe recommend using these features for further radio mics research in \nmyocardial T1 and T2 mapping. \nLimitations: This study was limited to healthy volunteers. The r eproducibility of \nradiomic features in patients with diffuse or focal  myocardial disease cannot be \ndirectly concluded. \nFunding for this study: The study was in part funded by the Federal Ministr y \nof Education and Research (BMBF) through the Networ k University Medicine \n„NUM 2.0“ (grant number 01KX2121). \nEthics committee - additional information: This study was approved by the \ninstitutional review board and written informed con sent was obtained from all \nvolunteers prior to enrollment. \nAuthor Disclosures:  \nFabian Christopher Laqua: Nothing to disclose \nBenjamin Böttcher: Nothing to disclose \nMathias Manzke: Nothing to disclose \nBettina Baeßler: Author: This study has been suppor ted by the Deutsche \nForschungsgemeinschaft (DFG, German Research Founda tion) within the \nPriority Programme SPP 2177 Radiomics (BA 6438/4–2)  and by the Federal \nMinistry of Education and Research (BMBF; “SWAG” pr oject). BB is founder \nand CEO of Lernrad GmbH and has received speaker fe es by Bayer Vital \nGmbH. \nAnn-Christin Klemenz: Nothing to disclose \nFelix G. Meinel: Author: Unrelated to this work, Dr . Meinel has received \ninstitutional research support from GE Healthcare a nd speaker’s honoraria \nfrom GE Healthcare, Circle Cardiovascular Imaging a nd Bayer Vital. \nMarc-André Weber: Nothing to disclose \n \n \nReproducibility of an AI-assisted plane positioning  tool for cardiac MRI \n*B. Böttcher*¹, K. K. Deyerberg¹, A-C. Klemenz¹, L- M. Watzke¹, M. Gorodezky², \nM. Manzke¹, M-A. Weber¹, F. G. Meinel¹; ¹Rostock/DE , ²Munich/DE \n(benjamin.boettcher@gmx.net) \n \nPurpose or Learning Objective: Plane positioning in cardiac magnetic \nresonance imaging (cMRI) is crucial for diagnostic image quality and \ncomparability of cardiac functional parameters in f ollow-up exams. Manual \nplanning is influenced by user’s training making it  susceptible for inter-reader \nvariability and errors. This prospective cohort stu dy aims to investigate the \nreproducibility of an artificial intelligence-based  planning approach against \nstate-of-the-art manual plane prescription. \nMethods or Background: 25 healthy participants (mean age 41.5, range: 23-\n65 years, mean BMI 25.2 kg/m²) underwent two identi cal cMRI exams on a \n1.5T scanner (Signa Artist, GE HealthCare). Short a xis, 2-, 3- and 4-chamber \nplanes (FOV: 34x34cm2, matrix size: 200x224, slice thickness: 8mm) were \nacquired using an AI-based planning tool (TeslaFlow  prototype, GE \nHealthCare) and manual planning. Short axis left ve ntricular volumetric \nanalysis (end-diastolic volume (EDV), end-systolic volume (ESV), stroke \nvolume (SV) and ejection fraction (EF)) were perfor med using an established \npost-processing software (cvi42, Circle Cardiovascu lar Imaging). The Wilcoxon \nmatched-pairs signed rank test with a significance level of p≤0.05 was used to \ncompare the first to the second exam for both the m anual and automated \nplanning. \n \nResults or Findings: Volumetric parameters calculated on manual and AI-\nassisted planned images showed following median of differences between both \nscans: EDV -5.0ml (p=0.220), -2.8ml (p=0.474); ESV 0.1ml (p=0.560), 2.0ml \n(p=0.367); SV -3.6ml (p=0.096), -4.2ml (p=0.043) an d EF -1.2% (p=0.329), -\n2.8% (p=0.045), respectively. The only statisticall y significant differences were \nobserved in SV and EF for AI-based planning, though  the deviation is not \nclinically relevant. \nConclusion: AI-based planning for cMRI showed high reproducibil ity without \nclinically relevant variability between follow-up s cans. This novel technique can \nsimplify and accelerate cMRI maintaining high diagn ostic quality. \nLimitations: This study was conducted on a cohort of healthy ind ividuals at a \nsingle MRI scanner, provided by a single vendor. \nFunding for this study: None. \nEthics committee - additional information: The study was designed as a \nprospective, single-center cohort study and approve d by the responsible \ninstitutional review board of the Medical Universit y Center of Rostock. \nAuthor Disclosures:  \nBenjamin Böttcher: Nothing to disclose \nMargarita Gorodezky: Employee: GE HealthCare \nMathias Manzke: Nothing to disclose \nAnn-Christin Klemenz: Nothing to disclose \nFelix G. Meinel: Nothing to disclose \nMarc-André Weber: Nothing to disclose \nLena-Maria Watzke: Nothing to disclose \nKarolin Kristina Deyerberg: Nothing to disclose \n \n \nAccelerated Deep Learning-Based Function Assessment  in \nCardiovascular Magnetic Resonance \n*F. Fanelli*, D. De Santis, L. Pugliese, G. G. Bona , C. Santangeli, T. Polidori, \nG. Tremamunno, D. Caruso, A. Laghi; Rome/IT \n(f.fanelli@uniroma1.it) \n \nPurpose or Learning Objective: Cardiovascular magnetic resonance (CMR) \nis the reference standard for the assessment of car diac function, achieved \nthrough conventional balanced steady-state free pre cession (bSSFP) cine \nsequences, which represent a considerable part of t he CMR exams, \ncontributing to patient discomfort. The aim of our study was to evaluate \ndiagnostic accuracy and image quality of deep-learn ing(DL)cine sequences for \nLV and RV parameters compared to bSSFP cine sequenc es in CMR. \nMethods or Background: From January to April 2024, patients with clinicall y \nindicated CMR were prospectively included. LV and R V were segmented from \nshort-axis bSSFP and DL cine sequences. LV and RV e nd-diastolic volume, \nend-systolic volume, stroke volume, ejection fracti on, and LV end-diastolic \nmass were calculated. The acquisition time of both sequences was registered. \nResults were compared with paired-samples t-test or  Wilcoxon signed-rank \ntest. Agreement between DL cine and bSSFP was asses sed using Bland-\nAltman plots. Image quality was graded by two reade rs based on blood-to-\nmyocardium contrast, endocardial edge definition, a nd motion artifacts, using a \n5-point Likert scale (1= insufficient quality; 5= e xcellent quality). \nResults or Findings: Sixty-two patients were included (mean age: 47±17 \nyears, 41 men). No significant differences between DL cine and bSSFP were \nfound for all LV and RV parameters (P≥ .176). DL cine was significantly faster \n(1.35 ±.55 m vs 2.83 ± .79 m; P< .001). The agreement between DL cine and \nbSSFP was strong, with near-zero bias and good limi ts of agreement. Overall \nimage quality was comparable (median: 5, IQR: 4-5; P= .330), while \nendocardial edge definition of DL cine (median: 4, IQR: 4-5) was lower than \nbSSFP (median: 5, IQR: 4-5; P= .002). \nConclusion: DL cine allows fast and accurate quantification of LV and RV \nparameters and comparable image quality with conven tional bSSFP. \nLimitations: Not applicable \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: This study has been approved by \nlocal Ethics committee . \nAuthor Disclosures:  \nLuca Pugliese: Nothing to disclose \nDamiano Caruso: Nothing to disclose \nCurzio Santangeli: Nothing to disclose \nFederica Fanelli: Nothing to disclose \nDomenico De Santis: Nothing to disclose  \nGiuseppe Tremamunno: Nothing to disclose \nTiziano Polidori: Nothing to disclose \nAndrea Laghi: Nothing to disclose \nGiovanna Grazia Bona: Nothing to disclose \n \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 152  \nSuper-Resolution Deep Learning Reconstruction to Im prove the \nAccuracy of CT Fractional Flow Reserve: Comparison to Model-based \nIterative Reconstruction \n*N. Tomizawa*, Y. Nozaki, R. Fan, Y. Kawaguchi,, K.  Takamura, F. Shinichiro, \nK. Kumamaru, T. Minamino, S. Aoki; Bunkyo-Ku/JP \n(tomizawa-tky@umin.ac.jp) \n \nPurpose or Learning Objective: The purpose of this study was to compare \nthe diagnostic performance of CT fractional flow re serve (CT-FFR) using \nmodel-based iterative reconstruction (MBIR) and sup er-resolution deep \nlearning reconstruction (SR-DLR) to detect function ally significant stenosis as \nassessed by invasive FFR. \nMethods or Background: This single-center retrospective study included 79 \npatients (mean age, 70 years ± 11 [SD]; 57 men) who  underwent coronary CT \nangiography showing intermediate stenosis (30% ‒70% stenosis) and \nsubsequent invasive FFR between February 2022 and M arch 2024. Vessels \nwith heavy calcification were not excluded from the  analysis. Computational \nfluid dynamics was used to calculate the CT-FFR usi ng MBIR and SR-DLR \nimages. Per-vessel diagnostic performance to detect  FFR ≤0.80 in coronary \nangiography was compared by analyzing receiver oper ating characteristic \n(ROC) curves. \nResults or Findings: Of the 98 vessels evaluated, 46 vessels (47%) had \nfunctionally significant stenosis. The median (inte rquartile range) calcium score \nwas 462 (134–932). CT-FFR values calculated using b oth MBIR (mean \ndifference: −0.088; 95% CI: −0.129, −0.048; p <0.001) a n d SR-DLR (mean \ndifference: −0.026; 95% CI: −0.050, −0.002; p = 0.03) we re  un dere s tima te d \ncompared to invasive FFR. The area under the ROC cu rve to diagnose \nfunctionally significant stenosis was higher for SR -DLR (0.88; 95% CI: 0.80, \n0.95) than for MBIR (0.76; 95% CI: 0.67, 0.86; p = 0.003). CT-FFR calculated \nusing SR-DLR had improved diagnostic accuracy (88% vs. 70%, p <0.001) and \nspecificity (87% vs. 63%, p <0.001) over MBIR, but had similar sensitivity (89% \nvs. 78%, p = 0.06). \nConclusion: SR-DLR images improved the diagnostic performance o f CT-FFR \nover MBIR images in detecting functionally signific ant stenosis as assessed by \ninvasive FFR. \nLimitations: This study is retrospective and used a single CT ve ndor. Multi-\nvendor multi-center study is necessary to confirm t he findings. \nFunding for this study: None \nEthics committee - additional information: Approved by the Ethics \nCommittee of Juntendo University on May 2, 2024 (No . E23-0040-H02) \nAuthor Disclosures:  \nFujimoto Shinichiro: Nothing to disclose \nYuko Kawaguchi,: Nothing to disclose \nShigeki Aoki: Nothing to disclose \nKazuhisa Takamura: Nothing to disclose \nTohru Minamino: Nothing to disclose \nYui Nozaki: Nothing to disclose \nNobuo Tomizawa: Nothing to disclose \nKanako Kumamaru: Nothing to disclose \nRuiheng Fan: Nothing to disclose \n \n \nEvaluating the Feasibility of a Customised GPT-4 Mo del for Extracting \nCAD-RADS Classification from Coronary CT Angiograph y Reports \n*V. Vingiani*, B. Proner, N. Cortellini, R. Vallett a, T. Gorgatti, A. Posteraro,  \nV. Corato, M. Bonatti; Bolzano/IT \n(vincenzovingiani@gmail.com) \n \nPurpose or Learning Objective: This study assessed the feasibility of a \ncustomized GPT-4 model in categorising cardiac radi ological reports using the \nCoronary Artery Disease Reporting and Data System ( CAD-RADS) \nclassification. \nMethods or Background: A customised GPT-4 model was developed using \nthe CAD-RADS 2.0-2022 guidelines, provided as a PDF , and fine-tuned on 30 \nclinical scenarios. The model was tested on 118 ano nymized Coronary CT \nAngiography (CCTA) reports. Data included patient m etrics and report details \n(e.g., length, conclusions). The reports were also reviewed by a radiologist with \n9 years of experience, who categorised them accordi ng to the CAD-RADS \nclassification. The time required for manual assess ment was recorded. The \nGPT-4 model's performance was evaluated using Cohen 's kappa for \nagreement and the Wilcoxon test for comparing proce ssing times between the \nmodel and the radiologist. \nResults or Findings: The mean patient age was 59.6 years (±10.9), with 4 1% \nwomen. Reports were authored by 10 radiologists, wi th 88% in Italian and 12% \nin German. The median report length was 1,456 chara cters, with conclusions \nin 76% of reports. The GPT-4 model showed substanti al agreement with the \nradiologist, achieving a Cohen's kappa of 0.79 (95%  CI: 0.68 - 0.89). It \nsignificantly reduced processing time, averaging 16  seconds per report \ncompared to 57 seconds for the radiologist (P < 0.0 001). \nConclusion: These findings suggest that a customized GPT-4 mode l is a \npromising tool for autonomously categorising radiol ogical findings using the \nCAD-RADS classification when the original report do es not include it, thereby \noffering a time-efficient alternative. Implementing  such a system could assist \nclinicians and cardiologists in consistently interp reting reports by providing \nCAD-RADS classifications when they are not explicit ly reported. \nLimitations: The GPT model was fine-tuned using only 30 scenario s. A larger \ndataset could be used to improve the model's perfor mance. \nFunding for this study: None \nEthics committee - additional information: This study was conducted in \naccordance with the principles of the Declaration o f Helsinki \nAuthor Disclosures:  \nAndrea Posteraro: Nothing to disclose \nMatteo Bonatti: Nothing to disclose \nValentina Corato: Nothing to disclose \nTommaso Gorgatti: Nothing to disclose \nBernardo Proner: Nothing to disclose \nRiccardo Valletta: Nothing to disclose \nNino Cortellini: Nothing to disclose \nVincenzo Vingiani: Nothing to disclose \n \n \nDLR-based Motion Correction of Coronary CTA: Prelim inary Evaluation \nF. Tatsugami¹, A. Streiff², T. Higaki¹, A. Labani²,  W. Fukumoto¹,  \nS. El Ghannudi², K. Haioun³, K. Awai¹, *M. Ohana*²;  ¹Hiroshima/JP, \n²Strasbourg/FR, ³Tokyo/JP \n(mickael.ohana@gmail.com) \n \nPurpose or Learning Objective: DLR-based Motion Correction (MC-DLR) for \nCoronary CTA has the potential to reduce/eliminate kinetic artifacts in CCTA \nmore effectively than traditional algorithms, but i ts clinical impact is still \nunknown. We aim to evaluate the effect of MC-DLR on  coronary luminal and \nstenosis assessment in a varied CCTA cohort. \nMethods or Background: Sixty CCTA (20 with HR<60bpm, 20 with HR=60-\n75, 20 with HR>75) with various degrees of stenosis  (50% CAD-RADS 1 & 2, \n50% CAD-RADS 3 & 4) were retrospectively selected f rom 2 tertiary centers. \nAll scans were acquired on 4th/5th-gen wide-area de tector CT within 1 \nheartbeat. Best phase for each included CCTA was re constructed without and \nwith MC-DLR, using Super Resolution DLR with 1024 m atrix-size. MC-DLR \nsubdivides the data required for volume reconstruct ion into smaller time \nsections to estimate coronary artery motion. Four r adiologists with varying \nlevels of expertise independently and randomly revi ewed all 120 datasets to: \n(A) grade the luminal/wall image quality using a 3- level scale, for the 9 \ncoronary artery segments, and (B) assess CAD-RADS. Statistical analysis \nused descriptive and Bayesian approaches. \nResults or Findings: For each reader and for the pooled analysis, overal l \nluminal/wall image quality score was significantly better with MC-DLR than \nwithout (p<0.05). Per segment, the positive effect was more consistent on the \nRCA (improvement in 62% of cases) than on the LAD ( 44%) and the Cx (21%). \nEffect of MC-DLR was non-existent in cases with abs ent/minimal coronary \nkinetic artifacts. Non-significant changes in intra /inter-reader variability were \nnoted in CAD-RADS 3/4. \nConclusion: MC-DLR significantly enhances coronary artery lumin al and wall \nimage quality in cases with moderate or severe kine tic artifacts, suggesting a \npotential clinical role in refining stenosis assess ment when above >50%. \nLimitations: Quantitative analysis with attenuation profile curv es was not \nperformed. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: IRB from Strasbourg University \nHospital \nAuthor Disclosures:  \nWataru Fukumoto: Nothing to disclose \nAissam Labani: Nothing to disclose \nSoraya El Ghannudi: Nothing to disclose \nFuminari Tatsugami: Nothing to disclose \nToru Higaki: Nothing to disclose \nKazuo Awai: Nothing to disclose \nAmandine Streiff: Nothing to disclose \nMickaël Ohana: Consultant: Boehringer Ingelheim Con sultant: Canon Medical \nSystems Europe \nKarim Haioun: Employee: Canon Medical Systems Japan  \n \n \nPerformance of AI-based automated coronary artery c alcium density \nquantification on CT \nY. J. Suh¹, C. Kim², W-S. Yoo¹, J. Y. Kim³, S. Chan g¹, C. H. Park⁴,  \n*S. J. Hong*⁵, D. H. Yang¹, H. S. Yong¹; ¹Seoul/KR, ²Ansan/KR, ³ Daegu/KR, \n⁴Cheonan/KR, ⁵Guri/KR \n \nPurpose or Learning Objective: Coronary artery calcium (CAC) density on \nelectrocardiogram (ECG)-gated CT has been suggested  as an inverse \nprognostic marker for prediction of future adverse cardiovascular events. We \naimed to evaluate the performance of artificial int elligence (AI)-based \nautomated CAC density quantification on ECG-gated c alcium scoring CT \n\n \n \nFriday \nAbstract-based Programme \n \n 153  \n(CSCT) and non-ECG-gated low-dose chest CT (LDCT), using multi-\ninstitutional datasets. \nMethods or Background: A total of 1,540 pairs of CSCT-LDCT scans from a \nmulticenter database were retrospectively included.  AI-based automated CAC \nquantification was conducted on the CSCT and LDCT. For cases with CAC \nscore>0, mean and peak CAC density was calculated f rom the labeled CAC. \nFor peak CAC density factors, a value of 1 to 4 was  assigned based on the \nmeasured peak density attenuation (1: 130-199HU; 2:  200-299HU; 3: 300-\n399HU; 4: >400HU). The reliability and agreement of  the mean CAC density \nand peak CAC density categories obtained from the a utomated scoring were \nanalyzed compared to manual measurement, using the intraclass correlation \ncoefficient (ICC), Bland-Altman analysis, and weigh ted kappa (κ) statistics, \nrespectively. \nResults or Findings: A total of 808 CSCT scans and 579 LDCT scans were \npositive for CAC. Automated mean density measuremen t demonstrated \nexcellent ICCs on CSCT and LDCT (0.988 [95% CI, 0.9 86-0.989] vs. 0.956 \n[95% CI, 0.948-0.962]). Mean bias with 95% limits o f agreement for the mean \ndensity was 0.7 ± 22.4 on CSCT and 0.6 ± 27.4 on LDCT. In terms of peak \ndensity category, automated measurement on CSCT and  LDCT exhibited \nexcellent reliability with manual measurement (weig hted κ 0.980 [95% CI, \n0.969-0.991) and 0.964 [95% CI, 0.945-0.982]). \nConclusion: AI-based automated CAC quantification can provide a ccurate \nand reliable measurement of CAC density on CSCT and  LDCT across multi-\ninstitutional datasets. \nLimitations: Prognostic value of AI-based CAC density should be further \ninvestigated. \nFunding for this study: The Researcher Supporting Program funded by \nKorean Society of Cardiovascular Imaging (KOSCI) an d the National Research \nFoundation of Korea (NRF) grant funded by the Korea  government (MSIT)(No. \n2021R1A2C4002195) \nEthics committee - additional information: Approval numbers: \nKC22RIDI0156, 2021-12-027, 2022GR0064, 2021AS0371, 2022-01-001, \n2021-0303, 2021-12-029, and 4-2021-1589 \nAuthor Disclosures:  \nDong Hyun Yang: Nothing to disclose \nYoung Joo Suh: Nothing to disclose \nSuyon Chang: Nothing to disclose \nChan Ho Park: Nothing to disclose \nJin Young Kim: Nothing to disclose \nSu Jin Hong: Nothing to disclose \nCherry Kim: Nothing to disclose \nHwan Seok Yong: Nothing to disclose \nWon-Seok Yoo: Nothing to disclose \n \n \n09:30-11:00 Research Stage 4 \nResearch Presentation Session: Neuro \nRPS 1311 \nDecoding the mind: sculpting \nneuroimaging with technology \n \nModerator \nA. Krainik; Grenoble/FR  \n(akrainik@gmail.com) \nAuthor Disclosures:  \nAlexandre Krainik: Advisory Board: Geodaisics \n \n \nClinical Evaluation of 3D Motion-Correction via Sco ut Accelerated Motion \nEstimation and Reduction (SAMER) framework versus C onventional T1-\nWeighted MRI at 1.5 T in Brain Imaging \n*L. Leukert*, A. Kronfeld, R. Paul, M. A. Brockmann , S. Altmann, A. Othman; \nMainz/DE \n(laura@s-leukert.com) \n \nPurpose or Learning Objective: To evaluate the presence of motion artifacts \nin 1.5 T T1-weighted MRI scans using 3D motion corr ection via the Scout \nAccelerated Motion Estimation and Reduction (SAMER)  framework versus \nconventional image reconstruction. \nMethods or Background: MRI long scan times often cause motion artifacts, \nreducing image quality. SAMER uses an ultrafast pre -scan and repeated \nacquisition of a minimal number of additional k-spa ce encoding lines, enabling  \n \n \nfeasible computation times. A preliminary study (14  volunteers) assessed \nSAMER’s effect on induced motion at 3T. The main st udy (82 patients) \ncompared conventional resonstruction (Non-Moco) and  SAMER (SAMER \nMoco) motion correction using 3D T1-weighted imagin g at 1.5T. Radiologists \nevaluated images with a 5-point Likert scale. \nResults or Findings: In the preliminary study, SAMER Moco showed \nsignificant improvements over Non-Moco across all i maging parameters (p < \n0.001), with 52.4% and 66.7% of cases rated as exce llent or good for artifact \nfreedom and image quality, compared to 21.4% for No n-Moco. The main study \nunderlined these findings. SAMER Moco demonstrated superior image quality \nand outperformed Non-Moco, particularly in diagnost ic confidence and overall \nimage quality (p < 0.0001). Diagnostic confidence w as rated excellent or good \nin 93.8% of SAMER Moco cases versus 72.0% for Non-M oco. Similarly, 84.6% \nof SAMER Moco cases had excellent or good image qua lity (56.8% for Non-\nMoco). Odds ratios favoured SAMER Moco (5.444 and 5 .807, respectively, p < \n0.0001). Multi-reader agreement was excellent acros s all parameters. \nConclusion: The use of SAMER in T1-weighted imaging is feasible  in clinical \npractice and significantly enhances the reliability  of 1.5 T brain MRI by \nsuccessfully mitigating motion artifacts. \nLimitations: This study's limitations include its single-centre design, reliance \non a single 3D MR sequence, and inclusion of both c ontrast-enhanced and \nnon-contrast scans. SAMER may also be affected by p atient-induced k-space \ngaps. \nFunding for this study: This research received no funding from any public, \ncommercial, or not-for-profit sources. The authors declare that they have no \ncompeting interests that are relevant for the conte nt of this article. \nEthics committee - additional information: This single-center prospective \nstudy was approved by our institution's local ethic s committee, and written \ninformed consent was obtained (approval number 2021 -15811). Our study was \nconducted in accordance with the Declaration of Hel sinki and its amendments. \nAuthor Disclosures:  \nRoman Paul: Nothing to disclose \nLaura Leukert: Nothing to disclose \nAndrea Kronfeld: Nothing to disclose \nMarc A Brockmann: Nothing to disclose \nAhmed Othman: Nothing to disclose \nSebastian Altmann: Nothing to disclose \n \n \nComparison of Photon-Counting CT and Conventional C T for \nDetermining Rotational Orientation of Directional D BS Electrodes:  \nA Phantom Study \n*D. Fedders*¹, A. Hellerbach², M. Eichner², C. Pank nin³, S. Faby³, J. Wirths², \nV. Visser-Vandewalle², H. Treuer², S. Hunsche²; ¹Ch emnitz/DE, ²Cologne/DE, \n³Forchheim/DE \n(dieter.fedders@gmail.com) \n \nPurpose or Learning Objective: Accurate determination of the rotational \norientation of directional deep brain stimulation ( DBS) electrodes is crucial for \noptimizing therapeutic outcomes in functional neuro surgery. Conventional CT \nmethods, relying on artifact analysis, face precisi on limitations, especially at \ncertain angles. Photon-counting detector CT (PCD-CT ), with its superior \nresolution, offers a potential alternative. This st udy compares the efficacy of \nPCD-CT against conventional CT-based artifact analy sis in determining DBS \nelectrode orientation. \nMethods or Background: A phantom study was conducted using directional \nleads from Boston Scientific, Medtronic, and Abbott  embedded in cylindrical \nphantoms. The phantoms were scanned with PCD-CT for  direct orientation \ndetection and conventional CT for stripe artifact a nalysis. Scans covered \nvarying polar angles to assess accuracy and consist ency. Key metrics included \norientation accuracy and dependency on lead positio n relative to the CT \ngantry. \nResults or Findings: PCD-CT demonstrated high accuracy across all tested  \nangles, independent of lead alignment. In contrast,  conventional CT showed \nreduced precision, particularly at extreme angles w here artifact detection was \nunreliable. PCD-CT enabled consistent, precise asse ssments of segmented \ncontacts, enhancing postoperative DBS programming. \nConclusion: PCD-CT offers a robust solution for determining the  rotational \norientation of DBS electrodes, overcoming limitatio ns of conventional artifact-\nbased methods. This supports more accurate electrod e positioning and \nprogramming, potentially improving functional neuro surgery outcomes. \nLimitations: The phantom-based design may not replicate clinical  complexity, \nlimiting generalizability. The study only evaluated  specific directional DBS \nleads, so results may not apply to other types. Add itionally, PCD-CT’s limited \navailability could hinder immediate clinical applic ation. \nFunding for this study: None beside scanning time at the research facility \nfrom Siemens \nEthics committee - additional information: Phantom study \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 154  \nAuthor Disclosures:  \nHarald Treuer: Nothing to disclose \nJochen Wirths: Nothing to disclose \nVeerle Visser-Vandewalle: Nothing to disclose \nDieter Fedders: Nothing to disclose \nChristoph Panknin: Other: Computed Tomography, Siem ens Healthineers AG, \nForchheim/DE \nSebastian Faby: Other: Computed Tomography, Siemens  Healthineers AGC \nAlexandra Hellerbach: Nothing to disclose \nMarkus Eichner: Nothing to disclose \nStefan Hunsche: Nothing to disclose \n \n \nA Multimodal MRI-Based Machine Learning Framework f or Classifying \nCognitive Impairment in Cerebral Small Vessel Disea se \n*G. Lin*, W. Chen, M. Chen, J. Ji; Lishui/CN \n \nPurpose or Learning Objective: This study aims to propose a multimodal \nmagnetic resonance imaging (MRI)-based machine lear ning framework to \neffectively classify mild cognitive impairment (MCI ) and no cognitive \nimpairment (NCI) in patients with cerebral small ve ssel disease (CSVD). \nMethods or Background: We enrolled 223 patients with CSVD, categorized \ninto NCI (n = 121) and MCI (n = 102) groups based o n neurocognitive \nassessments. Multimodal MRI data, including T1-weig hted, resting-state \nfunctional MRI, and diffusion tensor images, were c ollected. Image \npreprocessing, feature extraction, and feature sele ction methods were applied \nto obtain MRI features from the three modalities. T he AutoGluon platform was \nutilized for model development, and traditional mac hine learning algorithms \nwere applied for comparison. The models were valida ted using a validation \ncohort of 97 patients with CSVD, and their performa nce was assessed via \nreceiver operating characteristic curve (ROC) analy sis. \nResults or Findings: The AutoGluon model to distinguish MCI from NCI \nbased on multimodal MRI features demonstrated a hig h area under the ROC \ncurve (AUC), accuracy, sensitivity, specificity, an d F1-score in the testing set \n(0.894, 85.65%, 84.31%, 86.78%, and 84.31%, respect ively) and validation \ncohort (0.846, 79.38%, 81.82%, 77.36%, and 78.26%, respectively). Other \nmodels built using traditional machine learning alg orithms had AUCs of 0.661–\n0.732, and their prediction accuracies were signifi cantly lower than that of the \nAutoGluon model (P < 0.001). \nConclusion: Our study provides a multimodal MRI-based machine l earning \nframework, utilizing the AutoGluon platform, that o utperforms traditional \nalgorithms in classifying MCI and NCI in patients w ith CSVD, offering a \npromising tool for the early prediction of MCI in C SVD. \nLimitations: As a retrospective study, it is susceptible to sele ction bias, which \nmay limit its generalizability. \nFunding for this study: This study is supported by Zhejiang Public Welfare \nResearch Program (LGF20H220002, LGF19H180010), and Zhejiang \nProvincial Healthcare Program (2024KY562) \nEthics committee - additional information: This study was approved by the \nEthics Committee of the Fifth Affiliated Hospital o f Wenzhou Medical University \n(approval number: 2024-266) \nAuthor Disclosures:  \nMinjiang Chen: Nothing to disclose \nJiansong Ji: Nothing to disclose \nWeiyue Chen: Nothing to disclose \nGuihan Lin: Nothing to disclose \n \n \nDetection of intracranial hemorrhage using ultralow -dose brain computed \ntomography with deep learning reconstruction versus  conventional-dose \ncomputed tomography \n*C. Otgonbaatar*¹, H. Kim², P-H. Jeon², S. H. Jeon² , S. Cha², J-K. Ryu¹,  \nH. Shim¹, S. M. Ko², J. Kim²; ¹Seoul/KR, ²Wonju-si/ KR \n(chukarad@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the diagnostic \nperformance, image quality, and radiation dose amon g ultralow-dose protocol \nwith deep learning reconstruction (DLR), ultralow-d ose computed tomography \n(CT) with iterative reconstruction (IR), and conven tional-dose protocols for \ndetecting intracranial hemorrhage. \nMethods or Background: This retrospective study enrolled 93 patients. All \npatients underwent follow-up noncontrast CT with ul tralow-dose setting after \ninitial conventional-dose CT within 5 days. A conve ntional-dose CT was \nobtained using 123–188 mA and IR. Ultralow-dose CT was obtained using 50 \nmA with IR and DLR. Qualitative assessments and qua ntitative assessments \n(image noise, differentiation between gray and whit e matter, and artifact) were \nconducted. The diagnostic performance for detecting  intracranial hemorrhage \nusing ultralow-dose CT with IR and ultralow-dose CT  with DLR was assessed. \n \n \n \nResults or Findings: An approximately 84.0% reduction in median volume C T \ndose index was found in the ultralow-dose CT protoc ol (5.6 mGy) compared \nwith conventional-dose CT (35.02 mGy; IQR: 33.09–37 .36). Ultralow-dose CT \nwith DLR significantly (p < 0.001) improved image n oise, SNR, and CNR \ncompared with ultralow-dose CT with IR and conventi onal-dose CT. Ultralow-\ndose CT with DLR resulted in higher sensitivity (99 .3% vs. 98.6%) and \nspecificity (97.5% vs. 97.5%) for detecting intracr anial hemorrhage than \nultralow-dose CT with IR. \nConclusion: Ultralow-dose CT with DLR is an acceptable techniqu e that \nprovides higher image quality and diagnostic perfor mance with a reduction in \nradiation dose of approximately 87.7% compared with  conventional-dose CT. \nLimitations: We did not investigate the effect of a tube current  of <50 mA on \nthe diagnostic performance of intracranial hemorrha ge and image quality. \nFurther validation is required to investigate a low er effective dose of <0.21 \nmSv. Additionally, all results were limited to one scanner, and acquisition \nparameters may require adjustment for different CT vendors. \nFunding for this study: None \nEthics committee - additional information: No \nAuthor Disclosures:  \nSung Min Ko: Nothing to disclose \nPil-Hyun Jeon: Nothing to disclose \nJinwoo Kim: Nothing to disclose \nSungjin Cha: Nothing to disclose \nJae-Kyun Ryu: Nothing to disclose  \nHyunjung Kim: Nothing to disclose \nSang Hyeon Jeon: Nothing to disclose \nHackjoon Shim: Nothing to disclose \nChuluunbaatar Otgonbaatar: Nothing to disclose \n \n \nCorrelation of diffusion tensor imaging findings in  cerebral sensorimotor \nregions with neurophysiological deficits in patient s after spinal cord \ninjury \nA. Zimny¹, *W. N. Machaj*¹, P. Podgórski¹, W. Fortu na¹, J. Huber²,  \nB. Bobek-Billewicz³, P. Tabakow¹; ¹Wrocław/PL, ²Poz nań/PL, ³Gliwice/PL \n \nPurpose or Learning Objective: The aim of the study is to examine the \ncorrelation between DTI findings, clinical motor an d sensory deficits, and motor \nevoked potential (MEP) parameters. This will provid e insights into the potential \nof DTI metrics as biomarkers for predicting functio nal recovery and guide \ntherapeutic interventions in patients with chronic spinal cord injury (SCI). \nMethods or Background: A total of 29 patients with SCI (both paraplegic p-\nSCI and tetraplegic t-SCI), matched by sex and age to 29 healthy controls, \nwere neurologically and neurophysiologically evalua ted, including MEPs \nrecorded from upper and lower limb muscles. Diffusi on tensor imaging (DTI) \nwas performed using a 3 Tesla MRI scanner and proce ssed using Human \nMotor Area (HMAT) and Sensorimotor Area Tract (SMAT T) templates. \nResults or Findings: No significant DTI differences were found between p -\nSCI and t-SCI or p-SCI and healthy controls. Howeve r, patients with t-SCI had \nlower fractional anisotropy (FA) in primary motor ( M1) and sensorimotor (S1) \ntracts, pre-supplementary motor area (pre-SMA) trac ts, M1 and S1 cortices, \nand left pre-SMA cortex compared to controls. In t- SCI patients, higher motor \nscores correlated with increased FA in ventral prem otor area (PMv) tracts and \ncortices, and higher sensory scores with higher FA in S1 tracts. MEP \namplitudes from rectus femoris also positively corr elated with FA in motor \ntracts, M1, PMd, PMv, and SMA cortices. \nConclusion: DTI findings reveal distant degeneration in the sen sorimotor \ncortex and supraspinal tracts in chronic SCI, which  correlates with clinical \nmotor and sensory scores, as well as MEP parameters  from rectus femoris \nmuscles in t-SCI patients. DTI metrics can serve as  potential biomarkers to \npredict motor and sensory recovery in patients with  SCI and to guide and track \ntherapeutic interventions. \nLimitations: Cross-sectional design and the small sample size. \nFunding for this study: Grant NCBiR ERA-NET-NEURON/13/2018, Wroclaw \nMedical University grant SB. \nEthics committee - additional information: The study was performed in \naccordance with the Declaration of Helsinki and was  approved by the Bioethics \nCommittee of the Wroclaw Medical University. \nAuthor Disclosures:  \nJuliusz Huber: Author: Nothing \nAnna Zimny: Author: Nothing \nPaweł Tabakow: Author: Nothing \nWeronika Natalia Machaj: Author: Nothing \nPrzemysław Podgórski: Author: Nothing \nBarbara Bobek-Billewicz: Author: Nothing \nWojciech Fortuna: Author: nothing \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 155  \nMR Neurography at 3T and 7T for the assessment of p roximal nerve \ndamage in polyneuropathies \n*J. M. E. Jende*¹, C. Mooshage¹, K. Zhang¹, T. Plat t¹, C. Neelsen¹,  \nM. Bendszus¹, H-P. Schlemmer¹, M. Ladd¹, F. Kurz²; ¹Heidelberg/DE, \n²Geneva/CH \n \nPurpose or Learning Objective: Disorders of the peripheral nervous system \nsuch as polyneuropathies pose a huge challenge to t he global healthcare \nsystem. The exact pathophysiology underlying most p olyneuropathies remains \npoorly understood. Previous studies on 3T MR neurog raphy (MRN) have found \nthat the maximum of fascicular nerve damage in vari ous polyneuropathies is \nlocated at the level of the sciatic nerve although clinical symptoms usually \noccur further distally. \nMethods or Background: To understand the clinical impact and physiological  \nbackground of fascicular sciatic nerve lesions, 10 patients with distal symmetric \npolyneuropathy and 10 healthy controls matched for age and BMI underwent \nT2-weighted, high resolution MRN of the right thigh  at 3T and 7T. At 7T, the \nmaximum fascicular diameter and the average number of nerve fascicles per \nslice were measured. At 3T, additional diffusion-we ighted and T2-relaxometry \nsequences were acquired and the sciatic nerve’s fra ctional anisotropy (FA) and \nT2 relaxation times (T2R) were calculated. \nResults or Findings: At 7T, the sciatic nerve’s maximum fascicular diame ter \nin patients with polyneuropathies was larger compar ed to controls \n(1.41mm±0.16 vs. 1.03mm±0.07; p=0.049mm). In patients with \npolyneuropathies, the fascicular diameter was negat ively correlated with the \nnumber of nerve fascicles (r=-0.72;p=0.018) and the  FA (r=-0.78;p=0.017mm). \nPositive correlations were found between the fascic ular diameter and T2R \n(r=0.76;p=0.019). \nConclusion: The results indicate that polyneuropathies cause a fusion of \nnerve fascicles that results in a reduced number of  nerve fascicles and a larger \nfascicular diameter that is associated with a struc tural and functional decline \nrepresented by a decrease in nerve FA and an increa se in T2R. This study is \nthe first to show that the fascicular diameter of t he sciatic nerve is directly \nrelated to changes in FA, T2R and clinical neuropat hy status. The causes of \nfascicular fusion remain to be determined. \nLimitations: Cohort Size \nFunding for this study: Else Kröner Fresenius Foundation (EKFS) \nEthics committee - additional information: This study was appoved by the \nlocal ethics committee of Heidelberg University Hos pital. \nAuthor Disclosures:  \nTanja Platt: Nothing to disclose \nKe Zhang: Nothing to disclose  \nChristian Neelsen: Nothing to disclose \nJohann Malte Enno Jende: Nothing to disclose \nMark Ladd: Nothing to disclose \nFelix Kurz: Nothing to disclose \nMartin Bendszus: Nothing to disclose \nChristoph Mooshage: Nothing to disclose \nHeinz-Peter Schlemmer: Nothing to disclose \n \n \nDeep-learning-reconstructed 3D MR neurography of ex traforaminal \ncranial and spinal nerves \nF. Ensle¹, *F. Zecca*², B. J. Kerber¹, M. Lohezic¹,  J. Kroschke¹, K. Pawlus¹,  \nR. Guggenberger³; ¹Zurich/CH, ²Cagliari/IT, ³Winter thur/CH \n(fabio.zecca92@gmail.com) \n \nPurpose or Learning Objective: To assess and compare DESS and post-\ncontrast STIR sequences in deep-learning(DL)-recons tructed 3D MR \nneurography of the extraforaminal cranial and spina l nerves. \nMethods or Background: Eighteen consecutive exams of 18 patients with \nunclear cephalgia undergoing head-and-neck MRI at 1 .5T were retrospectively \nincluded (mean age: 51 ± 14 years, 11 female). 3D DESS and post-contrast 3D \nSTIR sequences were reconstructed with a prototype DL algorithm. Two \nblinded readers qualitatively evaluated visualizati on of the inferior alveolar \n(IAN), lingual (LN), facial (FN), hypoglossal (HN),  greater occipital (GON), \nlesser occipital (LON) and greater auricular (GAN) nerves, as well as overall \nimage quality, vascular suppression and artifacts. Apparent signal-to-noise \nratio (aSNR) and contrast-to-noise ratios (aCNR) we re measured. Qualitative \nratings were compared between sequences using Wilco xon signed-rank test, \nquantitative analysis with paired sample Student’s t-testing. \nResults or Findings: DESS demonstrated significantly improved visualizat ion \nof the LON and GAN and proximal GON (p < 0.015). Po st-contrast STIR \nshowed significantly enhanced visualization of the LN, HN and distal IAN (p < \n0.001). The FN, proximal IAN and distal GON did not  demonstrate significant \ndifferences in visualization between DESS and post- contrast STIR (p > 0.08). \nWith regard to overall image quality and artifacts,  there was also no significant \ndifference between sequences. Post-contrast STIR ac hieved superior vascular \nsuppression, reaching statistical signifi-cance for  one reader (p = 0.039). \nQuantitatively, there was no significant difference  between sequences (p > \n0.05). \nConclusion: Our findings suggest that 3D DESS generally provide s improved \nvisualization of spinal nerves (GON, LON, GAN), whi le post-contrast 3D STIR \nfacilitates enhanced delineation of extraforaminal cranial nerves (IAN, LN, HN). \n3D DESS and post-contrast 3D STIR could each add va lue to head-neck MRN \nprotocols, depending on the main clinical area of i nterest. \nLimitations: Retrospective study. Limited sample size. \nFunding for this study: Not applicable. \nEthics committee - additional information: This study was approved by the \ninstitutional review board. \nAuthor Disclosures:  \nMaelene Lohezic: Employee: GE HealthCare \nFabio Zecca: Nothing to disclose \nFalko Ensle: Nothing to disclose \nRoman Guggenberger: Nothing to disclose \nKarolina Pawlus: Nothing to disclose \nBjarne Jonas Kerber: Nothing to disclose \nJonas Kroschke: Nothing to disclose \n \n \nEnhancing Imaging Efficiency in Advanced Diffusion Imaging Using \nDenoising and Post-Processing Techniques \n*V. Sedlák*, K. Vambersky, A. Kavková, K. Sichova, D. Netuka, T. Belsan,  \nM. Majovsky; Prague/CZ \n \nPurpose or Learning Objective: The objective of this study is to demonstrate \nhow the application of advanced denoising technique s, such as MP-PCA and \nP2S, along with post-processing methods for enhanci ng angular resolution, \ncan significantly reduce imaging times in advanced diffusion imaging. This \napproach aims to preserve or improve image quality while reducing the \nacquisition time burden in clinical and research se ttings. \nMethods or Background: We acquired two sets of advanced diffusion MRI \ndata from patients with glial brain tumors: one usi ng a full-length acquisition \nand the other using a fast acquisition protocol. Th e full-length dataset was \nprocessed directly, while the fast protocol data wa s enhanced with denoising \nalgorithms (e.g. MP-PCA, P2S) and angular super-res olution reconstruction \ntechniques. We evaluated data quality by assessing signal-to-noise ratio \n(SNR), angular resolution, and diagnostic accuracy for predicting glioma grade \nand IDH mutation status. \nResults or Findings: Data from 100 patients with glial brain tumors were  \nprocessed and analyzed. The shortened protocol data , when enhanced by MP-\nPCA and P2S denoising, provided a significant impro vement in SNR, aligning \nclosely with or even superseding the quality of the  full-length acquisition. \nAngular super-resolution reconstruction techniques further enhanced the \nangular resolution without extending scan time. The  diagnostic accuracy for \npredicting glioma grade and IDH mutation status was  comparable between the \nfull-length and processed shortened datasets, demon strating that these post-\nprocessing methods can preserve clinical diagnostic  value while reducing scan \nduration \nConclusion: Denoising techniques and angular resolution enhance ment \nsignificantly improve the quality of shortened diff usion MRI acquisitions, \nmaintaining diagnostic accuracy for glioma grading and IDH status while \nreducing scan time. \nLimitations: Results are based on 100 glioma patients and requir e validation \nin broader populations. The computational demands m ay limit immediate \nclinical application, and performance may vary with  MRI hardware and \nsequence parameters. \nFunding for this study: This study was supported by the Grant Agency of \nCharles University, grant number GAUK 222623 \nEthics committee - additional information: Approved by the Ethics \ncommittee of the Military. University Hospital Prag ue \nAuthor Disclosures:  \nMartin Majovsky: Nothing to disclose \nVojtěch Sedlák: Nothing to disclose \nAnna Kavková: Nothing to disclose \nDavid Netuka: Nothing to disclose \nKamil Vambersky: Nothing to disclose \nKristyna Sichova: Nothing to disclose \nTomas Belsan: Nothing to disclose \n \n \nDTI-ALPS Mapping: A Novel Method that Can Comprehen sively Reflect \nthe DTI-ALPS pattern \nX. Fan, *G. Cheng*, X. Zhang, N. Zhang; Shenzhen/CN  \n \nPurpose or Learning Objective: Our objectives are to depict the whole white \nmatter DTI-ALPS changes pattern in cognitive impair ment and provide an \nintuitive and comprehensive method to assess the ac tivity of the glymphatic \nsystem. \nMethods or Background: The glymphatic system is increasingly recognized \nas a critical factor in the pathogenesis of dementi a. We creatively propose a \nnovel method to analyze the whole white matter diff usion tensor image \nanalysis along the perivascular space (DTI-ALPS). W e included 304 \n\n \n \nFriday \nAbstract-based Programme \n \n 156  \nparticipants from the Shenzhen Multimodal Aging Res earch (STAR) Cohort \nrecruited in Peking University Shenzhen Hospital, i ncluding 182 cognitively \nunimpaired (CU) participants, 93 participants with mild cognitive impairment, \nand 29 patients with dementia. All participants und erwent 3.0T MRI scans with \nDTI sequences. We calculated the DTI-ALPS values us ing the conventional \nmethod and depicted the whole white matter DTI-ALPS  mapping with the novel \nmethod we developed. We performed the analysis of v ariance (ANOVA) to find \nthe differences in ALPS values among three groups a nd used post-hoc tests to \nfind intergroup differences in the conventional met hod and DTI-ALPS mapping, \nrespectively. \nResults or Findings: We revealed the whole brain ALPS pattern using the \nDTI-ALPS mapping. Over 13 out of 30 regions showed significant differences \n(p < 0.05) in the inter-group analysis, which provi ded additional information \nbeyond the significant differences based on the con ventional ROI-based ALPS. \nConclusion: In conclusion, the DTI-ALPS mapping provides a robu st, intuitive, \nand comprehensive way to evaluate the changing patt ern of the glymphatic \nsystem, overcoming the limitations introduced by co nventional ROI-based DTI-\nALPS calculating methods. \nLimitations: Future studies should integrate additional statisti cs such as \nminimum, maximum, and standard deviation into ALPS- mapping analyses for a \ncomprehensive understanding of the DTI-ALPS mapping . \nFunding for this study: This study is principally supported by the Shenzhen  \nScience and Technology Program (KCXFZ 2021102016340 8012). \nEthics committee - additional information: the Ethics Committee of Peking \nUniversity Hospital \nAuthor Disclosures:  \nXiang Fan: Nothing to disclose \nXiqian Zhang: Nothing to disclose \nGuanxun Cheng: Nothing to disclose \nNa Zhang: Nothing to disclose \n \n \nHDD-Net: Hippocampus Dual Decoder Network for autom ated \nsegmentation of hippocampus from computed tomograph ic scans \n*S. J. Ahn*, W. J. Son, J. Y. Lee, H. Lee; Seoul/KR  \n(aahng77@yuhs.ac) \n \nPurpose or Learning Objective: Changes of brain hippocampal volumes are \nclosely associated with the development of Alzheime r’s disease. In this work, \nwe develop a new deep learning (DL) network model f or volumetric \nhippocampal segmentation from computed tomography ( CT) head images, a \ntask that has been challenged due to the modality’s  limited brain contrast. \nMethods or Background: HDD-Net: The proposed network model is \ncharacterized by four major elements – 1) an encode r , 2) two parallel \ndecoders (namely, seg-decoder and edge-decoder), 3)  a feature cross module \n(FCM) fusing features from the two decoders and 4) a cross loss computing \ndifferences between outputs. Datasets and preproces sing: 150 pairs of MRI-\nCT volumetric head images collected at Gangnam Seve rance Hospital were \nused for model training (N=120) and internal valida tion (N=30), while 47 pairs \nwere selected from Seoul St. Mary’s Hospital databa se for external validation. \nGround-truth hippocampal labels were generated from  T1-weighted MR \nimages using FreeSurfer, and then were processed by  a Gaussian high-pass \nfilter leading to reference edge maps. Each pair of  MR-CT images were \ncoregistered using SPM12. Training and Evaluation: The DL model was trained \nwith a cost function combining segmentation, edge, and cross losses. Its \nperformance was evaluated by calculating Dice coeff icient and intersection-\nover-union (IOU). The performance our model was com pared with that of \nconventional U-Net model. \nResults or Findings: Dice and IOU of our model is higher than those of U -net \nfor internal validation set (DICE : 0.840 vs. 0.822 , IOU: 0.726 vs. 0.699) and \nexternal validation (DICE : 0.784 vs. 0.751, IOU: 0 .650 vs. 0.613) \nConclusion: Results suggest feasibility of DL-based automated h ippocampal \nsegmentation from CT scans and its improved perform ance via edge decoding. \nLimitations: The applicability of the model could be enhanced by  training it on \npatients with dementia \nFunding for this study: None \nEthics committee - additional information: Gangnam severance hospital \nIRB \nAuthor Disclosures:  \nSung Jun Ahn: Nothing to disclose \nHyunyeol Lee: Nothing to disclose \nJi Young Lee: Nothing to disclose \nWon Jun Son: Nothing to disclose \n \n \n \n \n \n \n \nInitial experience with 60kVp craniocervical CT ang iography: achieving \n0.2 millisievert while maintaining diagnostic perfo rmance via artificial \nintelligence iterative reconstruction \nY. Han¹, L. Peng², T. Meng², Q. Sun¹, G. Zhang², *T . Wang*², X. Wang¹; \n¹Jinan/CN, ²Shanghai/CN \n(tiantian.wang@cri-united-imaging.com) \n \nPurpose or Learning Objective: To describe the initial experience and \nevaluate the clinical feasibility of 60kVp cranioce rvical CT angiography (CTA) \nwith artificial intelligence iterative reconstructi on (AIIR). \nMethods or Background: Sixty consecutive patients scheduled for \ncraniocervical CTA were prospectively enrolled and underwent a 60kVp (350 \nref. mAs) and a followed 120kVp (120 ref. mAs) cran iocervical CTA, in a 5-\nminute interval, using two separate contrast medium  (CM) injections. The \n120kVp scans were reconstructed with hybrid iterati ve reconstruction (HIR), \nwhile the 60kVp scans were reconstructed with HIR a nd AIIR. Two radiologists \ndiagnosed the stenosis, the intracranial aneurysm ( IA), and the vascular \nanatomic variant (VAV) in consensus using a 5-point  confidence scale \n(1=definitely absent, 5=definitely present) on a pe r-patient basis, which was \nused for a receiver operating characteristic analys is, and evaluated the vessel \nvisibility (1=blur, 5=clear). Image noise on the co mmon carotid artery (CCA) \nwas measured. The diagnostic performance and image quality of 60kVp scans \nwere evaluated, using 120kVp scans as the reference  standard. \nResults or Findings: The mean effective dose and CM dosage was 0.18 ± \n0.04 mSv and 30.58 ± 5.15 ml, respectively, for 60k Vp acquisition, \ncorresponding to an 83.02% and 38.46% reduction as compared to 120kVp \nacquisition. Under 60kVp acquisition, AIIR outperfo rmed HIR in diagnosing all \nthree abnormal manifestations, showing higher AUC ( stenosis: 0.98 vs 0.57; \nIA: 0.92 vs 0.62; VAV: 0.92 vs 0.79; all p<0.05). N o significant difference in \nvessel visibility was found between 60kVp AIIR and reference images \n(4.32±0.97 vs 4.42±0.72, p=0.412), while AIIR images showed lower image \nnoise (10.14±4.64 HU vs 14.68±5.23 HU; p<0.05). \nConclusion: The 60kVp craniocervical CTA with AIIR has the pote ntial for \nprofound dose reduction without compromising the im age quality and the \ndiagnostic performance. \nLimitations: N/A \nFunding for this study: N/A \nEthics committee - additional information: This study was approved by the \nlocal IRB \nAuthor Disclosures:  \nQizhong Sun: Nothing to disclose \nTiantian Wang: Employee: at United Imaging Healthca re \nYicheng Han: Nothing to disclose \nGuozhi Zhang: Employee: at United Imaging Healthcar e \nLiying Peng: Nothing to disclose \nXiming Wang: Nothing to disclose \nTing Meng: Nothing to disclose \n \n \n12:30-13:30 Research Stage 1 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1405 \nArtificial intelligence in musculoskeletal \nimaging \n \nModerator \nS. Gitto; Milan/IT  \nAuthor Disclosures:  \nSalvatore Gitto: Other: Abiogen Pharma, Biolive Alp inion, Bracco Imaging \n \n \nDevelopment and validation of deep learning model f or screening low \nbone mineral density using chest radiographs: A mul ticentre, \nmultinational study \n*J. Song*, M. Kim, G. Lee, J. Jeong, S. J. Bae, J-M . Koh, N. Kim; Seoul/KR \n(jmsong@promedius.ai) \n \nPurpose or Learning Objective: This study aimed to develop and validate a \ndeep learning model for screening of patients with low bone mineral density \n(BMD) using chest radiographs (CXRs). \nMethods or Background: We retrospectively collected CXR data paired with \nDXA results from patients aged 50 and above from fi ve different resources. \nEach patient's BMD was classified using a T-score t hreshold of -1.0, with \nscores of -1.0 or above defined as ‘normal’ and tho se below as ‘low BMD’. Of \n\n \n \nFriday \nAbstract-based Programme \n \n 157  \nthe 57,589 CXRs from Hospital A, 55,600 were utiliz ed for training, and 1,989 \nwere used for internal validation. For external val idation, 3338, 938, and 295 \nCXRs were collected from B, C hospitals and D platf orm, respectively, \nrepresenting diverse patient demographics and clini cal backgrounds. A deep \nlearning model was developed to perform binary clas sification of patients' BMD \nas either normal or low, based on their CXRs. \nResults or Findings: In the A dataset, the model yielded an AUC of 0.95 and \ndemonstrated sensitivity of 0.97, specificity of 0. 65, and F1 score of 0.90. In \nthe datasets B, C, and D, the model achieved AUCs o f 0.91, 0.89, and 0.82. \nThe model’s sensitivity was 0.87, 0.88, and 0.64; s pecificity was 0.77, 0.71, \nand 0.85; and F1 score was 0.80, 0.88, and 0.76, re spectively. \nConclusion: The proposed low BMD screening system demonstrated \nperformance exceeding an AUC of 0.8 in all external  datasets, highlighting the \nrobustness of the system. Notably, the system showe d promising performance \neven on the D dataset, which comprised individuals of completely different \nracial backgrounds. This suggests the potential to promptly identify patients \nwith low BMD from CXRs, the most widely used imagin g modality globally. \nLimitations: First, this is a retrospective study. Second, a sub stantial \nproportion of the dataset comprises a single nation al population. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Ethics approval was obtained \nfrom the Ethics Committee of Asan Medical Center (N o. 2019-1226), the Public \ninstitutional Bioethics Committee (No.2024-0256-001 ) and the Ethics \nCommittee of Korea VHS Medical Center (No.2022-10-0 03-001). \nAuthor Disclosures:  \nJeongmin Song: Employee: Promedius Inc. \nNamkug Kim: Shareholder: Promedius \nMinjee Kim: Employee: Promedius Inc. \nJung-Min Koh: Advisory Board: Promedius \nSung Jin Bae: Nothing to disclose \nGaeun Lee: Employee: Promedius Inc. \nJinhoon Jeong: Employee: Promedius Inc. \n \n \nEvaluating the Impact of Artificial Intelligence on  Fracture Detection:  \nA Multinational Randomized Crossover Study on diagn ostic thinking \nefficacy \n*B. J. Van Der Zwart*¹, H. C. Ruitenbeek¹, M. Boese n², M. W. Brejnebol²,  \nG. Gunes³, K-G. A. Hermann⁴, K. Ziegeler⁴, E. Oei¹, J. J. Visser¹; \n¹Rotterdam/NL, ²Copenhagen/DK, ³Ortaca/TR, ⁴Berlin/DE \n(basvanderzwart1@hotmail.com) \n \nPurpose or Learning Objective: To assess the impact of AI assistance on \nfracture detection on conventional radiographs by c onducting a multi-country, \nmulticenter randomised crossover study. \nMethods or Background: Radiography data from 1,500 consecutive adult \ncases with suspected posttraumatic fractures were g athered along with \nrelevant clinical information, with 500 cases from each of three European sites. \nAll cases were read by senior and junior radiologis ts and orthopedic surgeons \nboth without and with AI assistance in two sessions  separated by at least four \nweeks. A reference standard was established by expe rt radiologists with the \nhelp of clinical data and follow-up imaging. The me an change in diagnostic \naccuracy was measured both per case (sensitivity an d specificity) and per \nfracture (sensitivity). \nResults or Findings: Sensitivity at the case-level increased with the AI  \nassistance for all reader groups with +0.074 for se nior radiologists, +0.181 for \njunior radiologists, +0.095 for both senior and jun ior orthopedic surgeons. The \nspecificity was negatively impacted with the AI ass istance for the radiologists \nwith changes of -0.010 for senior and -0.027 for ju nior radiologists, \nrespectively. The specificity increased for orthopa edic surgeons with +0.016 \nand +0.024 for senior and junior surgeons, respecti vely. The changes in \nsensitivity per fracture with the AI assistance wer e +0.089 and +0.168 for \nsenior and junior radiologists and +0.082 and 0.111  for senior and junior \nsurgeons, respectively. \nConclusion: Our study demonstrates that AI assistance enhances fracture \ndetection on conventional radiographs, yielding imp roved patient-wise \nsensitivity across participating centers. The chang es in specificity was positive \nfor orthopedic surgeons, but negative for radiologi sts. \nLimitations: The study was conducted in a simulated setting, pot entially \nimpacting reader performance. Additionally, the Haw thorne effect may have \ninfluenced reader behaviour, as participants were a ware they were being \nobserved. \nFunding for this study: This project has received funding from the European  \nUnion’s Horizon 2020 Research and Innovation Progra mme under grant \nagreement no. 954221. The results presented in this  work reflect only the \nviews of the authors. The Commission is not respons ible for any use that may \nbe made of the information it contains. \nEthics committee - additional information: This study was approved by the \ninstitutional review board of the Erasmus MC, Rotte rdam (Study ID: MEC-\n2021-0430) and the need to obtain informed consent was waived by the \ninstitutional review boards of Charité Universitäts medizin–Berlin (no. \nEA4/079/22) and the Danish Patient Safety Authority . \nAuthor Disclosures:  \nHuibert C. Ruitenbeek: Nothing to disclose \nEdwin Oei: Nothing to disclose \nMikael Boesen: Advisory Board: Radiobotics \nKatharina Ziegeler: Nothing to disclose \nGözde Gunes: Employee: Radiobotics \nBastiaan Johannes Van Der Zwart: Nothing to disclos e \nMathias Willadsen Brejnebol: Nothing to disclose \nKay-Geert A. Hermann: Nothing to disclose \nJacob Johannes Visser: Nothing to disclose \n \n \nPost implementation validation - False Positives an d Negatives in AI \nFracture Detection in Clinical Workflow: A Deep Div e \n*R. Sivanandan*¹, J. Vardal²; ¹Sandvika/NO, ²Dramme n/NO \n(drprabu@gmail.com) \n \nPurpose or Learning Objective: This study evaluates post-implementation \nmonitoring of an AI fracture detection algorithm, f ocusing on false positives and \nnegatives, their clinical implications, and mitigat ion strategies. It also assesses \nthe algorithm's impact on workflow and patient outc omes and further possibility \nof research studies. \nMethods or Background: The algorithm was implemented sequentially across \n5 hospitals in Vestre Viken Health Trust, Norway, w ith a follow-up validation \nconducted at the primary hospital using 1284 cases.  AI results were negative \n(60.8%), positive (37.9%) and doubtful (1.3%). Our new workflow after AI \nimplementation allowed AI-negative patients to be d ischarged, while AI-positive \ncases were referred to clinicians. Radiographers in itially reviewed AI results \nand guided patients flow, while radiologists report ing all cases. Orthopedic \nresidents and consultants subsequently examined the se images with AI \nresults. \nResults or Findings: Validation revealed 86% true negatives, 2% false \nnegatives, 7% true positives, 4% false positives, a nd 4% doubtful cases. These \nresults were consistent with preliminary external v alidation. Patients with AI-\npositive had false positives (5.6%) were primarily attributed to old fractures, \nskin folds and heterotopic calcifications, with min imal clinical significance. \nAmong the patient sent home with AI-negative result , false negatives (2.3%) \nmainly included minimal knee effusion in adults, be nign bone lesions, and \nsmall avulsion fractures requiring conservative tre atment. Only one patient with \nan avulsion fracture and drop finger was erroneousl y discharged and recalled \nfor clinical re-evaluation. \nConclusion: The validation results aligned with pre-implementat ion findings, \nwith false positives and negatives having minimal c linical impact. This study \nhighlights the potential for further research that is planned to evaluate the \nnecessity of radiologist reporting for AI-positive cases, given that clinicians \nalready review these images. \nLimitations: Few patients’ clinical results were not analyzed du e to restricted \naccess to patient journal. Few patients status afte r imaging were missing. \nFunding for this study: No funding \nEthics committee - additional information: Information was collected as per \nthe Ethics \nAuthor Disclosures:  \nJonas Vardal: Nothing to disclose \nRamprabananth Sivanandan: Nothing to disclose \n \n \nAssessing the Generalisability of a Paediatric Wris t Fracture Detection AI \nModel Using a Novel Dataset \n*C. Pauling*, O. Arthurs, B. Kanber, S. C. Shelmerd ine, E. Allan, E. Ashworth; \nLondon/UK \n(cato.pauling.22@ucl.ac.uk) \n \nPurpose or Learning Objective: The purpose of this study was to assess the \ngeneralisability of an artificial intelligence (AI)  model, trained on open-source \ndata, for the detection of fractures and other abno rmalities in paediatric wrist \nradiographs using a novel, external, and multi-cent ric dataset. \nMethods or Background: A novel retrospective case dataset was curated \nfrom two paediatric trauma centres in London, Engla nd. The dataset comprises \n865 images with a mean patient age of 10.4 ± 3.5 [standard deviation] years. \nGround truth annotations for the external test data set were established by \nconsensus opinion of at least two paediatric radiol ogists. To imitate real-world \nprospective data, no pre-processing was applied to the external data and only \ninvalid scans were excluded. A YOLOv7-X model was t rained on \nGRAZPEDWRI-DX, an open-source paediatric wrist trau ma radiograph \ndataset. After achieving an optimal performance on the test split of data, the \nmodel was used to perform inference on the novel ex ternal data and the \nperformance metrics were compared. \nResults or Findings: The sensitivity of the model for the detection of f ractures \nwas 89.0% on the test split of the open-source data . When evaluating on the \nnovel external data, the sensitivity decreased by 3 2.6%. The reduction in the \nperformance of the model across all detection class es was less severe, with a \n\n \n \nFriday \nAbstract-based Programme \n \n 158  \nchange to mean Average Position (mAP) of -0.139mAP@ 0.5 (-0.067 \nmAP@[0.5:0.95]). \nConclusion: The model failed to adequately generalise to an ext ernal dataset \nevidenced by a notable decline in fracture detectio n sensitivity. It is of critical \nimportance to ensure that AI models intended for us e in a prospective clinical \nsetting are externally validated. Additionally, dat a quality and pre-processing \nprocedures can significantly impact model performan ce. \nLimitations: The open-source training dataset contains annotatio ns for \nadditional pathologies which are not included in th e external test dataset. \nFunding for this study: CP is funded by the Great Ormond Street Hospital \nChildren’s Charity (GOSHCC) (Award Number: VS0618).  OJA is funded by an \nNIHR Career Development Fellowship (NIHR-CDF-2017-1 0-037). SCS is \nfunded by an NIHR Advanced Fellowship Award (NIHR-3 01322). \nEthics committee - additional information: Ethical approval was provided by \nthe National Health Service (NHS) Health Research A uthority (HRA) (IRAS ID: \n274278, REC reference 22/PR/0334) \nAuthor Disclosures:  \nEmily Ashworth: Nothing to disclose \nCato Pauling: Nothing to disclose \nSusan Cheng Shelmerdine: Nothing to disclose \nOwen Arthurs: Nothing to disclose \nBaris Kanber: Nothing to disclose \nEmma Allan: Nothing to disclose \n \n \nAI in radiological imaging of soft-tissue and bone tumours: a systematic \nreview evaluating against CLAIM and FUTURE-AI guide lines \nD. J. Spaanderman¹, M. Marzetti², X. Wan¹, A. Scars brook², E. Oei¹,  \nD. Grünhagen¹, *S. Klein*¹, M. P. A. Starmans¹; ¹Ro tterdam/NL, ²Leeds/UK \n(s.klein@erasmusmc.nl) \n \nPurpose or Learning Objective: Soft-tissue and bone tumours (STBT) are \nrare, diagnostically challenging lesions with varia ble clinical behaviours and \ntreatment approaches. This systematic review aims t o provide an overview of \nArtificial Intelligence (AI) methods using radiolog ical imaging for diagnosis and \nprognosis of STBT, highlighting challenges in clini cal translation, and \nevaluating study alignment with the Checklist for A I in Medical Imaging \n(CLAIM) and the FUTURE-AI international consensus g uidelines for \ntrustworthy and deployable AI to promote clinical t ranslation of AI methods. \nMethods or Background: The systematic review identified literature from \nseveral bibliographic databases, covering papers pu blished before 17/07/2024. \nOriginal research published in peer-reviewed journa ls, focused on radiology-\nbased AI for diagnosis or prognosis of primary STBT  was included. Exclusion \ncriteria were animal, cadaveric, or laboratory stud ies, and non-English papers. \nAbstracts were screened by two of three independent  reviewers to determine \neligibility. Included papers were assessed against the two guidelines by one of \nthree independent reviewers. (PROSPERO Registration : CRD42023467970) \nResults or Findings: The search identified 15,015 abstracts, and 325 art icles \nwere included for evaluation. Studies performed mod erately on CLAIM, \naveraging a score of 28∙9±7∙5 out of 53, but poorly on FUTURE-AI, averaging \n5∙1±2∙1 out of 30. \nConclusion: Imaging-AI tools for STBT remain at the proof-of-co ncept stage, \nindicating significant room for improvement. Future  efforts by AI developers \nshould focus on design (define unmet clinical need,  intended clinical setting \nand integration), development (build on previous wo rk, training with data \nreflecting real-world usage, explainability), evalu ation (addressing biases, \nevaluating using best practices), and data reproduc ibility and availability. \nFollowing these recommendations could improve clini cal translation of AI \nmethods. \nLimitations: Limitations include single-reviewer scoring due to the high volumn \nof literature included and assessment against as-of -yet unpublished, FUTURE-\nAI guidelines. However, FUTURE-AI were developed by  a large group of \ninternational medical AI experts. \nFunding for this study: Hanarth Fonds, ICAI Lab, NIHR, EuCanImage \nEthics committee - additional information: Not applicable to the systematic \nreview. \nAuthor Disclosures:  \nStefan Klein: Grant Recipient: Payment to instituti on. Stefan Klein is scientific \ndirector of the ICAI lab “Trustworthy AI for MRI”, a public-private research \nprogram partially funded by General Electric Health care. \nEdwin Oei: Grant Recipient: Payment to institution.  Edwin Oei co-leads a \nproject embedded in the ICAI lab “Trustworthy AI fo r MRI”, a public-private \nresearch program partially funded by General Electr ic Healthcare. \nAndrew Scarsbrook: Nothing to disclose  \nDouwe Jan Spaanderman: Nothing to disclose  \nMatthew Marzetti: Nothing to disclose  \nMartijn Pieter Anton Starmans: Nothing to disclose \nDirk Grünhagen: Nothing to disclose \nXinyi Wan: Nothing to disclose \n \n \nMulti-Center External Validation of an Automated Me thod Segmenting \nand Differentiating Atypical Lipomatous Tumors from  Lipomas Using \nRadiomics and Deep-Learning on MRI \nD. J. Spaanderman¹, S. Hakkesteegt¹, D. Hanff¹, C. Messiou², L. Nardo³,  \nD. Grünhagen¹, C. Verhoef¹, M. P. A. Starmans¹, *S.  Klein*¹; ¹Rotterdam/NL, \n²London/UK, ³Sacramento, CA/US \n(s.klein@erasmusmc.nl) \n \nPurpose or Learning Objective: Differentiating between lipomas and atypical \nlipomatous tumors (ALTs) on imaging is challenging,  often requiring biopsies. \nThis study aimed to externally and prospectively va lidate a radiomics model to \ndistinguish between lipomas and ALTs using MRI acro ss three large, multi-\ncenter cohorts. Additionally, the model was extende d with automatic and \nminimally interactive segmentation methods to impro ve clinical applicability. \nMethods or Background: Three cohorts were analyzed: two for external \nvalidation (US data from 2008–2018 and UK data from  2011–2017), and one \nfor prospective validation (Netherlands, 2020–2021) . Patient data, including \nMDM2 amplification status and MRI scans, were colle cted. An automatic \nsegmentation method was developed for T1-weighted M RI scans, with \ninteractive segmentation applied in case of poor qu ality. Radiomics model \nperformance was compared with that of two radiologi sts. \nResults or Findings: The cohorts included 150 (54% ALT), 208 (37% ALT), \nand 86 patients (28% ALT) from the US, UK, and Neth erlands, respectively. \nAutomatic segmentation succeeded in 78% of cases, w hile 22% required \ninteractive segmentation, with only 3% needing manu al adjustments. External \nvalidation yielded AUCs of 0.74 (95% CI: 0.66, 0.82 ) (US) and 0.86 (0.80, 0.92) \n(UK), and prospective validation achieved an AUC of  0.89 (0.83, 0.96) \n(Netherlands). The radiomics model performed simila rly to radiologists in all \ncohorts. \nConclusion: The radiomics model, combined with automated and mi nimally \ninteractive segmentation methods, effectively diffe rentiated between lipomas \nand ALTs, matching the performance of expert radiol ogists and potentially \nreducing the need for biopsies. \nLimitations: First, the segmentation workflow was performed by a  single \nclinician, hence the potential impact of different users on radiomics \nperformance was not assessed. Second, MDM2 amplific ation status, \ndetermined by core needle biopsy or resected specim ens, may include false \nnegatives, affecting the accuracy of the ground tru th. \nFunding for this study: This research was supported by an unrestricted gran t \nof Stichting Hanarth Fonds, The Netherlands. MPAS a nd SK acknowledge \nfunding from the research project EuCanImage (Europ ean Union's Horizon \n2020 research and innovation programme under grant agreement Nr. 95210). \nThis study was supported by the National Institute for Health Research (NIHR) \nBiomedical Research Centre at The Royal Marsden NHS  Foundation Trust and \nThe Institute of Cancer Research, London, and by Th e Royal Marsden Cancer \nCharity. The work was also supported by the In Vivo  Translational Imaging \nShared Resources with funds from NCI P30CA093373. T he views expressed \nare those of the author(s) and not necessarily thos e of the NIHR or the \nDepartment of Health and Social Care. \nEthics committee - additional information: The study protocol was approved \nby the local medical ethics review committee (MEC-2 020-0175), and performed \nin accordance with national and international legis lation. Informed consent was \nrequired and obtained exclusively from participants  in the prospective study \ncohort. For the training and external validation co horts, approval by the local \nmedical ethics review committee and the waiver of i nformed consent were \npreviously reported. \nAuthor Disclosures:  \nStefan Klein: Nothing to disclose \nChristina Messiou: Nothing to disclose \nDouwe Jan Spaanderman: Nothing to disclose \nMartijn Pieter Anton Starmans: Nothing to disclose \nLorenzo Nardo: Nothing to disclose \nDirk Grünhagen: Nothing to disclose \nStefanie Hakkesteegt: Nothing to disclose \nCornelis Verhoef: Nothing to disclose \nDavid Hanff: Nothing to disclose \n \n \nRadiopsy, quantitative wb-mri adc and fat fraction sequences for \ndiscrimination of smoldering multiple myeloma and m ultiple myeloma: a \nprospective observational study \n*G. Feliciani*¹, A. Rossi¹, C. Cerchione¹, E. Loi¹,  E. Antognoni¹, A. Cattabriga², \nM. Marchesini¹, D. Barone¹, A. Sarnelli¹; ¹Meldola/ IT, ²Bologna/IT \n(giacomo.feliciani@gmail.com) \n \nPurpose or Learning Objective: To distinguish between Multiple Myeloma \nand High-Risk Smouldering Myeloma at staging using image-based \nbiomarkers obtained from Whole Body-MRI (WB-MRI) Ap parent Diffusion \nCoefficient (ADC) and Fat Fraction (FF) sequences. \nMethods or Background: From January 2021 to March 2024, we enrolled \nconsecutive myeloma patients at staging into an pro spective trial and divided \nthem into Smouldering Multiple Myeloma (SMM) and Mu ltiple Myeloma (MM). \n\n \n \nFriday \nAbstract-based Programme \n \n 159  \nAll patients underwent WB-MRI. We use the term \"Rad iopsy\" to indicate the \nquantification and modelling of image characteristi cs nearby the biopsy site to \npredict patient status. A radiologist placed a cyli ndrical VOI nearby the biopsy \nsite and 5 more identical VOIs on distant sites suc h the pelvis bone and on \nD11 and L5 vertebrae. LASSO was used to select the most predictive features \nand build logistic regression models, which were th en validated using the test \nset. ROC curves were used as metrics for models’ pe rformance assessment. \nResults or Findings: The study included 102 patients (46 males, mean age  \n63 ± 12 [SD]) with 60 diagnosed with MM and 42 with  SMM. 144 quantitative \nfeatures were extracted from the VOI at the biopsy site WB-MRI ADC and FF \nsequences for each patient. Radiopsy model showed a  median AUC of 0.80 \n(0.75-0.90) in the training phase and a median AUC of 0.70 (0.55-0.80) in the \ntest phase. The best predictive model had an AUC of  0.95 and 0.75 in the \ntraining and test phase, respectively. The models u sed to predict patient status \nat biopsy site were also predictive in distant VOIs . \nConclusion: Conclusions: Radiopsy models can distinguish betwee n MM and \nSMM with good performance nearby the biopsy site. R adiopsy can be used to \npredict disease invasion on distant sites where bio psy is not possible or not \nfeasable \nLimitations: Single center \nFunding for this study: None \nEthics committee - additional information: protocol name: AccuMRI code: \nIRST 100.15 \nAuthor Disclosures:  \nDomenico Barone: Nothing to disclose \nClaudio Cerchione: Nothing to disclose \nAnna Sarnelli: Nothing to disclose \nEleonora Antognoni: Nothing to disclose \nArrigo Cattabriga: Nothing to disclose \nAlice Rossi: Nothing to disclose \nEmiliano Loi: Nothing to disclose \nMatteo Marchesini: Nothing to disclose \nGiacomo Feliciani: Nothing to disclose \n \n \nFeasibility of generating sagittal radiographs from  coronal images using \ndeep learning in adolescent idiopathic scoliosis \n*M. E. Pellegrino*¹, T. Bassani¹, A. Cina², F. Galb usera², A. Cazzato¹,  \nD. Albano¹, L. M. Sconfienza¹; ¹Milan/IT, ²Zurich/C H \n(maria.pellegrino@unimi.it) \n \nPurpose or Learning Objective: Minimizing radiation exposure is crucial in \nclinical monitoring of adolescent idiopathic scolio sis (AIS). Generative \nadversarial networks (GANs) have gained prominence in medical imaging due \nto their ability to learn complex patterns and gene rate high-quality synthetic \nimages by transforming one type of image into anoth er. This study explores \nGANs to generate synthetic sagittal radiographs fro m coronal views in AIS \npatients. \nMethods or Background: A retrospective dataset of 3,935 AIS patients with \nmild-to-moderate scoliosis (Cobb angle <45°) was an alyzed. The subjects \nunderwent radiographic spine and pelvis examination  using the EOS system, \nwhich acquires coronal and sagittal images simultan eously. The dataset was \nsplit into training (85%, n=3,356) and validation ( 15%, n=579). A pix2pix-based \nGAN model was trained to generate sagittal images f rom coronal views, \ntargeting real sagittal views. To evaluate accuracy , 100 subjects from the \nvalidation set were randomly selected for manual me asurement of lumbar \nlordosis (LL), sacral slope (SS), pelvic incidence (PI), and sagittal vertical axis \n(SVA) by two radiologists in both synthetic and rea l images. \nResults or Findings: Of the 100 synthetic images, 69 were deemed \nassessable. Intraclass correlation coefficient rang ed from 0.93 to 0.99 for \nmeasurements in real images and from 0.83 to 0.88 f or synthetic images. \nCorrelations between parameters in real and synthet ic images (mean values \nbetween raters) were 0.52 (LL), 0.17 (SS), 0.18 (PI ), 0.74 (SVA). Errors in \nparameters showed minimal correlation with Cobb ang le. The mean±SD \nabsolute errors were 7±7° (LL), 9±7° (SS), 9±8° (PI), 1.1±0.8 cm (SVA). \nConclusion: While the model generates sagittal images consisten t with \nreference images, their quality is not sufficient f or clinical parameter \nassessment, except for promising results in SVA, wh ich describes lateral \nplumb line alignment. \nLimitations: The quality of sagittal images is insufficient for assessing clinical \nparameters, except for SVA. \nFunding for this study: Italian Ministry of Health. \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nAndrea Cazzato: Nothing to disclose \nAndrea Cina: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nFabio Galbusera: Nothing to disclose \nMaria Elena Pellegrino: Nothing to disclose \nTito Bassani: Nothing to disclose \nDomenico Albano: Nothing to disclose \n \n12:30-13:30 Research Stage 2 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 1401 \nInsights into gastroesophageal disease \n \nModerator \nG. Bagnacci; Siena/IT  \n(giulio.bagnacci@unisi.it) \n \n \nPredicting Efficacy of Neoadjuvant Therapy in Esoph ageal squamous cell \ncarcinoma Using Spectral CT Parameters \n*J. Qu*, W. Li, F. Chu; Zhengzhou/CN \n(qjryq@126.com) \n \nPurpose or Learning Objective: To explore the differences in spectral CT \nparameters among patients with esophageal squamous cell carcinoma \n(ESCC), categorized by different tumor regression g rade (TRG) groups. \nMethods or Background: This prospective study analyzed 31 patients with \nESCC who underwent neoadjuvant therapy. All patient s received energy-\nspectrum CT scans within 7 days before surgery, and  the following parameters \nwere quantified: 40KeV, 70KeV, and 100KeV CT values  in the arterial (AP) and \nvenous phase (VP), iodine concentration (IC) in the  normal esophageal wall, \nlesion and aorta, and Z Effective values. Finally, we assessed the differences \nin parameters among the different TRG groups (0, 1,  2, 3). \nResults or Findings: In this study, a total of 19 participants were anal yzed. 6 \n(31.58%) were in the TRG0 group, 3 (15.79%) in the TRG1 group, 6 (31.58%) \nin the TRG2 group, and 4 (21.05%) in the TRG3 group . Significant differences \nwere observed among the four groups for arterial ph ase K40–70 keV (F=4.86, \nP=0.015) and K70–100 keV (F=4.78, P=0.016). However , no significant \ndifferences were found for NICD, NICratio, venous p hase K40–70 keV and \nvenous phase K70–100 keV (P>0.05). \nConclusion: Arterial phase parameters differ significantly amon g TRG groups, \nindicating potential in predicting tumor regression , while NICD, NICratio, and \nvenous phase parameters do not. \nLimitations: One limitation of this study is the small sample si ze, which may \naffect the generalizability of the findings. Additi onally, the statistical methods \nemployed were relatively simple, which may not full y capture the complexity of \nthe data. Future research should involve larger sam ple sizes and develop \nmodels to further evaluate the significant paramete rs, enhancing the accuracy \nand reliability of the results. \nFunding for this study: This study has received funding by the Projects of the \nGeneral Program of the National Natural Science Fou ndation of China  \n(No.82271979), Henan Province Medical Science and T echnology Tackling \nProgram Joint Construction Project (No. LHGJ2023009 6)，Henan Province \nCentral Plains Talent Program (Nurturing talent Ser ies)（No.20240220） \nEthics committee - additional information: The Research Ethics Committee \nof the Affiliated Cancer Hospital of Zhengzhou Univ ersity (Henan Cancer \nHospital) has approved this study (NCT03635619). Co nsent to participate \nWritten informed consent was obtained from all subj ects (patients) in this study. \nAuthor Disclosures:  \nWenshi Li: Nothing to disclose \nJinrong Qu: Nothing to disclose \nFuning Chu: Nothing to disclose \n \n \nDiagnostic accuracy and reliability of CT-based Nod e-RADS for \nesophageal cancer \n*J. Leonhardi*, B. Schnarkowski, T. Denecke, H-J. M eyer; Leipzig/DE \n(jakob.leonhardi@medizin.uni-leipzig.de) \n \nPurpose or Learning Objective: Staging in patients with oesophageal cancer \nis of high importance for treatment decisions. Rega rding correct nodal staging, \nCT imaging was reported with low accuracy. Our obse rvational retrospective \nstudy tried to elucidate the potential diagnostic b enefit of the Node-RADS \nclassification to discriminate benign from malignan t lymph nodes. \nMethods or Background: 126 patients (n= 15 females, 11.9%) with a mean \nage of 62.1 ± 10.4 years were included. Patients were surgically resected and \nlymph nodes were analyzed histopathologically. CT s cans were acquired within \none month before surgery. N = 182 lymph nodes were independently scored by \ntwo radiologists, following the Node-RADS classific ation. This lymph node \nscoring system ranges from 1 to 5, reflecting proba bility of malignancy (1–very \nlow to 5–very high). \nResults or Findings: 54 patients were nodal positive (42.9 %), whereas 7 2 \npatients were nodal negative (57.1 %). N1 stage was  found in n= 28 cases \n\n \n \nFriday \nAbstract-based Programme \n \n 160  \n(22.2 %), N2 in n = 20 cases (15.9 %) and N3 stage in n = 6 cases (4.8 %). \nThe tumors were squamous cell carcinomas (36 cases) , adenocarcinomas (88 \ncases) and mixed adenoneuroendocrine carcinomas (2 cases). Inter-reader \nagreement reached k = 0.65 (p<0.001) for the overal l Node-RADS scoring. \nMalignancy rates for Node-RADS-scores, ranging from  score 1 - 5, were as \nfollowing: 30 %, 14%, 81%, 90.1%, 86.5%. Total Node -RADS score showed \nstatistically significant differences between N0 an d N1-3 stage (N0: 2.68 ± 1.31 \nversus N1-3: 3.54 ± 1.11, p<0.001). ROC curve analy sis yielded an AUC of \n0.69, a threshold of 2 resulted in a sensitivity of  0.77 and a specificity of 0.55. \nConclusion: Node-RADS scores were associated with malignancy of  lymph \nnodes and might help to improve staging in oesophag us carcinomas. \nLimitations: For potential clinical translation there is need fo r external \nvalidation. \nFunding for this study: This study did not receive funding of any kind. \nEthics committee - additional information: No. of the approval: 106-16-\n14032016 \nAuthor Disclosures:  \nTimm Denecke: Nothing to disclose \nHans-Jonas Meyer: Nothing to disclose \nJakob Leonhardi: Nothing to disclose \nBenedikt Schnarkowski: Nothing to disclose \n \n \nThe value of deep learning image reconstruction alg orithm in improving \ndual-energy CT image quality of gastric cancer \n*T. Bei*, J. Li; Zhengzhou/CN \n(beitianxia@163.com) \n \nPurpose or Learning Objective: To investigate the value of deep learning \nimage reconstruction algorithm (DLIR) in improving image quality of the dual-\nenergy CT scans of gastric cancer by comparing with  the adaptive statistical \niterative reconstruction Veo (ASiR-V) algorithm. \nMethods or Background: The original DECT images of 80 patients with \nsurgical pathology confirmed gastric cancer between  February 2023 and July \n2023 were retrospectively collected. The virtual mo no-energy images (VMI) \nand iodine images at three-phase including arterial  phase, venous and delayed \nphase, were reconstructed using ASiR-V50% and DLIR of three strengths \n(DLIR‑L/M/H). Objective evaluation indicators for assessi ng image quality \nincluded noise (SD), signal-to-noise ratio of gastr ic cancer lesions (SNR lesion) \nand muscle signal-to-noise ratio (SNR muscle), cont rast-to-noise ratio of \nlesions (CNR lesion), and normalized iodine concent ration (NIC). Subjective \nevaluation indicators included image noise and imag e sharpness. One-way \nanalysis of variance and Kruskal-Wallis test were p erformed to compare the \ndifferences of image quality among different recons truction images. \nResults or Findings: Compared with ASiR-V50%, DLIR-M and DLIR-H \nsignificantly reduced the SD value of the images (P <0.05). The SNR lesion and \nSNR muscle of DLIR-M and DLIR-H were significantly higher than those of \nASiR-V50% (P<0.05). The CNR lesion of DLIR-H were s ignificantly higher than \nthose of ASiR-V50% and DLIR-L (P<0.01). But compare d with ASiR-V50%, \nthere were no statistically significant difference in the NIC values of the iodine \nimages between the groups (P>0.05). Subjective scor ing results showed that \nthe DLIR‑H images displayed the lowest noise, and the highes t image \nsharpness, significantly higher than ASIR-V50% (P<0 .05). \nConclusion: Compared with ASiR-V50%, DLIR significantly reduced  the noise \nof DECT images of gastric cancer and improved the i mage quality. Among the \nthree deep learning reconstruction algorithms , DLI R-H had the best noise \nreduction performance, image noise and image sharpn ess. \nLimitations: Not applicable. \nFunding for this study: National Natural Science Foundation of China \n(82202146) \nEthics committee - additional information: This is a retrospective study \nAuthor Disclosures:  \nJing Li: Nothing to disclose \nTianxia Bei: Nothing to disclose \n \n \nPhoton-Counting CT: Can Virtual Non-Contrast Match True Non-Contrast \nImaging? \n*G. Marras*, F. Pisu, A. Palmisano, A. Esposito; Mi lan/IT \n(marras.gloria@hsr.it) \n \nPurpose or Learning Objective: Virtual Non-Contrast (VNC) has emerged as \na potential alternative to True Non-Contrast (TNC) imaging, offering dose \nreduction while maintaining good tissue contrast. P hoton-Counting CT (PCCT) \nfor improved spectral separation should improve mat erial decomposition \nproviding accurate VNC. The aim of the present stud y is to evaluate the \ndiagnostic reliability of VNC compared to TNC in PC CT images. \n \n \n \n \nMethods or Background: This retrospective study included consecutive \npatients submitted to Contrast-Enhanced CT examinat ion acquired with PCCT \nNaeotom Alpha (Iopromide or Iopamidol 370, 60 ml) f rom June to September \n2024 for clinical indications with the availability  of TNC. VNC were obtained \nfrom arterial (VNCa) and venous phases (VNCv) (Qr40 , Q3). Using PACS tool, \nTNC, VNCa and VNCv were aligned and for the extract ion of Hounsfield Unit, \nRegions of interest (ROI) were drawn in right liver  lobe, spleen, kidney, \npancreas, aorta, omental and subcutaneous fat, para vertebral muscle, \nvertebral cortical and spongious bone. Co-registrat ion and minor adjustments \nwere made to account for breathing variations. Atte nuation values were then \ncompared using Tukey’s HSD or Dunn’s test. \nResults or Findings: A total of 38 studies have been analyzed. Significa nt \ndifferences in mean HU values between TNC and both VNC were observed for \nmost tissues (p<.01), except the spleen. Highest at tenuation differences were \nfound for subcutaneous/omental fat and bone cortica l and spongious with HU \ndifference higher than 15HU in more than 50% of cas es for fat and 100% for \nbone. No significant HU differences were found betw een VNCa and VNCv, \nexcept in the spleen. \nConclusion: VNC obtained from PCCT provide different attenuatio n values \ncompared to TNC. Caution should be exercised in rou tine clinical practice. \nLimitations: Higher sample size is required to confirm the prese nt data. \nFunding for this study: None \nEthics committee - additional information: Aimomics \nAuthor Disclosures:  \nAntonio Esposito: Nothing to disclose \nAnna Palmisano: Nothing to disclose \nGloria Marras: Nothing to disclose  \nFrancesco Pisu: Nothing to disclose \n \n \nSarcopenia effects on surgical outcomes and surviva l after gastrectomy \n(SESGa study) \nV. Sbacco, F. Puccetti, D. Palumbo, *G. J. Ortu*, A . Campisi, E. Mazza,  \nF. De Cobelli, U. Elmore, R. Rosati; Milan/IT \n(gabrielejacopoo@gmail.com) \n \nPurpose or Learning Objective: Sarcopenia, defined as a progressive and \ngeneralized loss of muscle mass, strength, and func tion, is linked to increased \nmorbidity and mortality in several nonmalignant con ditions. Recently, it has \nemerged as an important factor in oncology, correla ting with higher \nchemotherapy toxicity. However, its impact on survi val and surgical outcomes \nin cancer patients remains unclear. This study eval uates the effects of \nsarcopenia on the response to neoadjuvant chemother apy (NACT), overall \nsurvival, and postoperative outcomes in gastric can cer patients. \nMethods or Background: A retrospective analysis was conducted on a \nprospective database of patients treated between Se ptember 2017 and \nDecember 2023. All patients received NACT with the FLOT regimen followed \nby total or subtotal gastrectomy (open/laparoscopic ). Body composition \nparameters, including skeletal muscle area (SM), vi sceral (VAT), and \nsubcutaneous adipose tissue (SAT), were measured fr om CT images before \nand after NACT using a semi-automated segmentation software (Sliceomatic \n5.0). Sarcopenia was defined by the skeletal muscle  index (SMI) adjusted for \ngender. \nResults or Findings: The study included 101 patients, predominantly male  \n(60.4%), with distal adenocarcinoma (71.3%). Post-N ACT sarcopenia was \npresent in 65 patients (64.4%), while visceral obes ity was found in 33 (32.7%). \nNACT-induced sarcopenia had no significant effect o n postoperative \noutcomes, except for a higher incidence of wound co mplications (p=0.012). \nLongitudinal changes in body composition showed a c orrelation between \nsarcopenia and a poorer tumor regression grade (TRG ). \nConclusion: There is a strong association between NACT administ ration and \nthe onset of sarcopenia (p = 0,001) , which, howeve r, had minimal impact on \nsurvival and clinical outcomes in our center. This underscores the importance \nof completing the full neoadjuvant FLOT regimen. \nLimitations: Although our goal is to expand it, our cohort of pa tients is \ncurrently quite limited. Further exploration of sar copenia in gastric cancer is \nwarranted. \nFunding for this study: Unfunded study. \nEthics committee - additional information: 28/Int/2015 \nAuthor Disclosures:  \nFrancesco Puccetti: Nothing to disclose \nDiego Palumbo: Nothing to disclose \nGabriele Jacopo Ortu: Nothing to disclose \nElena Mazza: Nothing to disclose \nValentina Sbacco: Nothing to disclose \nAntonino Campisi: Nothing to disclose \nRiccardo Rosati: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \nUgo Elmore: Nothing to disclose \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 161  \n12:30-13:30 Research Stage 3 \nResearch Presentation Session: Cardiac \nRPS 1403 \nValves and pulmonary hypertension: the \nrole of cardiac CT and MRI \n \nModerator \nR. P. J. Budde; Rotterdam/NL  \n(r.budde@erasmusmc.nl) \nAuthor Disclosures:  \nRicardo P.J. Budde: Advisory Board: Bayer, payment to Erasmus MC; Board \nMember: ESCR; Research Grant/Support: Siemens, Hear tflow, Bayer, Bracco \nall payements to Erasmus MC; Speaker: Bayer \n \n \nMitral valve annulus assessment: a comparison betwe en 3D-TOE, CCT \nand surgical ring \n*E. Moliterno*, L. Giarletta, A. Pasquini, M. Masse tti, L. Natale, R. Marano; \nRome/IT \n(eleonora.moliterno@gmail.com) \n \nPurpose or Learning Objective: To compare the size assessment of the \nmitral annulus (MA) obtained by preoperative CCT wi th the intraoperative 3D-\nTOE in patients undergoing surgical mitral valve re pair (MVr) and assess their \nagreement with the MA measured intraoperatively by the surgeon. \nMethods or Background: 55 patients with severe primary mitral regurgitatio n \nand candidates to MVr by the Carpentier technique w ith annuloplasty \nunderwent pre-surgery CCT to exclude coronary arter y disease (CAD). We \ncompared CCT-derived MA sizing on short-axis view a ccording to the D-shape \nMV segmentation model obtaining inter-trigonal dist ance, septal-to-lateral and \ninter-commissural distances, with those obtained by  3D-reconstruction of intra-\noperative TOE and the surgical implanted ring sizes , using intraclass \ncorrelation. \nResults or Findings: Good agreement resulted between the inter-trigonal \ndistance measured by CCT and the surgical ring (ICC  0.89 [CI 0.330-0.985; p \n< 0.05]) and between inter-commissural distance obt ained by 3D-TOE and \nsurgical ring (ICC 0.81 [CI 0.458-0.936; p < 0.05]) , while the inter-trigonal \ndistance measured by 3D-TOE showed a moderate agree ment (ICC 0.63 [CI \n0.056-0.852; p < 0.05]). Excellent agreement result ed between the inter-\ntrigonal distance assessed by intraoperative 3D-TOE  and CCT (ICC 0.95 [CI \n0.755-0.989; p<0.05]). \nConclusion: Our study shows the good accuracy of the pre-proced ural CCT-\nbased MA size assessment in comparison with the int raoperative 3D-TOE and \nthe conventional surgical sizing, proposing CCT as a complementary non-\ninvasive imaging technique in predicting eligibilit y and surgical ring sizing in \npatients candidate to MVr. \nLimitations: The major study limitations are the relatively smal l sample size \nand the monocentric design. \nFunding for this study: None \nEthics committee - additional information: No additional information \nAuthor Disclosures:  \nMassimo Massetti: Nothing to disclose \nEleonora Moliterno: Nothing to disclose \nLuigi Natale: Nothing to disclose \nAnnalisa Pasquini: Nothing to disclose \nRiccardo Marano: Nothing to disclose \nLorenzo Giarletta: Nothing to disclose \n \n \nCT-based planning of transcatheter pulmonary valve implantation in \npatients operated for Tetralogy of Fallot and Doubl e Outlet Right \nVentricle \n*P. Marakhouskay*, P. Chernoglaz, K. Marakhouski, K . Drozdovski; Minsk/BY \n(pmarakh@gmail.com) \n \nPurpose or Learning Objective: The aim of the study was to quantify the \nmethod of planning transcatheter pulmonary valve im plantation (TPVI) in \npatients who have undergone surgery for congenital heart defects(CHD) by \nmeasuring in both ECG and non-ECG gated CTA sizes o f pulmonary \nartery(PA). \nMethods or Background: 22 patients with CHD as Tetralogy of Fallot and \nDouble outlet Right Ventricle(DORV), operated in in fancy, were included into \nstudy and divided into two groups: study group, in which direct intravascular \nballoon sizing of PA was performed, and control gro up in which TPVI was  \n \nrejected for excessive artery size without performi ng angiography. CTA of all \ncases were retrospectively analyzed and 4 sizing zo nes were measured: \nRVOT, supravalvular zone, middle-segment of and bif urcation zone of PA. \nAfter that calculations to find appropriate patient  -specific sizing methods by \ncomparing parameters with an actual stent size were  made (MedCalc Software \nLtd, Ostend, Belgium; https://www.medcalc.org;2022) . \nResults or Findings: Analysis showed no significant correlation between \ndiastolic RVOT sizing with any other measuring poin ts or an actual stent size \n(P=0,0004). Only supravalvular zone was accepted as  an independent residual \nin regression analysis (F-ratio 22, P=0,0004). Afte r implementing a regression \nequation to calculate appropriate size of PA stent on the control group, all the \nstents exceeded 30 mm (TPVI exclusion point). Mann- Whitney U-Test showed \nthat the significant difference between an actual s tent size and calculated size \nfor study group (U 91,00,P = 0,5678) and no signifi cant difference between \ncalculated sizes of two groups (U 0,50 P = 0,0004).  \nConclusion: Our study showed that any cardiac CT-based preTPVI landing \nzone sizing in diastolic phase can be based solely on measuring supravalvular \nzone of PA (r= 0,7992, F-ratio 22, P=0,0004), espec ially as an exclusion \ncriteria (U 0,50 P = 0,0004). \nLimitations: None \nFunding for this study: No funding \nEthics committee - additional information: Approved by the Ethics \nCommittee of the NACPS No. 11 22.03.24. \nAuthor Disclosures:  \nKirill Marakhouski: Nothing to disclose \nPalina Marakhouskay: Nothing to disclose \nKonstantin Drozdovski: Nothing to disclose  \nPavel Chernoglaz: Nothing to disclose \n \n \nLung Perfusion Changes Following Pulmonary Valve Re placement in \nRepaired Tetralogy of Fallot Patients: A Time-Resol ved MR Angiography \nStudy \nA. F. F. Tekin, *B. Tütüncüoğlu*, T. Banaz, Ö. Altun, Y. C. Kartal, S. Ozkök; \nIstanbul/TR \n(berk.tutun97@gmail.com) \n \nPurpose or Learning Objective: Tetralogy of Fallot (TOF) is a congenital \nheart disease surgically repaired in early life.Gui delines recommend pulmonary \nvalve replacement (PVR) for these patients. However , the effects of PVR on \nlung perfusion are not yet fully understood.In our study, we aimed to evaluate \nthe changes in lung perfusion following PVR in pati ents with surgically \ncorrected TOF. \nMethods or Background: The study included 10 patients with surgically \ncorrected TOF (M/F: 6/4, mean age: 16 years, median : 11-19 years).Lung \nperfusion was assessed using 4D time-resolved magne tic resonance \nangiography (MRA) before and 1 year after pulmonary  valve replacement \n(PVR) .Gadolinium-based contrast was administered a nd images were \nanalyzed with Philips Intellispace workstation.For measurement, ROI 's were \nmanually placed in the main pulmonary artery before  bifurcation, the aortic \narch, upper, middle, and lower lobes of the right a nd left lungs.Changes in both \nventricular parameters, pulmonary insufficiency rat io, and perfusion were \nassessed before and after PVR. Statistical analyses  were performed using \nSPSS .Data were reported as medians and the Wilcoxo n-sum rank test was \nused. \nResults or Findings: Following PVR, a reduction in pulmonary regurgitati on \nfraction and right ventricular end-diastolic volume  was observed, while no \nchanges were seen in right ventricular ejection fra ction and left ventricular \nvolume. After PVR, a difference was noted in right and left lung perfusion but \nno differences were observed between lung zones. \nConclusion: Lung perfusion assessment is a non-invasive, radiat ion-free \nimaging technique for the diagnosis and monitoring of various respiratory \ndiseases.Improved pulmonary artery flow volumes, re duced perfusion \nheterogeneity, and enhanced perfusion dynamics foll owing PVR reflect a \npositive impact on pulmonary hemodynamics.MRA-perfu sion imaging can be a \nvaluable tool for evaluating perfusion changes in t his patient population. \nLimitations: The planning of retrospective studies using a large r number of \npatient groups is necessary. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was approved by \nIstanbul Medeniyet University Göztepe Training and Research Hospital Clinical \nTrials Ethics Commitee \nAuthor Disclosures:  \nSerçin Ozkök: Nothing to disclose \nBerk Tütüncüoğlu: Nothing to disclose \nAli Fuat Fuat Tekin: Nothing to disclose  \nYiğit Can Kartal: Nothing to disclose \nTuba Banaz: Nothing to disclose \nÖmer Altun: Nothing to disclose \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 162  \nRole of CT-derived Extracellular Volume Fraction fo r Predicting \nResponse to Percutaneous Aortic Valve Replacement \n*C. Gnasso*, D. Vignale, A. Palmisano, S. Barbieri,  E. Agricola, A. Esposito; \nMilan/IT \n(gnasso.chiara@hsr.it) \n \nPurpose or Learning Objective: Aortic stenosis (AS) causes myocardial \nmicrostructural modifications: the development of i nterstitial fibrosis, in \nparticular, negatively impacts patients’ prognosis.  Echocardiography, the gold \nstandard for AS phenotypization, can’t investigate microstructural changes. CT \n-routinely performed for pre-procedural TAVI planni ng- can be implemented \nwith delayed scans to assess myocardial fibrosis as  ECV quantification. The \nprognostic role of ECV has already been demonstrate d. The study aims to \ninvestigate the capacity of ECV to improve risk str atification on top of \nechocardiographic classification (low-flow low-grad ient AS -LF-LG- vs high-\ngradient AS -HG-). \nMethods or Background: Prospective, single-center study (IRCCS OSR \nMilan), enrolling consecutive patients undergoing C T for TAVI planning \n(Oct2020-Mar2023); clinical, echocardiographic, and  laboratoristic data were \ncollected. Patients were categorized as LF-LG or HG . The CT protocol \nincluded a low-dose delayed scan (5 mins after iodi xanol 320 administration, \n85-110 mL based on patient’s BMI). ECV was calculat ed with the formula: (1-\nhematocrit)x(ΔHUmyocardium/ΔHUblood), with ΔHU being the HU-difference \nin pre and post-contrast scans. The composite endpo int (death; heart failure \nhospitalization) was collected after a 1-year follo w-up. \nResults or Findings: The final cohort consisted of 415 patients (82 year s [78-\n85]); 87 reached the endpoint. At Cox multivariable  analysis including clinical, \necho, and CT data, ECV (using the cut-off of 29% de rived from Youden index) \nresulted an independent prognosticator of the endpo int, along with sex, \nhypercholesterolemia, diabetes and transvalvular gr adient, with HR of 1.828 \n(95%CI1.151-2.903, p=0.0106), significantly higher in the LF-LG group \n(31.9(27.6-35.6) vs 28 (25.4;31), p<0.001). At Kapl an-Meier analysis, adding \nECV to echo classification led to an improvement in  risk stratification: HG-AS \npatients with higher ECV had a worse prognosis even  compared to LF-LG-AS \nwith low ECV (log-rank<.0001). \nConclusion: CT-ECV improves risk stratification in AS patients on top of \nechocardiographic evaluation. \nLimitations: Lack of MR as reference standard \nFunding for this study: None \nEthics committee - additional information: All patients signed informed \nconsent. \nprotocol number: CTMyoC 112/INT/2019 \nAuthor Disclosures:  \nDavide Vignale: Nothing to disclose \nAntonio Esposito: Nothing to disclose \nAnna Palmisano: Nothing to disclose \nSimone Barbieri: Nothing to disclose \nEustachio Agricola: Nothing to disclose \nChiara Gnasso: Nothing to disclose \n \n \nRelationship between myocardial strain and extracel lular volume: \nExploratory study in patients with severe aortic st enosis undergoing \nphoton-counting detector CT \n*C. Lisi*¹, V. Mergen², L. J. Moser², K. Klambauer² , H. Alkadhi², M. Eberhard²; \n¹Milan/IT, ²Zurich/CH \n(costanza.lisi@hotmail.it) \n \nPurpose or Learning Objective: Diffuse myocardial fibrosis and altered \ndeformation are relevant prognostic factors in aort ic stenosis (AS) patients. \nThe aim of this exploratory study was to investigat e the relationship between \nmyocardial strain, and myocardial extracellular vol ume (ECV) in patients with \nsevere AS with photon-counting detector (PCD)-CT. \nMethods or Background: We retrospectively included 77 patients with severe  \nAS undergoing PCD-CT imaging for transcatheter aort ic valve replacement \n(TAVR) planning between January 2022 and May 2024 w ith a protocol \nincluding a non-contrast cardiac scan, an ECG-gated  helical coronary CT \nangiography (CCTA), and a cardiac late enhancement scan. Myocardial strain \nwas assessed with feature tracking from CCTA and EC V was calculated from \nspectral cardiac late enhancement scans. \nResults or Findings: Patients with cardiac amyloidosis (n=4) exhibited \nsignificantly higher median mid-myocardial ECV (48. 2% versus 25.5%, \np=0.048) but no significant differences in strain v alues (p>0.05). Patients with \nprior myocardial infarction (n=6) had reduced media n global longitudinal strain \nvalues (-9.1% versus -21.7%, p<0.001) but no signif icant differences in global \nmid-myocardial ECV (p>0.05). Significant correlatio ns were identified between \nglobal longitudinal, circumferential, and radial st rains, and CT-derived left \nventricular ejection fraction (EF) (all, p<0.001). Patients with low-flow, low-\ngradient AS and reduced EF exhibited lower median g lobal longitudinal strain \n(GLS) values compared with those with high-gradient  AS (-15.2% versus -25.8, \np<0.001). In these patients, the baso-apical mid-my ocardial ECV gradient \ncorrelated with GLS values (R=0.33, p<0.05). \nConclusion: In patients undergoing PCD-CT for TAVR planning, EC V and \nGLS may enable to detect patients with cardiac amyl oidosis and patients with \nreduced myocardial contractility after myocardial i nfarction. Patients with low-\nflow, low-gradient AS and reduced EF showed lower m edian GLS correlating \nwith basal LV fibrosis. \nLimitations: Retrospective design, monocentric design, small sam ple number \nFunding for this study: No funding \nEthics committee - additional information: No additional information \nAuthor Disclosures:  \nCostanza Lisi: Nothing to disclose \nVictor Mergen: Nothing to disclose \nMatthias Eberhard: Nothing to disclose \nLukas Jakob Moser: Nothing to disclose \nKonstantin Klambauer: Nothing to disclose \nHatem Alkadhi: Nothing to disclose \n \n \nCardiac magnetic resonance 4d flow in mitral annulo plasty: impact on left \nventricular flow dynamics and functional correlatio ns with different types \nof devices \n*G. C. Pambianchi*, G. Cundari, L. Marchitelli, C. Catalano, N. Galea; Rome/IT \n(giacomo.pambianchi@gmail.com) \n \nPurpose or Learning Objective: To evaluate functional and fluidodynamic \nmodifications after MAP with different types of ann uloplasty prosthetics using \nCMR 4D-flow techniques and left atrial (LA) strain feature-tracking \nMethods or Background: We enrolled 12 patients treated with MAP (7 semi-\nflexible and 5 flexible incomplete rings), and had them undergo CMR at least \n12 months after the procedure. The protocol include d cineMR and 4D-Flow \nsequences with a whole heart coverage. We quantifie d TotalFlowVolume \n(TFV), PeakVelocity (Vmax), and WallShearStress (WS S) at LV inflow. We \nanalyzed LA volumes and strain. MAP were compared w ith 30 age- and sex-\nmatched healthy controls (10 studied with 4D-flow) \nResults or Findings: MAP patients had lower EF (49.3±4.2% vs62.34±6.3%; \np<0.021), reduced TFV (52.3±7.8ml vs69.6±9.7ml; p=0.039), and increased \nVmax (159.7±21.3cm/s vs125.1±35.3 cm/s; p=0.002) compared to controls; the \nWSS resulted comparable between the groups (0.23±0. 11 vs0.21±0.44; \np=0.354). As for the flow patterns analysis, intrav entricular and beneath the \nmitral valve flow, there was similar vortex formati on among the two groups. \nMAP patients had decreased Reservoir (20.6±20.1% vs 22.9±2.5%; p=0.033), \nlower Conduit (9.1±3.48% vs12.7±1.8%; p = 0.005), and increased Booster \nPump strain (12.4±1.8% vs8.9±2.3%; p=0.001) compared to controls. The time \nbetween the surgery and CMR was inversely correlate d with TVF (r: -0.95; \np=0.04) but didn't affect the WSS. \nConclusion: Mitral annuloplasty with leaflets preservation did not considerably \nalter intraventricular flow patterns compared to he althy controls; while the \nprosthetic ring causes slight stenosis, WSS isn't s ignificantly increased. Atrial \nfunction was preserved in MAP but still reduced if compared to healthy \ncontrols. Detailed evaluation of hemodynamic change s post-mitral repair and \npotential impact on the choice of annuloplasty devi ces and techniques to \noptimize long-term outcomes. \nLimitations: The study had a numerically limited sample, patient s didn't have \npre-annuloplasty CMR,. We didn't analyze TKE and vi scous energy loss. \nFunding for this study: Not applicable \nEthics committee - additional information: The study was submitted to local \nethics committee \nAuthor Disclosures:  \nGiacomo Carlo Pambianchi: Nothing to disclose \nGiulia Cundari: Nothing to disclose \nLivia Marchitelli: Nothing to disclose \nNicola Galea: Nothing to disclose \nCarlo Catalano: Nothing to disclose \n \n \nThe non-invasive right heart catheter: Hemodynamic classification of \npulmonary hypertension using 4D flow MRI \n*G. Reiter*, G. Kovacs, C. Reiter, H. Olschewski, M . Fuchsjäger, U. Reiter; \nGraz/AT \n(gert.reiter@siemens-healthineers.com) \n \nPurpose or Learning Objective: Mean pulmonary arterial pressure (mPAP), \npulmonary arterial wedge pressure (PAWP), and pulmo nary vascular \nresistance (PVR) are the hemodynamic parameters mea sured during right \nheart catheterization (RHC) for diagnosis and class ification of pulmonary \nhypertension (PH). This study aimed to assess the a ccuracy of 4D flow MRI in \npredicting these parameters non-invasively. \nMethods or Background: 103 patients with known or suspected PH (PH/non-\nPH, 77/26) prospectively underwent both RHC and who le-heart 4D flow MRI at \n3T. From 4D flow data, the duration of vortical blo od flow along the main \npulmonary artery (t_vortex), the left atrial accele ration factor (acc) and the \ncardiac output (CO_MR) were determined to derive es timates for mPAP via \nmPAP_MR = 16 + 0.63·t_vortex, PAWP via PAWP_MR = −6.2 + 10.1·a c c , a n d \n\n \n \nFriday \nAbstract-based Programme \n \n 163  \nPVR via PVR_MR = (mPAP_MR – PAWP_MR)/CO_MR. Relatio nships \nbetween invasive and 4D flow MRI-derived parameters  were analyzed by \ncorrelation analysis and t-test. The performance of  4D flow MRI-derived \nparameters to predict the hemodynamic classificatio n of PH was assessed by \nreceiver operating characteristic curve analysis. \nResults or Findings: The area under the curve (AUC) for predicting PH \n(mPAP > 20 mmHg) using mPAP_MR was 0.96. In PH pati ents, mPAP and \nmPAP_MR correlated strongly (r = 0.94) and showed n o significant bias (0.6 ± \n4.5 mmHg, p=0.24). Within PH patients, the AUCs for  predicting post-capillary \nPH (PAWP > 15 mmHg) from PAWP_MR and PVR > 2 WU fro m PVR_MR \nwere 1.00 and 0.95, respectively. PAWP and PAWP_MR correlated strongly (r \n= 0.94) and demonstrated no significant bias (0.4 ± 1.6 mmHg, p=0.05). PVR \nand PVR_MR correlated strongly (r = 0.85) but demon strated a significant bias \n(0.8 ± 2.2 WU, p<0.01). \nConclusion: 4D flow MRI allows accurate non-invasive diagnosis and \nhemodynamic classification of PH. \nLimitations: Single-center study \nFunding for this study: None \nEthics committee - additional information: Medical University of Graz, \nAustria \nAuthor Disclosures:  \nHorst Olschewski: Nothing to disclose \nUrsula Reiter: Nothing to disclose \nGert Reiter: Employee: Research and Development Sie mens Healthineers \nMichael Fuchsjäger: Nothing to disclose \nClemens Reiter: Nothing to disclose \nGabor Kovacs: Nothing to disclose \n \n \nFeasibility of pulmonary arterial pulse wave veloci ty assessment from 4D \nflow MRI \n*C. Reiter*, G. Reiter, G. Kovacs, D. Scherr, A. Sc hmidt, H. Olschewski,  \nM. Fuchsjäger, U. Reiter; Graz/AT \n(Clemens.Reiter@medunigraz.at) \n \nPurpose or Learning Objective: To assess the feasibility of measuring \npulmonary arterial pulse wave velocity (PWV) – a po tential prognostic marker \nin pulmonary hypertension (PH) – from 4D flow MRI. \nMethods or Background: Thirty-one healthy subjects (15 females; age, 60 ± \n10 years) and 10 patients with PH (6 females; age, 66 ± 11 years; mean \npulmonary arterial pressure, 46 ± 11 mmHg) were pro spectively recruited for \n4D flow MRI at 3T. PWV was calculated using the tra nsit-time approach \n(cvi42). Centerline segmentation of the pulmonary a rtery was performed twice, \nfrom the main pulmonary artery once to the left and  once to the right \npulmonary artery; PWV was measured in the main pulm onary artery as well as \nthe entire segmented vessels. The pulmonary artery cross-section area was \nobtained from a multiplanar-reconstructed plane thr ough the center of the main \npulmonary artery. Results were analyzed using t-tes ts and correlation analysis. \nResults or Findings: In healthy subjects, the main pulmonary arterial PW V \nwas 2.4 ± 0.2 m/s with no significant difference between the main-to-right (2.7 \n± 0.3 m/s) and main-to-left pulmonary PWV (2.7 ± 0.3 m/s, p=0.583). \nMoreover, no significant sex differences were obser ved (p=0.430). In PH \nsubjects the PWV was higher than in controls in the  main (6.1 ± 1.8 m/s, \np<0.001), the main-to-right (7.7 ± 3.0 m/s, p<0.001) and the main-to-left \npulmonary artery (7.1 ± 2.5 m/s, p<0.001). Main, main-to-right and main-to-left \npulmonary artery PWV correlated significantly with the average pulmonary \nartery cross-section area (r = 0.78, 0.61 and 0.59,  respectively; p<0.001 in all \ncases). \nConclusion: Pulmonary arterial PWV assessment from 4D flow MRI is \nfeasible. Normal ranges align with values reported from 2D flow \nmeasurements, show small variations in healthy subj ects, and differ compared \nPWV in patients with PH. \nLimitations: Small sample size. \nFunding for this study: OeNB Anniversary Fund 17934 \nEthics committee - additional information: Approval was obtained from the \nEthics committee of the Medical University of Graz,  Austria. \nAuthor Disclosures:  \nHorst Olschewski: Nothing to disclose \nUrsula Reiter: Nothing to disclose \nGert Reiter: Employee: Research & Developement Siem ens Healthineers \nMichael Fuchsjäger: Nothing to disclose \nDaniel Scherr: Nothing to disclose \nClemens Reiter: Nothing to disclose \nAlbrecht Schmidt: Nothing to disclose \nGabor Kovacs: Nothing to disclose \n \n \n \n12:30-13:30 Research Stage 4 \nResearch Presentation Session: Neuro \nRPS 1411 \nCharting the brain's next frontier: \nglymphatic system imaging \n \nModerator \nE. T. Tali; Ankara/TR  \n(turgut.tali@gmail.com) \n \n \nAssociation of Sleep Quality and Memory Performance  With Glymphatic \nFunction in Medical Interns and Residents \n*H. Wu*; Chongqing/CN \n(cefradine1005@gmail.com) \n \nPurpose or Learning Objective: Poor sleep quality is prevalent among \nmedical interns and residents, often characterized by insufficient sleep .Our \nobjective is to examine the associations between gl ymphatic function, sleep \nquality, and neuropsychological performance within this population. \nMethods or Background: 30 medical Interns and 15 residents on 2-week \nInternal Medicine and Radiology rotations were incl uded . Sleep profile was \naccessed using questionnaires and polysomnography. Wechsler Memory \nScale (WMS) was used for evaluation of the memory p erformance. The MRI for \neach subject was scheduled at least 72 hours after the completion of the night \nshift. Diffusion tensor imaging analysis along the perivascular space (DTI-\nALPS) index was used to evaluate glymphatic functio n. \nResults or Findings: Multivariate linear regression model for interns gr oup \ndetermined that intern year (unstandardized β=-0.0045 [SE 0.0001]; p < \n0.001), and the N3 sleep duration, index (unstandar dized β=0.0071 \n[SE=0.0004]; p = 0.0013) were independently associa ted with DTI-ALPS. \nIncreased DTI-ALPS was linked to enhanced memory sc ores (unstandardized \nβ = −0.77 [SE = 0.23]; p = 0.014). Interns group had  higher percentages of N1 \nand N2 sleep and a modestly higher amount of REM sl eep than residents \ngroup. The higher percentages of N1, N2, and REM sl eep were offset by a \nlower percentage of N3 sleep in interns. There was no significant difference of \nmemory scales between these two groups. \nConclusion: Interns tend to have poorer sleep structure compare d to \nresidents, likely due to their more frequent extend ed overnight shifts, whereas \nresidents rarely or never work such extended shifts . This study identified \nassociations between DTI-ALPS and memory performanc e, suggesting the \npotential of DTI-ALPS as a biomarker for evaluating  the neuropsychological \nstatus of young doctors. \nLimitations: Our study does not include longitudinal data on tem poral changes \nin DTI-ALPS between night shifts for individual par ticipants. \nFunding for this study: None \nEthics committee - additional information: The authors report no \ndisclosures relevant to the Abstract. \nAuthor Disclosures:  \nHao Wu: Nothing to disclose \n \n \nSmall vessel disease and glymphatic changes in hosp italized and non-\nhospitalized COVID-19 patients: A study on peak ske letonised mean \ndiffusivity and DTI-ALPS \nB. Genç, *A. Özçağlayan*, M. S. Buruk, L. Incesu, K. Aslan; Samsun/TR  \n(ozcaglayan.ali@gmail.com) \n \nPurpose or Learning Objective: COVID-19 has also been associated with the \ndevelopment of dementia, cortical atrophy, and cogn itive impairments such as \nbrain fog, which are challenging or often impossibl e to detect using \nconventional MRI. Peak width of skeletonized mean d iffusivity (PSMD), a \nrecently developed quantitative marker, has been pr oposed as a sensitive \nbiomarker for small vessel disease. The aim of this  study is to investigate \nchanges in PSMD, associated with small vessel disea se, and DTI-ALPS \nparameters, related to the glymphatic system, in CO VID-19 patients. \nMethods or Background: Clinical, demographic data, and MRI images were \nobtained from the \"neuroCOVID MRI dWIand fMRI with reversal learning\" \ndataset available on OpenNeuro (https://openneuro.o rg/datasets/ds005364/ \nversions/1.0.0 . After denoising and eddy current c orrections, DTI-ALPS and \nPSMD index measurements were performed similarly to  the previous literature. \nThe values of the hospitalized and non-hospitalized  COVID groups were \ncompared separately with the control group. \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 164  \nResults or Findings: The PSMD index was significantly higher in the \nhospitalized COVID-positive group (326 x10 ⁻⁶ mm²/s) compared to the control \ngroup (298 x10⁻⁶ mm²/s, p=0.028), while there was no significant di fference \nbetween the non-hospitalized COVID-positive group ( 301 x10⁻⁶ mm²/s) and the \ncontrol group (p=0.953). no significant differences  were observed in left or right \nDTI-ALPS values between the hospitalized, non-hospi talized COVID-positive \ngroups, and the control group. \nConclusion: Our study shows that hospitalized COVID-19 patients  exhibit an \nincrease in PSMD index, indicative of small vessel disease, while no such risk \nexists in non-hospitalized COVID-19 patients. Addit ionally, contrary to initial \nconcerns, COVID-19 does not appear to cause glympha tic dysfunction. \nLimitations: The use of the DTI-ALPS method to measure glymphati c function \nremains controversial. Perivascular spaces account for only 1% of cerebral \ntissue, making it difficult for this method to diff erentiate diffusion within \nperivascular spaces from diffusion along other axes . \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: We did not apply for ethics \napproval as the data utilized in this study were ob tained from the OpenNeuro \ndatabase. \nAuthor Disclosures:  \nMehmet Seyfi Buruk: Nothing to disclose \nBariş Genç: Nothing to disclose \nLütfi Incesu: Nothing to disclose \nKerim Aslan: Nothing to disclose \nAli Özçağlayan: Nothing to disclose \n \n \nChanges in cerebrospinal and vitreous fluid density  after iodine contrast \nadministration in acute ischemic stroke patients: p otential role of the \nglymphatic system? \nV. Khasminsky, J. Naftali, *G. Danieli*, E. Uriel; Tel Aviv/IL \n(guydanieli87@gmail.com) \n \nPurpose or Learning Objective: The aim of our study was to measure the \nchanges in density of the ventricular CSF and vitre ous fluid of the eyes in AIS \npatients who underwent emergent diagnostic CT work- up and endovascular \nthrombectomy. Our assumption was that iodine contra st medium leaked into \ninfarcted brain parenchyma would be cleared by the glymphatic system into \nCSF and would be detected there by CT. \nMethods or Background: A cohort of 119 subjects with the diagnosis of AIS \nwho underwent head NCCT, CT-angiography, CT-perfusi on, endovascular \nthrombectomy, and another follow-up head NCCT in up  to 36 hours was \nselected. The density of CSF in Hounsfield units (H U) was measured in the \nfrontal horns of lateral ventricles, in the third v entricle and in the vitreous body \nof both eyes on the admission head NCCT and compare d to the same \nmeasurements on the follow-up head NCCT. \nResults or Findings: Small but statistically significant increases in HU  were \nobserved across various locations on the follow-up CT. The average HU in the \nfrontal and third ventricles combined rose from 3.8 9±1.49 to 5.01±2.24, a \ndifference of 1.12±2.14 (P < 0.001). The average HU in both eyeballs rose \nfrom 5.93±2.75 to 7.14±4.18, a difference of 1.2±4.57 (P = 0.003). In the \nsubgroup with ASPECT score below 8 we found a diffe rence of 1.68±2.14 (P < \n0.001) in the ventricles and 3.16±5.29 (P < 0.011) in the vitreous. \nConclusion: Our findings can be explained in the framework of g lymphatic \nsystem theory by the clearence of leaked iodine via  the bulk flow of CSF in the \nperivenular space to the lateral ventricles. The de nsity increase in the vitreous \nfluid may be due to secondarily increased permeabil ity of blood-retinal barrier \nor connections between intra-cranial and intra-ocul ar glymphatic systems. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Uid: Rmc-0825-23 \nAuthor Disclosures:  \nGuy Danieli: Nothing to disclose \nVadim Khasminsky: Nothing to disclose \nJonathan Naftali: Nothing to disclose \nEitan Uriel: Nothing to disclose \n \n \nDiffusion tensor imaging along the perivascular spa ce (DTI-ALPS) for the \nglymphatic system evaluation in idiopathic and GBA- associated \nParkinson’s disease \nL. De Carolis, *R. Pascuzzo*, D. Aquino, S. A. Dell a Seta, R. Eleopra,  \nF. Doniselli, R. Cilia; Milan/IT \n(riccardo.pascuzzo@istituto-besta.it) \n \nPurpose or Learning Objective: Glymphatic system dysfunction has been \nassociated with neurodegeneration in Parkinson's di sease (PD). However, its \nrole in PD patients carrying GBA mutations (GBA-PD) , a key genetic risk factor \nfor PD, remains unclear. This study aimed to compar e glymphatic function in \nidiopathic PD (iPD) and GBA-PD using Diffusion Tens or Imaging Along the \nPerivascular Spaces (DTI-ALPS) and investigate corr elations with motor and \nnon-motor symptoms. \nMethods or Background: Brain DTI sequences (32 directions, b=1000 \ns/mm2) acquired with a 3T MRI scanner were retrospe ctively collected from 40 \nPD (20 iPD, 20 GBA-PD) and 40 age- and sex-matched patients with essential \ntremor (ET, serving as controls). DTI-ALPS indeces were computed for both \nhemispheres using Taoka’s method. Severity of motor  and non-motor \nsymptoms of PD patients were assessed by Hoehn and Yahr (H&Y) and \nUnified Parkinson’s Disease Rating scales (UPDRS). DTI-ALPS values were \ncompared between groups of patients using Mann-Whit ney test and were \ncorrelated with clinical variables using Spearman’s  (rho) correlation coefficient. \nResults or Findings: DTI-ALPS values were similar between iPD and GBA-\nPD patients. However, DTI-ALPS values in the hemisp here contralateral to the \nless affected side (DTI-ALPS-ctrl) were significant ly higher (p<0.001) in PD \npatients with H&Y stage <2 (1.42±0.07) compared to others (PD with H&Y \nstage ≥2: 1.26±0.18; ET: 1.27±0.19). In GBA-PD patients, DTI-ALPS-ctrl \nvalues were positively correlated with visual memor y (rho=0.726, p<0.001) but \nnegatively correlated with sleep disturbances (rho= -0.576, p=0.008), while in \niPD they were negatively correlated with age at ons et (rho=-0.591, p=0.006), \nworse response to dopaminergic medication (rho=-0.5 88, p=0.007), and total \nUPDRS score (rho=-0.591, p=0.006). \nConclusion: Glymphatic dysfunction is linked to disease severit y in PD and \nmay play distinct roles in the pathology of iPD and  GBA-PD. \nLimitations: The study limitations are the relatively small samp le size and the \nabsence of healthy controls. \nFunding for this study: Funding was provided by the Italian Ministry of Hea lth \nEthics committee - additional information: The study is retrospective \nAuthor Disclosures:  \nSara Adriana Della Seta: Nothing to disclose \nDomenico Aquino: Nothing to disclose \nFabio Doniselli: Nothing to disclose \nRoberto Eleopra: Nothing to disclose \nRoberto Cilia: Nothing to disclose \nLanfranco De Carolis: Nothing to disclose \nRiccardo Pascuzzo: Nothing to disclose \n \n \nDiffusion-Tensor MRI Study of the relationship betw een glymphatic \nsystem asymmetry and onset lateralization in Parkin son's disease \n*Z. Li*; Nanjing/CN \n(1600372734@qq.com) \n \nPurpose or Learning Objective: Parkinson's disease (PD) is commonly \ncharacterized by asymmetric motor symptoms. Previou s studies have \nimplicated abnormal α-synuclein aggregation as a key factor in PD \npathogenesis and suggested that glymphatic system d ysfunction may be \nresponsible for the disease progression. However, t he relationship between \nglymphatic system asymmetry and the side of motor s ymptom onset in PD \nremains unclear. \nMethods or Background: 27 left-onset PD (LPD) patients, 36 right-onset PD \n(RPD) patients, and 49 age- and sex-matched healthy  controls (HCs) were \nincluded in this study. The bilateral hemispheric A LPS indices were used to \nevaluate glymphatic function. Asymmetry of the glym phatic system was \nassessed by the asymmetry index (AI). Partial corre lation analysis was \nconducted to examine the relationship between impai red glymphatic system \nfunction and motor deficits. \nResults or Findings: Compared to HCs, RPD patients exhibited a significa nt \nreduction in the left ALPS index, while both left a nd right ALPS indices were \nsignificantly reduced in LPD patients. In both LPD patients and HCs, the right \nALPS index was lower than the left, suggesting a na tural leftward asymmetry. \nHowever, this asymmetry was diminished in RPD patie nts, as indicated by a \nreduced AI. Moreover, in RPD patients, the Unified Parkinson's Disease Rating \nScale Part III score showed a negative correlation with the left ALPS index, \nand with AI. \nConclusion: This study demonstrated PD patients with different onset sides \nhave different patterns of glymphatic system functi on. The glymphatic system \nasymmetry may provide new insights into the mechani sm underlying the \nlateralized onset of PD. \nLimitations: Firstly, being a retrospective cohort study conduct ed at a single \ncenter, it involved a relatively small sample size.  Secondly, further research is \nneeded to investigate the potential therapeutic imp lications of targeting the \nglymphatic system in PD \nFunding for this study: National Natural Science Foundation of China \n(81671258) Nanjing Medical University-Qilu Clinical  Research Fund Project \n(2024KF0254) \nEthics committee - additional information: This study was approved by the \nEthics Committee of the First Affiliated Hospital of Nanjing Medical University, \nand the requirement for written informed consent wa s achieved. IRB number: \nNo. 2014-SRFA-097. \nAuthor Disclosures:  \nZihan Li: Nothing to disclose \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 165  \nGlymphatic System Evaluation in Essential Tremor an d Parkinson’s \nDisease: an MRI Study Using Diffusion Tensor Imagin g Analysis Along \nPeri-vascular Spaces (DTI – ALPS) \n*A. Innocenzi*, M. Cella, G. Saltarelli, P. Badini,  F. Pistoia, A. Catalucci,  \nF. Bruno, E. Di Cesare, A. Splendiani; L'Aquila/IT \n \nPurpose or Learning Objective: This study aims to assess the presence of \nglymphatic system alterations in patients with Esse ntial tremor (ET) and \nParkinson’s disease (PD) through the analysis of di ffusion tensors along \nperivascular spaces (DTI-ALPS index). \nMethods or Background: We retrospectively evaluated 35 patients (19 PD, \n16 TE, 28 males, mean age 67 years) who were eligib le for Vim thalamotomy \nusing MRgFUS (2018-2022) and 17 healthy controls (1 3 female, mean age 39 \nyears). All MR images were obtained using a 3.0 Tes la MRI and a 32-channel \nhead coil. DTI images were analyzed using open-sour ce software. The ALPS \nindices were compared between the ET and PD patient  populations, \nrespectively. Both these populations were compared to healthy controls. \nResults or Findings: PD and ET groups were matched for disease duration,  \nage, gender and cognitive score. The ALPS index in PD patients showed a \nmean of 1.33, with no statistically significant differences compared to ET \npatients (1.31, p-value = 0.31), both reduced compa red to healthy controls \n(1.62; p-value < 0.001). We observed a more intense  tremor in ET patients \nrather than in PD patients (FTM 34,14 VS 28,64; p-v alue = 0.041). There was a \nsignificant correlation between the ALPS index and tremor intensity in ET \npatients (R = -0.76; p-value < 0.001). We did not f ind statistically significant \ncorrelations between ALPS –index and MoCA score in PD and TE groups. \nConclusion: There were no differences observed in the MRI indic es of the \nglymphatic system between ET and PD patients, and b oth are reduced with \nrespect to healthy subjects. Our results suggest th e need for further studies to \nbetter define the role of the ALPS index as a marke r of disease progression \nand to evaluate the possibility of neuro-degenerati ve pathophysiology in ET. \nLimitations: Small sample size. \nFunding for this study: This research received no external funding. \nEthics committee - additional information: Our study received approval from \nthe Internal Review Board of the University of L'Aq uila (protocol code 21 \nJanuary 2020 n. 01/2020), and all participating pat ients provided signed \ninformed consent to take part in the study. \nAuthor Disclosures:  \nAlessandra Splendiani: Nothing to disclose \nGaspare Saltarelli: Nothing to disclose \nAntonio Innocenzi: Nothing to disclose \nErnesto Di Cesare: Nothing to disclose \nFrancesca Pistoia: Nothing to disclose \nMarco Cella: Nothing to disclose \nAlessia Catalucci: Nothing to disclose \nFederico Bruno: Nothing to disclose \nPierfrancesco Badini: Nothing to disclose \n \n \nGlymphatic system evaluation in idiopathic intracra nial hypertension- \ncan ALPS index solve all issues of the disease? \n*R. Dahiya*, S. Rohilla; Rohtak/IN \n \nPurpose or Learning Objective: idiopathic intracranial hypertension (IIH) is a \ncommon headache disorder in young and middle age fe males. A few theories \npostulate impaired cerebral glymphatic clearance in  IIH, however there is a \npaucity of methods to quantify glymphatic activity in human brains. The \npurpose of this study was to use diffusion-tensor i maging to evaluate the \nglymphatic clearance of IIH patients and how it may  relate to various clinical \nparameters. \nMethods or Background: This observational cross-sectional study was \nconducted on 25 patients clinically diagnosed with IIH. DTI was used to \nseparately evaluate the diffusivity in lateral asso ciation and projection fibers, \nwith the degree of diffusivity used as a surrogate for glymphatic function \n(diffusion tensor image analysis along the perivasc ular space). Glymphatic \nclearance was correlated with several clinical metr ics, including lumbar \npuncture opening pressure and Frisen papilledema gr ade and combined \nconduit score for transverse sinus stenosis. \nResults or Findings: At an ALPS Index cut off of 1.64, ROC curve analysi s \nshowed sensitivity of 100%, specificity of 84.6%, P PV of 83.3%, NPV of 100%, \nDA of 91.3% and PLR of 6.5 with AUC of 0.973 demons trating excellent \ndiagnostic performance in differentiating severe fr om mild IIH. Also ALPS index \nshowed significant negative correlation with BMI(P= 0.012), CSF \npressure(p=0.032), papilledema(p=0.001) indicating increasing ALPS \nindex(i.e.worsened glymphatic clearance) with incre asing severity of disease. \nCSF pressure and papilledema showed positive correl ation(p=0.008). \nConclusion: Our results show that patients with IIH possess imp aired \nglymphatic clearance, which is directly related to the extent of clinical severity \nand ALPS Index can be used for clinical diagnosis a s a non invasive \ninvestigation. \nLimitations: Since we did not record response to treatment on mo st recent \nfollow up and its correlation with ALPS, we could a ssess whether ALPS can be \nused to guide treatment for IIH \nFunding for this study: Not provided \nEthics committee - additional information: The ethics committee notification \ncan be found under the  \nnumber BREC/22/231. \nAuthor Disclosures:  \nSeema Rohilla: Nothing to disclose \nRavi Dahiya: Nothing to disclose \n \n \nEvaluation of the glymphatic system in patients wit h intracranial tumours \nwith blood-brain barrier disruption \n*L-P. Schmill*, S. Seehafer, S. Aludin, S. Peters, O. Jansen, N. Larsen; Kiel/DE \n(lars-patrick.schmill@uksh.de) \n \nPurpose or Learning Objective: After decades of doubt, it has finally been \nproven that the human brain has its own glial-lymph atic system. Its role in \nvarious diseases has been studied and discussed. Al though its exact function \nis still unclear, perivascular spaces and fluid tra nsport by glial cells are thought \nto be the main components, and drainage appears to occur via the perivenous \nspaces and meningeal lymphatic vessels. These have been delineated by MRI \nfor the first time in recent years, but there is st ill a lack of established methods \nto assess their functionality. The aim of this stud y was to investigate whether \nintracranial tumours have increased drainage via ad jacent meningeal lymphatic \nvessels. \nMethods or Background: MRI scans were performed on patients with \nglioblastomas and cerebral metastases of different primary tumours. The \nprotocol included a fluid-attenuated inversion reco very sequence and a \ncontrast-enhanced T1-weighted black blood sequence,  which can be used to \ndelineate meningeal lymphatic vessels. This study l ooked specifically at the \ncontrast enhancement of these vessels and compared the signal intensities of \nlymphatics near the intracranial mass with those co ntralateral to it. \nResults or Findings: Results from the first 5 out of 30 patients show \nincreased visual contrast enhancement in the mening eal lymphatics adjacent \nto the tumour compared to the contralateral side. T hese results indicate an \nincreased lymphatic drainage of intracranial masses  in addition to their venous \ndrainage. \nConclusion: The current aim is to gain new insights into the gl ymphatic \nsystem and its relationship with the pathogenesis o f brain diseases. This study \nhas already provided evidence for increased glympha tic drainage in \nintracranial tumours. However, the exact drainage a reas and which sequences \nthat can be used for optimal detection of lymphatic  vessels are still unclear. \nLimitations: Not applicable \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Ethic approval was obtained from \nthe IRB of the Medical Faculty of the Christian-Alb rechts-University Kiel. \nAuthor Disclosures:  \nSönke Peters: Nothing to disclose \nLars-Patrick Schmill: Nothing to disclose \nSchekeb Aludin: Nothing to disclose \nSvea Seehafer: Nothing to disclose \nOlav Jansen: Nothing to disclose \nNaomi Larsen: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 166  \n14:00-15:30 Research Stage 1 \nResearch Presentation Session: \nEmergency Imaging \nRPS 1517 \nNovel approaches in emergency imaging \nstrategies \n \nModerator \nA. Ugarte; San Sebastian/ES  \n(augarte8689@gmail.com) \n \n \nRisk factors and prognostic significance of inferio r vena cava volume \ndefined by initial polytrauma CT-imaging \n*H-J. Meyer*, V. Sotikova, T. Denecke, M. Struck; L eipzig/DE \n(jonas90.meyer@web.de) \n \nPurpose or Learning Objective: The role of the inferior vena cava (IVC) \nvolume measurement in trauma patients is not yet fu lly understood. The aim of \nthe present study was to identify associations betw een the IVC volume and red \nblood cell (RBC) transfusion and massive transfusio n (≥10 RBC) within 24 \nhours after admission, as well as 24-hour and 30-da y mortality in trauma \npatients. \nMethods or Background: A retrospective analysis was conducted on all \nconsecutive trauma patients who required emergency tracheal intubation and \nmechanical ventilation before initial whole-body CT  imaging at a level-1 trauma \ncenter over a 12-year period. The IVC volume was de termined in the initial \ntrauma CT scan. \nResults or Findings: A total of 438 patients (75.3% male) with a median age \nof 50 years, and a median injury severity score (IS S) of 26 points were \nincluded in the analysis. Median IVC volume was 36. 25 cm3, and RBC \ntransfusion and massive transfusion were performed in 197 and 90 patients, \nrespectively. The 24-hour and 30-day mortality rate s were 7.3% and 23.3%, \nrespectively. IVC volume was found to be independen tly associated with the \nnecessity of RBC transfusion and 24-hour mortality (HR 0.98, 95% CI 0.96–\n0.99, p =0.01 and HR 0.96, 95% CI 0.93–0.99, p =0.0 25, respectively), while \nassociations with massive transfusion and 30-day mo rtality were not \nstatistically significant in multivariable analyses . \nConclusion: The initial IVC volume may serve as a predictor of the general \nneed for RBC transfusion, although it does not reac h the prognostic threshold \nfor massive transfusion. The association with 24-ho ur mortality rather than 30-\nday mortality suggests the possibility of its diagn ostic efficacy in short-term \noutcomes. \nLimitations: First, it is a single center retrospective study. S econd, there might \nbe bias induced by infusion therapy before the CT. \nFunding for this study: None \nEthics committee - additional information: Ethics committee at the Medical \nFaculty of Leipzig University, Leipzig, Germany (IR B00001750, project ID \n441/15ek, September 14, 2020) \nAuthor Disclosures:  \nTimm Denecke: Nothing to disclose \nManuel Struck: Nothing to disclose \nHans-Jonas Meyer: Nothing to disclose \nVeronika Sotikova: Nothing to disclose \n \n \nUnmasking the bleeding: post-traumatic CT-scan angi ography of pelvic \nring with and without pelvic belt \n*L. J. Pavan*, N. Ouamrane, G. Paesani, P-A. Ranc, K. Desalos,  \nT. Vivarrat-Perrin, M-E. Amoretti, N. Amoretti; Nic e/FR \n(lucajpavan@gmail.com) \n \nPurpose or Learning Objective: Patients who suffered a high energy trauma \nwith pelvic ring fracture usually come to whole-bod y CT-room with a pelvic \nbinder that improves outcome but may mask an active  bleeding during CT \ncontrast injection. Aim of study was to evaluate th e importance of removing \npelvic binder in the early imaging of pelvic ring f racture. \nMethods or Background: All consecutive post-traumatic whole-body CT-\nscans performed in our emergency department from Ja nuary 2022 to \nDecember 2023 were reviewed. Patients with pelvic r ing fracture and a 3-\nphase CT evaluation (non-contrast, late arterial an d portal venous) were \nincluded. CT-scan were directly performed without p elvic binder for \nhemodynamically stable patients. For hemodynamicall y unstable patients a \nfirst acquisition was performed with a tighten pelv ic binder, and if no bleeding \nwas found a second 2-phase contrast injection was r ealized after loosening the \npelvic binder to reveal any possible hidden bleedin g. \nResults or Findings: Out of 847 whole-body CT scan performed in the \nconsidered period, a total of 149 patients (87 men,  62 women, mean age \n43,6±17 years) with a pelvic ring fracture (78 Tile-A, 42 Tile-B, 29 Tile-C) were \nincluded. Seven patients were hemodynamically unsta ble, requiring a first CT \nexamination with tighten pelvic binder and a second  acquisition with loosen \npelvic binder. Of these, 2 patients (2/7, 29%) show ed a pelvic active bleeding \nonly with loosen pelvic binder. \nOf the 142 hemodynamically stable patients who dire ctly underwent CT without \nbinder, 10 presented an active bleeding. \nConclusion: Pelvic binders are a useful tool but may interfere with CT \nimaging, masking an active pelvic bleed. An acquisi tion with loosen pelvic \nbinder should always be performed in order to avoid  false negative \nexamination. \nLimitations: The retrospective design of the study. \nFunding for this study: None \nEthics committee - additional information: None since observational and \nretrospective \nAuthor Disclosures:  \nLuca Jacopo Pavan: Nothing to disclose \nKevin Desalos: Nothing to disclose \nNadine Ouamrane: Nothing to disclose \nThomas Vivarrat-Perrin: Nothing to disclose \nGaelle Paesani: Nothing to disclose \nNicolas Amoretti: Nothing to disclose \nPaul-Alexis Ranc: Nothing to disclose \nMarie-Eve Amoretti: Nothing to disclose \n \n \nSingle-pass split-bolus abdominal computed tomograp hy (CT) versus \nconventional biphasic CT in abdominal trauma patien ts \n*S. Gautam*, R. Gupta, A. Sharma; New Delhi/IN \n(shubhamgautam78866@gmail.com) \n \nPurpose or Learning Objective: To compare the image quality in single-pass \nsplit-bolus abdominal computed tomography (CT) and conventional biphasic \nCT in abdominal trauma patients. \nMethods or Background: 66 consecutive patients of abdominal trauma \nreferred for CT were randomized into two groups: th e study group (n = 33) \nscanned using the split-bolus technique; and the co ntrol group (n = 33) \nscanned using the conventional biphasic technique. CT image quality was \nanalyzed subjectively by two observers based on a 5 -point Likert scale. The \nimages were also analyzed quantitatively for attenu ation values achieved by \nregion of interest (ROI) placements in major arteri es, veins and solid organs. In \naddition, radiation dose in terms of Dose Length Pr oduct (DLP) was compared \nin the two groups. \nResults or Findings: The image quality in both groups ranged from good t o \nexcellent in most cases. There was no statistically  significant difference in \nsubjective image quality in both the groups as asse ssed by Likert score. \nAttenuation values in solid organs and major venous  structures were \nsignificantly higher in the split-bolus group (p <0 .001). Arterial attenuation \nvalues were significantly higher in the control gro up (p <0.001) but diagnostic \nlevels were achieved in all patients. There was a r eduction of 31.1% in DLP in \nthe split-bolus group. \nConclusion: Split-bolus technique offers comparable image quali ty and higher \nsolid organ and venous enhancement than conventiona l biphasic protocol at a \nreduced radiation dose. \nLimitations: The sample size was limited, excluding pediatric pa tients. The \nsplit-bolus protocol used a fixed 120 ml contrast d ose, unlike the weight-based \nregime in dual-phase CT. Most patients were young m ales, so the average \ncontrast in the dual-phase group was similar. Altho ugh whole-body CT is \nincreasingly used in trauma centers, we applied bot h protocols only for \nabdominal scans. \nFunding for this study: None \nEthics committee - additional information: Abstract is approved by \ninstitutional ethics committee \nAuthor Disclosures:  \nShubham Gautam: Author: 1st author \nAnuradha Sharma: Author: 2nd author \nRohini Gupta: Author: Corresponding author \n \n \nContrast Timing Pulmonary CT-Angiography: fixed tri gger delay in the \nascending aorta vs. pulmonary trunk \n*G. G. De Almeida*, O. Krzystek, J. Heimer, T. Niem ann, A. Euler; Baden/CH \n \nPurpose or Learning Objective: CT angiography (CTA) is the gold standard \nfor diagnosing pulmonary artery embolism (PE). Opti mal scan timing is crucial \nfor homogeneous enhancement of the pulmonary arteri es (PA). Clinical \ndifferentiation between PE and aortic dissection is  not always clear. This study \ntested whether bolus tracking in the ascending aort a provides sufficient \n\n \n \nFriday \nAbstract-based Programme \n \n 167  \ndiagnostic enhancement of the PA, whilst improving enhancement in the \nthoracic aorta in patients with suspected PE, with the objective of allowing \nadditional diagnostic information within the same e xam. \nMethods or Background: Retrospective image analysis of 200 patients \nscanned for PE between 03.2024 and 07.2024 was cond ucted. Patients were \nimaged using a third-generation dual-source CT with  application of 70 mL of \niodinated contrast medium and bolus tracking trigge ring in either the pulmonary \ntrunk (A) or the ascending aorta (B). A fixed trigg er delay of 7 seconds and \nautomatic tube voltage selection were applied. CT a ttenuation and contrast-to-\nnoise ratio (CNR) were measured at the pulmonary tr unk, main pulmonary \narteries, segmental superior lobe arteries, ascendi ng, and descending aorta. A \nmixed-effects model and post-hoc tests were applied . \nResults or Findings: No significant difference in CNR was found when \ncomparing both techniques for the pulmonary tree (a ll p>.05). Mean CNR at \nthe pulmonary trunk, main pulmonary artery, and seg mental pulmonary arteries \nwere 13.8, 12.9, 12.5 for group A and 13.9, 13.6, 1 3.6 for group B, \nrespectively. In the ascending and descending aorta , CNR was significantly \nhigher in group B when compared to group A (13.1 vs . 8.8 and 11.8 vs. 6.9; \nboth p<.001). \nConclusion: Bolus tracking in the ascending aorta showed simila r contrast \nenhancement of the pulmonary tree compared to trigg ering in the pulmonary \ntrunk while improving enhancement in the thoracic a orta, allowing \nsimultaneous evaluation of both regions. \nLimitations: Single center, rectrospective study. \nFunding for this study: No dedicated funding \nEthics committee - additional information: Rectrospective study with \nanonymised patient data. \nAuthor Disclosures:  \nJakob Heimer: Nothing to disclose \nOezlem Krzystek: Nothing to disclose \nTilo Niemann: Nothing to disclose \nGonçalo Garcia De Almeida: Nothing to disclose \nAndre Euler: Nothing to disclose \n \n \nContrast-Enhanced Ultrasound for Post-Traumatic Spl een Lesion \nAssessment: A Radiation-Free Diagnostic Alternative  \n*N. Finardi*, F. Cicchetti, E. Xhepa, C. Lanza, A. M. Ierardi, G. Carrafiello; \nMilan/IT \n \nPurpose or Learning Objective: Spleen injuries are common in abdominal \ntrauma. While FAST and contrast-enhanced CT are com monly used to detect \nparenchymal or vascular spleen lesions, contrast-en hanced ultrasound (CEUS) \nis emerging as a radiation-sparing alternative. CEU S can detect lesions within \nminutes using microbubble agents, providing a fast and safe method for \nmonitoring damage. This study aims to develop a CEU S protocol for \nmonitoring spleen trauma cases managed non-operativ ely. \nMethods or Background: A prospective, single-center observational cohort \nstudy was conducted at the Radiology Department of Fondazione IRCCS Ca’ \nGranda Ospedale Maggiore Policlinico in Milan. The study involves 28 patients \nwho presented to the emergency department with abdo minal trauma and were \ndiagnosed with parenchymal or vascular spleen injur ies managed non-\noperatively. Initial assessments included contrast- enhanced CT and \nsubsequent follow-up evaluations using contrast-enh anced ultrasound (CEUS) \nat 48 hours, 12-15 days, 30 days, and 60 days post- trauma to monitor lesion \nprogression and healing. \nResults or Findings: The primary endpoint was to evaluate the diagnostic  \naccuracy of CEUS in detecting complications followi ng non-operative \nmanagement of splenic trauma. Among 28 patients, 25  (89.3%) underwent \nembolization, with CEUS at 48 hours revealing splen ic infarcts in 10 (35.7%). \nNo increased free fluid, vascular lesions, or hemat oma progression cases were \nobserved in 22 patients (78.6%). \nConclusion: CEUS proved to be a reliable, radiation-free tool f or early \ndetection of complications in non-operatively manag ed splenic trauma. Its high \ndiagnostic accuracy, particularly at 48 hours, supp orts its use as a valuable \nalternative to CT in follow-up care. \nLimitations: Small population and short follow-up. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Pending approval by the ethics \ncommittee. \nAuthor Disclosures:  \nAnna Maria Ierardi: Nothing to disclose \nNiccolò Finardi: Nothing to disclose  \nFrancesco Cicchetti: Nothing to disclose  \nEdon Xhepa: Nothing to disclose  \nCarolina Lanza: Nothing to disclose \nGianpaolo Carrafiello: Nothing to disclose \n \n \n \n \nMultiplanar reconstruction of MDCT-images of the mi dface for CT-\nexophthalmometry: performance and reproducibility b etween different \npost-processing programs \n*U. G. Müller-Lisse*, K. Donij, S. G. Priglinger, S . Otto, C. R. Hintschich; \nMunich/DE \n(ullrich.mueller-lisse@med.uni-muenchen.de) \n \nPurpose or Learning Objective: Measuring ocular protrusion (OP) on MDCT-\nimages is particularly useful in cranio-facial trau ma. We step-wisely performed \nmulti-planar reconstruction of MDCT-images (MPR) fo r CT-exophthalmometry \nand investigated if results differ between physicia ns and between different \nMPR-programs. \nMethods or Background: One dentist and one radiologist independently \nreformatted MDCT-images from primary multidetector- row-CT reconstructions \nand measured inter-frontozygomatic base-length and OP, applying two \ncommercial and one freely available post-processing  programs for MPR, \nrespectively, in fifteen consecutive patients with cranio-facial trauma (five \nfemale, age 24-88 years), with ethics-committee app roval. Ease of patient-and-\nexam selection, 3D-reconstruction, alignment of orb its, fine adjustment, and \nmeasurements were rated 1-10 for each MPR-program. Wilcoxon-matched-\npairs-signed-ranks-tests and two-tailed Student-T-t ests for paired data \ncompared MPR-steps and distance-measurements, respe ctively (significance-\nlevel, p<0.05). \nResults or Findings: All MPR-programs allowed physician-generated MPR-\nreformatting and measurements of base-length and OP , although with different \nlevels of ease (range, 5-10). Results of distance m easurements varied by 0.2-\n0.4 mm, correlated highly (Pearson-r=0.9411-0.9956)  and did not differ \nsignificantly between different MPR-programs and di fferent observers with few \nexceptions. \nConclusion: CT-exophthalmometry results appear highly reproduci ble and \nstable between different MPR-programs when differen t physicians \nindependently reformatted CT-images and measured ba se-length and ocular \nprotrusion. \nLimitations: The study is limited to three exemplary post-proces sing \nprograms, two independent observers with different training background and a \nsmall number of consecutive patients. However, it d emonstrates that the \nprinciples of observer-performed multiplanr reconst ruction of MDCT-images \nand measurements relating to CT-exophthalmometry tr ansfer between different \npost-processing programs and Physicians with differ ent subspecialty training. \nFunding for this study: No funding has been obtained for this study. \nEthics committee - additional information: Ethics committee of the Medical \nFaculty, LMU Ludwig-Maximilians-Universität München , Vote No. 20-633 KB \nAuthor Disclosures:  \nKathleen Donij: Nothing to disclose \nSiegfried Georg Priglinger: Nothing to disclose \nChristoph Rudolf Hintschich: Nothing to disclose \nUllrich G. Müller-Lisse: Nothing to disclose \nSven Otto: Nothing to disclose \n \n \nAnalyzing the Prevalence of Injury and Violence in Transgender Females \nUsing Radiology Reports \n*R. Chopra*, K. Patel, B. Rosner, O-P. Hamnvik, B. Khurana; Boston, MA/US \n(rrchopra@bwh.harvard.edu) \n \nPurpose or Learning Objective: Given the high risk of violence with \nsignificant underreporting among transgender and ge nder diverse patients, this \nstudy aims to investigate the prevalence and dispar ities in injuries and potential \nviolence between transgender female and cisgender f emale patients by \nanalyzing radiology reports. \nMethods or Background: We utilized our hospital's Research Patient Data \nRegistry to identify 263 transgender female patient s and 525 age, race, and \nethnicity-matched cisgender women. Adjusted inciden ce rate ratios (aIRR) and \nOdds ratios were calculated to compare imaging and injury patterns. Two \nradiologists blinded to the study's purpose assesse d the likelihood of intimate \npartner violence (IPV) based on radiology reports. EMRs were reviewed for \nviolence documentation in all patients with radiolo gically evident injuries. \nResults or Findings: In our cohort, 25.4% (67/263) of cases sustained 14 1 \ninjuries, compared to 14.7% (77/525) of controls wh o sustained 98 injuries. \nInjury rates were higher in cases (aIRR: 3.3 [2.5-4 .3] P<0.0001), particularly for \ncranial (7.8 [2.1-29.1] P<0.0001), facial (36.4 [8. 6-153.8] P<0.0001), and \nthoracic injuries (4.9 [1.4-17] P=0.01), with 78.9%  of facial fractures (15/19) \ninvolving the midface. The percentage of imaging st udies in emergency \ndepartments was significantly higher in the cases t han in the controls (OR = \n5.3 [3.3, 8.3]) (P<0.0001). Radiologists suspected IPV in 12 cases and 1 \ncontrol, with 75% of cases confirming violence and 50% reporting IPV. A \nhigher number of cases with radiologically evident injuries reported \nexperiencing IPV (OR 6.5; [2.7-15.9]; P<0.0001) com pared to controls. \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 168  \nConclusion: Transgender females experience significantly higher  injury rates, \nparticularly to the head, face, and chest, with fre quent presentations to \nemergency departments, indicating an elevated risk of violence and gaps in \npreventive care. By recognizing these patterns, rad iologists can help identify \nat-risk patients and facilitate timely IPV screenin g and support. \nLimitations: Retrospective single-institution, self-reporting by  patient. \nFunding for this study: National Institute of Biomedical Imaging and \nEngineering (NIBIB), National Institute of Health \nEthics committee - additional information: Approved by Mass General \nBrigham IRB \nAuthor Disclosures:  \nKrishna Patel: Nothing to disclose \nBharti Khurana: Nothing to disclose \nBernard Rosner: Nothing to disclose \nRohan Chopra: Nothing to disclose \nOle-Petter Hamnvik: Nothing to disclose \n \n \nAbbreviated MRI in traumatic injury of spine \n*C. Loberg*, L. Küsters, A. Gisevius, D. Roggenland , E. Yilmaz,  \nT. Schildhauer, C. Kruppa, M. Aach; Bochum/DE \n(Christina.Loberg@bergmannsheil.de) \n \nPurpose or Learning Objective: Traumatic ligamentous injury is common in \ntrauma of spine and can be overlooked easily. Patie nts outcome is based on \nfast diagnosis and surgery. MRI is the standard of care in evaluation of \nligamentous injury. We invested if an abbreviated p rotocol (AP) consisting only \nof one T2wSTIR acquisition is suitable for detectio n of ligamentous injury. \nMethods or Background: A cohort of 100 patients with underlying spinal \ntrauma who underwent MRI were selected. Full MRI pr otocol (FP) comprised \nT1w, T2w, T2wSTIR sequences at 1.5T. For abbreviate d MRI protocol (AP) we \nchose a rapid protocol that ensures the acquisition  of maximum of contrast and \nmaximum spatial-resolution images based on T2wSTIR acquisition. Two \nradiologist with 5 years and 16 years of experience  reviewed the images to \ncharacterize ligaments, fracture, spinal instabilit y and traumatic disc herniation. \nResults or Findings: MRI acquisition for FP was 23 minutes versus 4.12 f or \nthe AP. Average time to read the single T2wSTIR and  complete FP was 3.4 \nminutes versus 14.3 minutes. 42 ligamentous injurie s were detected. \nSpecificity and positive predictive value (PPV) of AP versus FDP were \nequivalent (96.5% to 94.2% and 23.9% v 23.2%). \nConclusion: An MRI acquisition of 4.20 minutes and expert radio logist reading \ntime of 3.40 minutes are sufficient to confirm liga mentous injury of spine. With \na reading time < 4 minutes diagnostic accuracy was equivalent to that of the \nFDP. \nLimitations: This was a retrospective single center study. \nFunding for this study: There was no funding \nEthics committee - additional information: Ethic Comittee Ruhr University \nHospital Bochum 178/ 24 \nAuthor Disclosures:  \nEmre Yilmaz: Nothing to disclose \nAstrid Gisevius: Nothing to disclose \nThomas Schildhauer: Nothing to disclose \nMirko Aach: Nothing to disclose \nLeonie Küsters: Nothing to disclose \nChristina Loberg: Nothing to disclose \nDaniela Roggenland: Nothing to disclose \nChristiane Kruppa: Nothing to disclose \n \n \nDiagnostic accuracy and time efficiency of a novel deep learning \nalgorithm for the assessment of intracranial hemorr hage \n*C. Booz*¹, T. Vogl¹, V. Koch¹, L. D. Gruenewald¹, A-I. Nica¹, T. D'Angelo²,  \nM. Dimitrova¹, G. M. Bucolo¹, I. Yel¹; ¹Frankfurt/D E, ²Messina/IT \n \nPurpose or Learning Objective: To evaluate diagnostic accuracy and time \nefficiency of a deep learning-based pipeline using a Dense U-net architecture \nfor the assessment of intracranial hemorrhage (ICH)  in unenhanced head CT \nscans. \nMethods or Background: This retrospective study included 1004 CT scans of \n1004 patients (mean age, 71 ± 11 years; 496 men and  508 women) who had \nundergone an unenhanced head CT scan for the assess ment of ICH. All CT \nscans were analyzed by the algorithm and a board-ce rtified radiologist \nindependently for the presence of ICH. In case of I CH presence, ICH had to be \ndefined as intraparenchymal hemorrhage (IPH), intra ventricular hemorrhage \n(IVH), subarachnoid hemorrhage (SAH), subdural hemo rrhage (SDH) and \nepidural hemorrhage (EDH). Additionally, the time u ntil first temporary \ndiagnosis of ICH was measured. Three experienced bo ard-certified radiologists \nanalyzed the CT scans in consensus reading sessions  to establish the \nstandard of reference for hemorrhage presence and c lassification. \nResults or Findings: The reference standard revealed a total of 1108 dif ferent \nICH presences (IPH, n=344; IVH, n=52; SAH, n=326; S DH, n=356; EDH, \nn=30). The algorithm showed a high diagnostic accur acy for the assessment of \nICH with a sensitivity of 92%, specificity of 95% a nd an accuracy of 93%. \nConcerning the most frequently present different IC H types in this study, the \nsensitivity was 92%, 93% and 93% (IPH, SAH and SDH,  respectively), and the \nspecificity was 95%, 96% and 95% (IPH, SAH and SDH,  respectively). \nRegarding analysis time, the algorithm was signific antly faster compared to the \ntemporary report of the assigned radiologist (16 ± 3 s vs 273 ± 11 s, p < 0.001). \nConclusion: A novel deep learning algorithm provides high diagn ostic \naccuracy combined with time efficiency for the iden tification and classification \nof ICH in unenhanced CT scans. \nLimitations: Single-center retrospective study \nFunding for this study: No funding was received. \nEthics committee - additional information: The local IRB approved this \nstudy. \nAuthor Disclosures:  \nChristian Booz: Speaker: Siemens Healthineers \nLeon D. Gruenewald: Nothing to disclose \nIbrahim Yel: Speaker: Siemens Healthineers \nMirela Dimitrova: Nothing to disclose \nThomas Vogl: Nothing to disclose \nVitali Koch: Nothing to disclose \nAndreea-Ioana Nica: Nothing to disclose \nTommaso D'Angelo: Speaker: Bracco Speaker: Philips \nGiuseppe Mauro Bucolo: Nothing to disclose \n \n \nBeyond ASL: SWI as the Superior Diagnostic Tool for  Status Epilepticus \nWhen Conventional MRI Falls Short \n*S. K. Kondapavuluri*, S. K. Patan, R. P. Yadav; Vi jayawada/IN \n \nPurpose or Learning Objective: This study evaluates the diagnostic utility of \nsusceptibility-weighted imaging (SWI) in status epi lepticus (SE), focusing on its \nrole in identifying cerebral perfusion and oxygenat ion changes, particularly \nwhen conventional MRI sequences (diffusion, T2, FLA IR) appear normal. \nSWI’s superior resolution for detecting subtle veno us and metabolic changes \nmakes it a better alternative to arterial spin labelling (ASL) in certain clinical \nsettings. The study aims to demonstrate SWI’s value  as both a practical and \nsuperior tool in perfusion imaging. \nMethods or Background: This observational study involved 50 patients with \nconfirmed SE who underwent MRI within 6 hours of se izure onset. The imaging \nprotocol included diffusion-weighted imaging (DWI),  fluid-attenuated inversion \nrecovery (FLAIR), SWI, and ASL. SWI patterns of ven ous oxygenation were \ncompared with ASL perfusion maps to assess concorda nce in detecting hyper- \nand hypo-perfused regions, particularly in cases wh ere conventional MRI \nsequences were unremarkable. \nResults or Findings: SWI revealed distinct patterns of cerebral venous \nchanges, categorised into four groups: Group 1 (20 patients) exhibited \ngeneralised hyperperfusion on ASL with globally dim inished cortical veins on \nSWI, indicating global hyperoxygenation. Group 2 (1 3 patients) showed focal \nhyperperfusion with focally diminished cortical vei ns, reflecting focal \nhyperoxygenation. Group 3 (10 patients) displayed f ocal hyperperfusion with \nfocally prominent cortical veins due to focal deoxy genation. Group 4 (7 \npatients) demonstrated generalised hyperperfusion w ith globally prominent \nveins, indicating global deoxygenation. \nConclusion: SWI is a valuable tool for detecting oxygenation ch anges in SE, \neven when conventional MRI sequences appear normal.  Its superior resolution \nand availability make it a better alternative for d etecting subtle venous and \nmetabolic changes, supporting its broader adoption as a complementary or \nstandalone tool in epilepsy management. \nLimitations: Not applicable \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Ethics approval for this \nobservational study was obtained from the relevant ethics committee. Informed \nconsent was secured from all participants, ensuring  compliance with ethical \nstandards for research involving human subjects. Th e study adhered to the \nprinciples outlined in the Declaration of Helsinki.  \nAuthor Disclosures:  \nSushen Kumar Kondapavuluri: Nothing to disclose \nSharuq Khan Patan: Nothing to disclose \nRatan Pal Yadav: Nothing to disclose \n \n \nPrognostic Indicators of Conservative Treatment Fai lure in Adhesive \nSmall Bowel Obstruction: Insights from CT Imaging \n*A. Ammirabile*, E. Desiato, A. M. A. Lucia, S. Giu dici, M. Francone,  \nD. Del Fabbro, E. Lanza; Milan/IT \n(angela.ammirabile@humanitas.it) \n \nPurpose or Learning Objective: This study aimed to identify the CT imaging \nfeatures associated with the failure of conservativ e management using oral \nwater-soluble contrast medium in patients presentin g to the Emergency Room \nwith Adhesive Small Bowel Obstruction (ASBO). \n\n \n \nFriday \nAbstract-based Programme \n \n 169  \nMethods or Background: This retrospective single-center study included all  \nconsecutive patients admitted to the ER from Februa ry 2019 to February 2023 \nwith ASBO, who underwent contrast-enhanced CT at di agnosis and received \nconservative treatment. The assessed CT findings we re type and location of \nthe transition zone, ASBO severity, presence of fat  notch sign, beak sign, small \nbowel feces sign, peritoneal free fluid, and pneuma tosis intestinalis. \nUnivariable and multivariable logistic regression a nalyses were performed to \nevaluate the association between these radiological  parameters and treatment \noutcomes. \nResults or Findings: Among the 106 patients included (median age 74.5 \nyears), conservative management succeeded in 59 cas es (55.7%), while 47 \npatients (44.3%) required surgery after initial non -operative treatment failure. \nFailure was more common in patients with previous A SBO episodes (p = 0.03), \nfemale gender (p = 0.04), and was associated with a  longer hospital stay (p < \n0.001). At multivariable analysis, the fat notch si gn (OR = 2.95; p = 0.04) and \nthe beak sign (OR = 3.42; p = 0.04) were significan tly associated with \nconservative treatment failure. \nConclusion: Two CT features - the fat notch sign and the beak s ign - were \nsignificantly correlated with the failure of conser vative management in ASBO. \nThese findings highlight the importance of an early  identification of patients \nwho may benefit from undelayed surgical interventio n. \nLimitations: The limitations of the study are the retrospective,  single-center \ndesign and the evaluation of a limited number of CT  signs and laboratory \nvalues. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective. \nAuthor Disclosures:  \nElena Desiato: Nothing to disclose \nAngela Ammirabile: Nothing to disclose \nAda Maria Antonella Lucia: Nothing to disclose \nEzio Lanza: Nothing to disclose \nDaniele Del Fabbro: Nothing to disclose \nMarco Francone: Nothing to disclose \nSimone Giudici: Nothing to disclose \n \n \n14:00-15:30 Research Stage 2 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1505 \nGenerative AI in radiology \n \nModerator \nC. Blüthgen; Zurich/CH  \n \n \nPrecision, Non-classifications, and Misclassificati ons of General and \nMedical Large Language Models in Liver Lesions Clas sification using LI-\nRADS from Unstructured Radiology Reports \nJ. Lu¹, F. F-Y. Tang¹, J. Ng¹, C. Chan², H. M. Chen g¹, P. L. H. Yu¹,  \nW. K. W. Seto¹, *W. H. K. Chiu*¹; ¹Hong Kong/HK, ²H ampshire/UK \n(Keith.chiu@ha.org.hk) \n \nPurpose or Learning Objective: Large Language Models (LLM) are powerful \ntools for data extraction and summarization. Howeve r, scant evidence exists as \nto whether a medial-specific LLM is necessary to pe rform radiology tasks. This \nstudy evaluates the performance between a general a nd a medical LLM in \nextracting and categorizing liver lesions from radi ology reports according to the \nLI-RADS. \nMethods or Background: A total of 273 anonymized unstructured Computed \nTomography (CT) reports, written by 115 radiologist s from 5 institutions \ncontaining 599 liver observations were retrospectiv ely collected. These reports \nwere fed into GPT-4 and MedLM to assign LI-RADS cat egories for each \nobservation using zero-shot prompts (GPT4sp and Med LMsp) and instructions \npost-prompt engineering (GPT4pe and MedLMpe). Groun d truths and quality \nof the CT reports were derived by 2 board-certified  radiologists. \nResults or Findings: At lesion level, the accuracies for correctly class ifying \nmalignant lesions (LR-4/5/M) were 0.584, 0.634, 0.6 68, and 0.84 for GPT-4sp, \nMedLMsp, GPT-4pe, and MedLMpe respectively with Med LM outperforming \nGPT-4 using both simple prompts (p=0.023) and promp t engineering \n(p<0.001). At patient level, the accuracies were 0. 762, 0.744, 0.791, and 0.883, \nrespectively, with prompt engineering outperforming  simple prompts in MedLM \n(p < 0.001). Prompt engineering improved performanc e by reducing non-\nclassification in both MedLM (11.5% vs 33.7%, p<0.0 01) and GPT-4 (29.4% vs \n38.4% p<0.001). The quality of the CT reports of th e 31 misclassified/non-\nclassified patients on MedLMpe were considered aver age (median LIKERT \nscore 3/5) with a Fleiss’ ĸ value 0.563 (95%CI 0.356 - 0.770). \nConclusion: While general LLM exhibits potential in text-based medical tasks, \nour findings suggest that medical LLM yields superi or performance. \nLimitations: Limitations include a small sample size, lack of pr ompt \nengineering exploration, and only one of general an d medical LLM used. \nFunding for this study: None \nEthics committee - additional information: The ethics committee notification \ncan be found under Ref: KC/KE-23-0083/ER-3. \nAuthor Disclosures:  \nPhilip L. H. Yu: Nothing to disclose  \nHo Ming Cheng: Nothing to disclose  \nWan Hang Keith Chiu: Nothing to disclose  \nChelsea Chan: Nothing to disclose  \nJustin Ng: Nothing to disclose  \nWai Kay Walter Seto: Nothing to disclose  \nFanny Fong-Yi Tang: Nothing to disclose  \nJianliang Lu: Nothing to disclose \n \n \nEvaluating the Performance of LLaMA 3.1 in Classify ing Mammography \nReports Based on BIRADS Scores \n*A. Kumar*, V. K. Venugopal; New Delhi/IN \n(mitugarg75@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the \nperformance of the LLaMA 3.1 large language model ( LLM) in classifying \nmammography reports based on the Breast Imaging-Rep orting and Data \nSystem (BIRADS) classification without fine-tuning the model. \nMethods or Background: A total of 930 mammography reports, covering a \nrange of BIRADS classifications from 0 to 6, were p rocessed using the LLaMA \n3.1 open-source LLM (8B version). The model was pro mpted using a five-shot \nprompting technique. The classification accuracy of  the algorithm was \nanalyzed, and errors in classification were recorde d. Among the 930 reports, 8 \ninstances of errors were identified, with 4 cases w here BIRADS 2 was \nincorrectly classified as BIRADS 4 by the model. \nResults or Findings: The LLaMA 3.1 model demonstrated a classification \naccuracy of 921 correct classifications out of 930 reports, yielding an overall \naccuracy rate of 98.99%. Despite the model's strong  performance, errors were \npresent, particularly in the misclassification of l ower BI-RADS scores, with \nsome benign reports (BIRADS 2) being classified at a higher risk level \n(BIRADS 4) \nConclusion: LLaMA 3.1, even without fine-tuning, shows signific ant potential \nfor accurately classifying mammography reports base d on BIRADS scoring. \nThis indicates that large language models could ser ve as valuable tools in \nmedical imaging analysis, offering high accuracy wi th minimal adjustments. \nLimitations: The study is limited by the occurrence of misclassi fication in a \nsmall number of cases, particularly in distinguishi ng between benign and \nhigher-risk categories. Further studies with larger  datasets and fine-tuning may \nbe needed to improve reliability. \nFunding for this study: The study didn't receive any funding \nEthics committee - additional information: Anonymized data was used \nAuthor Disclosures:  \nAmit Kumar: Nothing to disclose \nVasantha Kumar Venugopal: Nothing to disclose \n \n \nEvaluating local open-source large language models for data extraction \nfrom unstructured reports on mechanical thrombectom y in patients with \nischemic stroke \n*A. Meddeb*¹, A. Othman², N. F. Grauhan², M. Scheel ³, J. Nawabi³; ¹Reims/FR, \n²Mainz/DE, ³Berlin/DE \n(aymenmeddeb@gmail.com) \n \nPurpose or Learning Objective: To assess the effectiveness of open-source \nLarge Language Models (LLMs) in extracting clinical  data from unstructured \nmechanical thrombectomy reports in patients with is chemic stroke caused by a \nvessel occlusion. \nMethods or Background: We deployed local open-source LLMs to extract \ndata points from free-text procedural reports in pa tients who underwent \nmechanical thrombectomy between September 2020 and June 2023 in our \ninstitution. The external dataset was obtained from  a second university hospital \nand comprised consecutive cases treated between Sep tember 2023 and \nMarch 2024. Ground truth labeling was facilitated b y a human-in-the-loop \n(HITL) approach, with time metrics recorded for bot h automated and manual \ndata extractions. We tested three models—Mixtral, Q wen, and BioMistral—\nassessing their performance on precision, recall, a nd F1 score across 15 \nclinical categories such as National Institute of H ealth Stroke Scale (NIHSS) \nscores, occluded vessels, and medication details. \nResults or Findings: The study included 1000 consecutive reports from ou r \nprimary institution and 50 reports from a secondary  institution. Mixtral showed \nthe highest precision, achieving 0.99 for first ser ies time extraction and 0.69 for \n\n \n \nFriday \nAbstract-based Programme \n \n 170  \noccluded vessel identification within the internal dataset. In the external \ndataset, precision ranged from 1.00 for NIHSS score s to 0.70 for occluded \nvessels. The HITL approach yielded an average time savings of 65.6% per \ncase, with variations from 45.95% to 79.56%. \nConclusion: LLMs showed high performance in automated clinical data \nextraction from medical reports. Incorporating HITL  annotations enhances \nprecision and also ensures the reliability of the e xtracted data. This \nmethodology presents a scalable privacy-preserving option that can \nsignificantly support clinical documentation and re search endeavors. \nLimitations: Variability in the quality and consistency of the i nput data, such \nas differences in terminology, formatting, or detai l level in the reports, can \naffect the performance of the models \nFunding for this study: None \nEthics committee - additional information: This retrospective study was \napproved by the ethics committee of the Charité Uni versity Hospital in Berlin \n(No. EA4/062/20).The requirement for informed conse nt was waived due to the \nretrospective design of the study. \nAuthor Disclosures:  \nNils F. Grauhan: Nothing to disclose \nAymen Meddeb: Research/Grant Support: Berlin Instit ute of Health \nJawed Nawabi: Nothing to disclose \nMichael Scheel: Nothing to disclose \nAhmed Othman: Nothing to disclose \n \n \nLarge language models in healthcare: DRAGON perform ance benchmark \nfor clinical NLP \n*J. S. Bosma*¹, K. Dercksen¹, M. De Rooij¹, F. Ciom pi¹, A. Hering¹,  \nJ. Geerdink², H. E. Huisman¹; ¹Nijmegen/NL, ²Almelo /NL \n(Joeran.Bosma@radboudumc.nl) \n \nPurpose or Learning Objective: Artificial Intelligence (AI) requires large-scale \nannotated datasets to train clinical algorithms to perform at an expert level. \nNatural Language Processing (NLP) shows great poten tial to annotate large \nvolumes of data from clinical routine and facilitat e the training of these \nalgorithms. This study aims to introduce a benchmar k for clinical NLP \nalgorithms, including Large Language Models (LLMs),  to assess the ability of \nalgorithms to extract information from medical repo rts. \nMethods or Background: The DRAGON (Diagnostic Report Analysis: \nGeneral Optimization of NLP) challenge has three ob jectives. First, it provides \na unique and publicly available cloud-based benchma rk for clinical NLP that \nspans 28 clinically relevant tasks. 28,824 annotate d medical reports from five \nDutch care centers from multiple imaging modalities  (MRI, CT, X-ray, \nhistopathology) and conditions spanning the entire body (lungs, pancreas, \nprostate, skin, etc.) are used. The tasks are desig ned to facilitate automated \ndataset curation and include predicting diagnoses, extracting lesion sizes, \nidentifying protected health information, and more.  Second, we release \nfoundational LLMs pretrained using four million cli nical reports from a sixth \nDutch care center. Third, we investigate three pret raining strategies across five \narchitectures by evaluating LLMs using the DRAGON b enchmark. \nResults or Findings: Results showed the superiority of domain-specific \npretraining (benchmark score of 0.770, 95% CI 0.755 -0.785) and mixed-\ndomain pretraining (0.756, 95% CI 0.739-0.773), com pared to general-domain \npretraining (0.734, 95% CI 0.717-0.752, p<0.005). T he best model achieved \nexcellent or good performance for 18/28 tasks and p oor or moderate \nperformance for 10/28 tasks. \nConclusion: The DRAGON benchmark showed that NLP is ready to fa cilitate \ndata curation in some settings, enabling high-quali ty, low-cost, and large-scale \nannotation, and uncovered where innovations are nee ded to improve clinical \nNLP. \nLimitations: Half of the tasks were sourced from a single academ ic tertiary \ncare center (14/28, 50%). \nFunding for this study: Funding was provided by Health~Holland \n(LSHM20103), European Union HORIZON-HLTH-2022: COMF ORT \n(101079894), European Union HORIZON-2020: ProCAncer -I project (952159), \nEuropean Union HORIZON-2020: PANCAIM project (10101 6851), and NWO-\nVIDI grant (number 18388). The collaboration projec t is co-funded by PPP \nAllowance awarded by Health~Holland, Top Sector Lif e Sciences \\& Health, to \nstimulate public-private partnerships. Views and op inions expressed are \nhowever those of the author(s) only and do not nece ssarily reflect those of the \nEuropean Union or European Health and Digital Execu tive Agency (HADEA). \nNeither the European Union nor the granting authori ty can be held responsible \nfor them. \nEthics committee - additional information: Retrospective use of anonymous \npatient data was approved by institutional or regio nal review boards at each \ncontributing center (identifiers: CMO 2016-3045; IR Bd22-159; A21-0349 2; \nA20-0777), and was conducted in accordance with the  principles of the \nDeclaration of Helsinki. Informed consent was waive d. \n \n \n \n \nAuthor Disclosures:  \nJoeran Sander Bosma: Nothing to disclose \nHenkjan En Huisman: Nothing to disclose \nFrancesco Ciompi: Nothing to disclose \nAlessa Hering: Nothing to disclose \nKoen Dercksen: Nothing to disclose \nMaarten De Rooij: Nothing to disclose \nJeroen Geerdink: Nothing to disclose \n \n \nImplementing Local Large Language Models and using Clinical Data \nWarehouse for Clinical Summarization and Decision S upport \n*M. Segeroth*, M. Bach, J. Wasserthal, J. Cyriac, M . Pradella, H-C. Breit,  \nB. Stieltjes, E. M. Merkle, S. Yang; Basel/CH \n(martin.segeroth@gmail.com) \n \nPurpose or Learning Objective: Recent advances in Large Language Models \n(LLMs) have improved medical text summarization and  decision support but \nraised data privacy concerns. We aim to integrate l ocal LLMs into clinical \nworkflows for testing with real-world patient data.  \nMethods or Background: Within our institutional healthcare network, a clin ical \ndata warehouse (CDWH) serves as a central hub for q uerying all patient \nrecords and parameters, while ensuring data privacy . Exemplary parameters \nlike temporal evolution of chemotherapies, dates an d outcomes of resections, \nfindings from previous imaging examinations, etc. w ere extracted for oncology \npatients. The collected data were fed via a prompt into local LLMs. We utilized \nprivateGPT and Ollama as the primary platform, allo wing integration of clinical \ntreatment guidelines. Regarding LLMs we tested Llam a3-70B and the German-\nlanguage SauerkrautLM Mixtral 8X7B Instruct which b oth ran on a Nvidia A100 \nGPU with 80 GB memory. A set of anonymized data was  processed with cloud-\nbased ChatGPT-4 and Claude-3 for comparison. \nResults or Findings: Using the privateGPT platform both tested LLMs ran on \na single GPU with maximally 65 GB of memory usage. Both LLMs created text \nsummaries within 15 seconds and provided decision s upport in under 5 \nseconds per request. For all brain cancer cases the  local LLMs provided a \ncorrect and reasonable summary of medical history. In decision-making for a \nprostate tumor board, the decision accuracy amounte d to 7 out of 10 test \ncases. For anonymized data, accuracy between the lo cal LLMs and both \nChatGPT-4 and Claude-3 was 8 out of 10 test cases. \nConclusion: Integration of local LLMs into clinical workflow or  research task is \npossible. Local LLMs were able to summarize medical  history or clinical data \nfor tumor boards, preserving local data privacy pol icies. \nLimitations: Only two local LLMs were evaluated on sophisticated  datasets. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nBram Stieltjes: Nothing to disclose \nJacob Wasserthal: Nothing to disclose \nMichael Bach: Nothing to disclose \nHanns-Christian Breit: Nothing to disclose \nMaurice Pradella: Nothing to disclose \nJoshy Cyriac: Nothing to disclose \nMartin Segeroth: Nothing to disclose \nElmar M. Merkle: Nothing to disclose \nShan Yang: Nothing to disclose \n \n \nTraining and Evaluation of Sentence Transformer Mod el for Retrieval \nAugmented Generation on Radiology Reports \n*K. Arzideh*, H. Schäfer, A. Idrissi-Yaghir, C. S. Schmidt, J. Haubold,  \nR. Hosch, F. Nensa; Essen/DE \n(kamyar.arzideh@uk-essen.de) \n \nPurpose or Learning Objective: In many medical settings physicians often \nhave to sift through unstructured documents to find  important information. This \nmanual process is time-consuming and can lead to mi ssed details. Retrieval \nAugmented Generation (RAG) can help physicians to q uickly locate relevant \ninformation. By using Sentence Transformer models f ine-tuned for retrieval \ntasks, similarity search between input query and do cument passages can be \nperformed to find relevant context. However, most p ublicly available models \nare not specifically fine-tuned for the radiology d omain and are therefore very \nlimited in finding clinically relevant information.  \nMethods or Background: Document chunks from 400,000 German clinical \nnotes including radiology reports and doctoral note s, were provided as input to \nthe SauerkrautLM-SOLAR-Instruct Large Language Mode l. The model was \nprompted to generate clinically related questions a nd answers based on these \nchunks. The model generated 11 million clinically r elated question-answer \npairs to fine-tune a multilingual-e5-large model. F or evaluation, 1,717 question-\nanswer pairs were generated from 215 radiology repo rts. A radiologist filtered \nout unrelated or incorrect pairs for a realistic ev aluation. The fine-tuned model \nwas then integrated into a RAG system, and its answ ers were compared to \nthose from a non-fine-tuned model using the same da taset. \n\n \n \nFriday \nAbstract-based Programme \n \n 171  \nResults or Findings: Fine-tuning the model resulted in improved performa nce \nmetrics. The BLEURT score increased from 0.551 to 0 .563, indicating \nenhanced alignment with human judgment. Similarly, the BERTScore F1 rose \nfrom 0.750 to 0.756. \nConclusion: By using LLM to generate synthetic questions out of  real world \ndocuments and fine-tuning sentence transformer mode ls on these question \nand document pairs, information retrieval performan ce can improve as \nindicated by automated evaluation metrics. \nLimitations: The evaluation was only carried out for documents i n German. \nFine-tuning on documents written in other languages  and from other hospital \nsites could lead to a broader applicability. \nFunding for this study: None \nEthics committee - additional information: This study was approved by the \nEthics Committee of the Medical Faculty of the Univ ersity of Duisburg-Essen \n(approval number 23-11557-BO). Due to the study's r etrospective nature, the \nrequirement of written informed consent was waived by the Ethics Committee \nof the Medical Faculty of the University of Duisbur g-Essen. All methods were \ncarried out in accordance with relevant guidelines and regulations. \nAuthor Disclosures:  \nJohannes Haubold: Nothing to disclose \nHenning Schäfer: Nothing to disclose \nCynthia Sabrina Schmidt: Nothing to disclose \nKamyar Arzideh: Nothing to disclose \nRené Hosch: Nothing to disclose \nAhmad Idrissi-Yaghir: Nothing to disclose \nFelix Nensa: Nothing to disclose \n \n \nLarge Language Models for Simplified Interventional  Radiology Reports: \nA Comparative Analysis \n*E. Can*¹, W. Uller¹, K. Vogt¹, F. Busch², N. Bayer l³, A. Kader²,  \nM. R. Makowski², K. K. Bressem², L. C. Adams²; ¹Fre iburg/DE, ²Munich/DE, \n³Erlangen/DE \n(elif.can@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: To quantitatively and qualitatively evaluate \nand compare the performance of leading large langua ge models (LLMs), \nincluding proprietary models (GPT-4, GPT-3.5 Turbo,  Claude-3-Opus, and \nGemini Ultra) and open-source models (Mistral-7b an d Mistral-8x7b), in \nsimplifying 109 interventional radiology reports. \nMethods or Background: Qualitative performance was assessed using a five-\npoint Likert scale for accuracy, completeness, clar ity, clinical relevance, \nnaturalness, error rates, including trust-breaking and post-therapy misconduct \nerrors. Quantitative readability was assessed using  Flesch Reading Ease \n(FRE), Flesch-Kincaid Grade Level (FKGL), SMOG Inde x, and Dale-Chall \nReadability Score (DCRS). Paired t-tests and Bonfer roni-corrected p-values \nwere used for analysis. \nResults or Findings: Qualitative evaluation showed no significant differ ences \nbetween GPT-4 and Claude-3-Opus for any metrics (al l Bonferroni-corrected p-\nvalues: p=1), while they outperformed other models across five qualitative \nmetrics (p < 0.001). GPT-4 had the fewest content a nd trust-breaking errors, \nwith Claude-3-Opus second. All models exhibited som e trust-breaking and \npost-therapy misconduct errors, with GPT-4-Turbo an d GPT-3.5-Turbo with \nfew-shot prompting showing the lowest error rates, and Mistral-7B and Mistral-\n8x7B the highest. Quantitatively, GPT-4 surpassed C laude-3-Opus in \nreadability metrics (all p < 0.001), with a median FRE score of 69.01 (IQR: \n64.88-73.14) versus 59.74 (IQR: 55.47-64.01) for Cl aude-3-Opus. GPT-4 also \noutperformed GPT-3.5-Turbo and Gemini Ultra (both p  < 0.001). Inter-rater \nreliability was strong (κ = 0.77-0.84). \nConclusion: GPT-4 and Claude-3-Opus demonstrated superior perfo rmance \nin generating simplified IR reports, but the presen ce of errors across all \nmodels, including trust-breaking errors, highlights  the need for further \nrefinement and validation before clinical implement ation. \nLimitations: This study was based on predefined metrics, which, while \ncomprehensive, may not capture all aspects of patie nt understanding and \nengagement. Future research should include real-wor ld data, a broader range \nof medical documents, and consider patient feedback  to more accurately \nassess the clinical utility of these models. \nFunding for this study: This study did not receive any specific funding fro m \npublic, commercial, or not-for-profit sectors. \nEthics committee - additional information: Since the reports did not include \nany real patient data, institutional review board a pproval was not required. This \nensures that the study adhered to ethical standards  by avoiding the use of real \npatient information and thereby eliminating the nee d for formal ethical approval \nprocesses typically required for studies involving human subjects. \n \n \n \n \n \n \n \nAuthor Disclosures:  \nKeno K. Bressem: Nothing to disclose \nElif Can: Nothing to disclose \nMarcus R. Makowski: Nothing to disclose  \nLisa C. Adams: Nothing to disclose \nKatharina Vogt: Nothing to disclose \nNadine Bayerl: Nothing to disclose \nAvan Kader: Nothing to disclose \nFelix Busch: Nothing to disclose \nWibke Uller: Nothing to disclose \n \n \nAutomated Radiology Controlling - Using Large Langu age Models for \nPrediction of Radiological Services based on Radiol ogical Reports \n*K. Arzideh*, A. Idrissi-Yaghir, H. Schäfer, K. A. Borys, J. Haubold, F. Nensa, \nR. Hosch; Essen/DE \n(kamyar.arzideh@uk-essen.de) \n \nPurpose or Learning Objective: In hospitals worldwide, controlling of \nradiological services is a manual process. In Germa ny, the so-called \n“Gebührenordnung für Ärzte” (GOÄ) regulates the bil ling of private medical or \ndental services, i.e. services outside the public h ealth insurance scheme. GOÄ \nnumbers can be used to indicate which private relat ed clinical interventions \nwere performed during treatment. These numbers are documented by going \nthrough radiological reports and picking out releva nt information, which is time-\nconsuming and error-prone. \nMethods or Background: In order to automate billing of radiological servic es, \na Large Language Model (LLM) was fine-tuned to gene rate GOÄ digits out of \nradiology reports. In total, 1,000,000 radiology re ports and GOÄ digit pairs \nwere split into 80 % training and 20 % test dataset . Training was performed on \na Phi-3-small-8k-instruct model. For evaluation, th e test dataset was compared \nagainst the numbers generated by the model. \nResults or Findings: The fine-tuned LLM achieved an accuracy of 75 % \ncalculated for the generation of GOÄ numbers. These  generated GOÄ codes \nwere identical to the ground truth. 83 % of the pre dicted codes were present in \nthe ground truth, but may not have been a complete match. \nConclusion: LLM are capable of automatically extracting relevan t controlling \ncodings based on radiology reports only. Therefore,  LLMs could be used as an \nenhanced method for the automation of controlling t asks in radiology. \nLimitations: The LLM needs human feedback and manual correction in order \nto achieve human-like results. The radiology report s used in this study were \nalso written in German language. The use of dataset s in other languages and \nfrom other hospitals could enable broader generaliz ability. \nFunding for this study: None \nEthics committee - additional information: This study adhered to all \nguidelines defined by the approving institutional r eview board of the \ninvestigating hospital. The Institutional Review Bo ard waived written informed \nconsent due to the study's retrospective nature. Co mplete anonymization of all \ndata was performed before inclusion in the study. \nAuthor Disclosures:  \nKatarzyna Anna Borys: Nothing to disclose \nJohannes Haubold: Nothing to disclose \nHenning Schäfer: Nothing to disclose \nKamyar Arzideh: Nothing to disclose \nRené Hosch: Nothing to disclose \nAhmad Idrissi-Yaghir: Nothing to disclose \nFelix Nensa: Nothing to disclose \n \n \nInsights and Challenges in Implementing Vision Tran sformers for Thorax \nRadiography \n*S. Hyska*, A. Wollek, T. Lasser, M. Ingrisch, B. O . T. Sabel; Munich/DE \n \nPurpose or Learning Objective: This study aimed to evaluate the \nperformance of a Vision Transformer (ViT)-based AI model, trained on publicly \navailable chest radiography datasets, when applied to real-world data from our \nclinic. The model's performance in detecting pleura l effusion, pneumothorax, \ncardiomegaly, and consolidation was examined, along  with potential \nconfounders. \nMethods or Background: The AI model, pre-trained on ImageNet and fine-\ntuned on >700,000 public chest X-rays (CXR), was te sted on an internal \ndataset of 113 CXR, including 23 pneumothorax, 29 c ardiomegaly, 31 \nconsolidation, 52 pleural effusion cases, and 29 no rmal CXR. The model’s \nperformance was assessed through ROC-curves, AUC, Y ouden Coefficient, \nand sensitivity/specificity metrics. Logistic regre ssion, odds ratios, and Fisher's \ntest were used to analyse confounding factors. \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 172  \nResults or Findings: The model correctly identified all normal CXRs. For  \npleural effusion, sensitivity was 96.2% and specifi city 98.4%, indicating strong \nperformance. For pneumothorax, sensitivity was only  26.1% with 96.7% \nspecificity. Pneumothorax size and presence of thor acic tubes were significant \nconfounders. Cardiomegaly was detected with 55.2% s ensitivity and 96.4% \nspecificity, whereas concomitant pleural effusions,  obscuring the heart \ncontours, act as a potential confounder. Consolidat ion was detected with \n45.2% sensitivity and 91.5% specificity, and higher  density consolidations were \nmore easily identified. \nConclusion: This study emphasizes the challenges AI models face  when \nintegrated into clinical practice, demonstrating th e importance of carefully and \nclinically assessing model performance on real-worl d-data, especially in the \ncontext of confounding factors. While our ViT model  showed strong \nperformance for pleural effusion and normal finding s, its detection of \npneumothorax, cardiomegaly, and consolidation was l imited. Known \nconfounders, e. g. pneumothorax size and presence o f thoracic tubes, were \nconfirmed, and new ones, such as pleural effusion i n cardiomegaly and density \nof consolidations, were identified. \nLimitations: The exploratory nature and limited number of CXR we re key \nlimitations. \nFunding for this study: This work was funded in part by the German federal \nministry of health’s program for digital innovation s for the improvement of \npatient- centered care in healthcare [grant agreeme nt no. 2520DAT920]. \nEthics committee - additional information: Approval by an ethics committee \nis present. \nAuthor Disclosures:  \nBastian Oliver Theodor Sabel: Nothing to disclose \nSardi Hyska: Nothing to disclose \nAlessandro Wollek: Nothing to disclose \nMichael Ingrisch: Nothing to disclose \nTobias Lasser: Nothing to disclose \n \n \n14:00-15:30 Research Stage 3 \nResearch Presentation Session: Paediatric \nRPS 1512 \nBody imaging in children: from head to toe \n \nModerator \nL. B. Laborie; Bergen/NO  \n(lene.bjerke.laborie@helse-bergen.no) \nAuthor Disclosures:  \nLene Bjerke Laborie: Other: Involved in development  of AI-algorithms for \npaediatric hip radiographs with software company Vi siana (DK). No financial \nbenefits. \n \n \nAutomatic Identification and Classification of Pedi atric \nGlomerulonephritis on Ultrasound Images Based on De ep Learning and \nRadiomics \n*J. Kou*, Y. Tang; Chongqing/CN \n(koujun@my.swjtu.edu.cn) \n \nPurpose or Learning Objective: Glomerulonephritis (GN) includes a diverse \nrange of kidney diseases that often exhibit subclin ical manifestations in \nchildren. While renal biopsy is the gold standard, its invasiveness, susceptibility \nto sampling errors, and time requirements impede ra pid diagnosis. This study \naimed to create a noninvasive diagnostic model for childhood GN by \nintegrating deep learning and radiomics techniques using renal ultrasound \nimages. \nMethods or Background: A total of 469 renal ultrasound images were \nselected from children undergoing ultrasound-guided  biopsy and split into \ntraining and validation sets at an 8:2 ratio to tra in a U-Net model for kidney \nsegmentation. Radiomic features were extracted from  the segmented regions \nand categorized by GN types: IgA nephropathy (127 c ases), minimal change \ndisease (83 cases), and Henoch-Schönlein purpura ne phritis (103 cases). \nThese categories were also split into training and validation sets at an 8:2 ratio. \nANOVA was used for feature selection in the trainin g set, followed by LASSO \nregression for dimensionality reduction, yielding 3 7 features. A random forest \nalgorithm was then used to develop a GN classificat ion model, which was \nevaluated using the validation set. \nResults or Findings: The segmentation model demonstrated excellent \nperformance, achieving 95.19% accuracy on the valid ation set. Thirty-seven \nselected features were used to build a strong class ification model, which \nshowed high accuracy and predictive power across GN  categories, with AUC \nvalues between 0.91 and 0.98. \nConclusion: The combination of deep learning and radiomics usin g renal \nultrasound images shows great potential for classif ying childhood GN \nsubtypes, offering a noninvasive method to enhance diagnostic efficiency and \npatient outcomes. \nLimitations: Firstly, the relatively limited data sources may ha ve introduced \nsome regional bias to our findings. Furthermore, we  focused only on the \npathological subtypes of three GN, which inevitably  limited the scope of the \nmodel. \nFunding for this study: 0 \nEthics committee - additional information: the Ethics Committee of \nChildren's Hospital at Chongqing Medical University . \nAuthor Disclosures:  \nJun Kou: Nothing to disclose \nYi Tang: Nothing to disclose \n \n \nBiomarkers of Primary Sclerosing Cholangitis detect ed with delayed \ngadolinium-enhanced Magnetic Resonance Imaging in p ediatric patients \n*F. Maccioni*, V. Cardinale, E. Damato, L. Busato, S. Veraldi, A. Valenti,  \nC. Catalano; Rome/IT \n(francesca.maccioni@uniroma1.it) \n \nPurpose or Learning Objective: Primary sclerosing cholangitis (PSC) is a \nsevere liver disease frequently associated with inf lammatory bowel disease \n(IBD) with a late diagnosis, mostly based on biliar y changes at MRCP. To \nidentify specific biomarkers for PSC using delayed phases of gadolinium \nenhancement to detect ductal fibrosis, such as inte stinal fibrosis in Crohn's \ndisease. \nMethods or Background: A prospective study based on the association of \nMRCP and gadolinium-enhanced MRI, including delayed  (7 minutes) phases, \nwas performed in 3 groups of pediatric patients, on e with PSC and IBD, one \nwith IBD only, and one of controls. Three radiologi sts blindly and independently \nanalyzed: a) intra and extrahepatic bile ducts stri ctures (IHBDs, EHBDs) at \nMRCP; b) gallbladder volume; c) gallbladder wall ga dolinium-enhancement; d) \nIHBDs-EHBDs gadolinium-enhancement. \nResults or Findings: We included 39 patients, 12 with PSC and IBD (31%) , \n16 with IBD only (41%) and 11 controls (28%). At MR CP, IHBDs strictures \nwere detected in 82% PSC-IBD patients (p<0.001). De layed enhancement of \ngallbladder wall was observed in 100% of PSC-IBD pa tients (100% sensitivity, \n90% specificity, (p<0.001); delayed enhancement of the extrahepatic biliary \nducts in 89% (89% sensitivity, 100% specificity, (p <0.001); and delayed \nenhancement of the intrahepatic biliary duct in 55, 6% (56% sensitivity, 100% \nspecificity). \nConclusion: Delayed gadolinium-enhancement of the gallbladder w all and \nextrahepatic bile duct showed remarkable sensitivit y and specificity for PSC. \nThese biomarkers may potentially increase MRI diagn ostic accuracy in high-\nrisk IBD patients. \nLimitations: The main limitation is the small number of patients . \nFunding for this study: No funding \nEthics committee - additional information: The study was approved by the \nethics commitee of our hospital. \nAuthor Disclosures:  \nVincenzo Cardinale: Nothing to disclose \nAlessandra Valenti: Nothing to disclose \nSilvio Veraldi: Nothing to disclose \nElio Damato: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLudovica Busato: Nothing to disclose \nFrancesca Maccioni: Nothing to disclose \n \n \nA nomogram model based on Combi-Elastography for pr eoperative \ndifferential diagnosis of biliary atresia \n*J. Chen*, F. Xu, Y. Gao, M. Yu, Y. Tang; Chongqing /CN \n(cjy419103@163.com) \n \nPurpose or Learning Objective: This study constructs a nomogram prediction \nmodel based on combi-elastography indexes and labor atory indicators for the \ndifferential diagnosis of biliary atresia (BA) and other cholestatic liver diseases \n(non-BA), with a view to recognizing BA at an early  stage and carrying out \ntreatment in time. \nMethods or Background: A total of 111 children aged < 180 d with cholestat ic \nhepatitis are included in the study, 75 in the BA g roup and 36 in the non-BA \ngroup. Conventional ultrasound, combi-elastography,  and laboratory tests are \nperformed on each patient before pathologic biopsy.  The variables are \nselected through logistics regression to construct a nomogram model, and the \neffectiveness of the model is evaluated. \nResults or Findings: Multifactorial logistic regression analysis shows t hat \ngamma-glutamyl transferase (GGT) , total bilirubin (TBIL) , and liver fibrosis-\nrelated F index (FI) of the combi-elastography inde x could be used as \nindependent predictors to differentiate BA from oth er causes of cholestasis. A \nnomogram model of these three indexes is constructe d which shows better \n\n \n \nFriday \nAbstract-based Programme \n \n 173  \nperformance, with an area under the operating chara cteristic curve (AUC) of \n0.887 (p < 0.001), sensitivity of 83.3%, and specif icity of 81.3%. The internal \nvalidation of the model is performed using 1,000 bo otstrap resamples and \nDecision curve analysis indicates that this model h ad a better diagnostic \nefficacy and accuracy. \nConclusion: The nomogram model based on combi-elastography inde xes and \nlaboratory indicators has certain value in differen tiating BA from other \ncholestatic liver diseases. \nLimitations: First, our sample size was small and it was a singl e-center \nstudy.Futher, we can expand the sample size to veri fy the validity of the \nnomogram model and provide more valuable insights f or clinicians. Second, \nadditional indicators can be included in the future , such as MMP-7 to improve \nthe diagnostic efficacy of the nomogram model. \nFunding for this study: Chongging Municipal Science and Health Joint \nMedical Research Project(2024MSXM050) \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board of the Children's Hospit al of Chongqing Medical \nUniversity (ethical approval number 2024-216), and all examinations and \nsurveys were conducted after obtaining consent from  the parents. \nAuthor Disclosures:  \nFenglin Xu: Nothing to disclose \nYi Tang: Nothing to disclose \nJingyu Chen: Nothing to disclose \nYang Gao: Nothing to disclose \nMingzhu Yu: Nothing to disclose \n \n \nUltrasound spleen stiffness as a marker of portal v ein anastomotic \nstenosis following paediatric liver transplantation : correlation with \ntranshepatic portal venography \n*D. C. Missud*, S. Le Cam, I. Mannes, M. Duché, S. Franchi-Abella;  \nLe Kremlin-Bicêtre/FR \n(david.missud@aphp.fr) \n \nPurpose or Learning Objective: Spleen stiffness measurement (SSM) is \nknown as a biomarker of significant portal hyperten sion in chronic liver \ndisease, but has been poorly studied in the context  of liver transplantation (LT). \nSSM may be particularly interesting to evaluate por tal vein anastomotic \nstenosis, a common and serious complication of LT. \nMethods or Background: Retrospective study including all portal vein \nstenoses assessed with transhepatic portal venograp hy (TPV) among \npaediatric liver recipients between 2015 and 2024. Diagnostic performance of \nSSM for the diagnosis of portal vein stenosis prior  to TPV was evaluated. \nResults or Findings: 36 children who underwent 58 TPV (stenosis group) \nwere included, with a median age at TPV of 3.0 year s. Findings were \ncompared to those of 58 randomly selected paediatri c liver recipients without \nportal vein complication (control group). In the st enosis group, there were 17 \nmild, 20 moderate, and 21 severe stenoses. SSM was significantly increased \nin the moderate and severe groups vs no or mild ste nosis groups (p < 0.005). \nThe Area Under the Curve was 0.96 for significant v s no to mild stenosis. A \nSSM cut-off of 33.7 kPa led to a sensitivity of 0.9 76 and a specificity of 0.840 \nfor the diagnosis of significant stenosis. \nConclusion: SSM correlates very well with transhepatic portal v enography \nwhen there is a suspicion of significant portal vei n anastomotic stenosis \nfollowing LT in paediatric patients. Further studie s may demonstrate that SSM \nis a good biomarker of portal hypertension followin g LT in paediatric patients, \nregardless of the aetiology. \nLimitations: This is a retrospective study which doesn't have th e strength of a \nprospective clilnical trial. This is a series of on ly 58 TPVs, but it is nonetheless \none the largest series ever published regarding pae diatric liver recipients with \nportal vein complications. \nFunding for this study: No external funding \nEthics committee - additional information: This is a usual care clinical \nretrospective study, under review by the local ethi cs committee. \nAuthor Disclosures:  \nInès Mannes: Nothing to disclose \nDavid Charles Missud: Nothing to disclose \nMathieu Duché: Nothing to disclose \nSolène Le Cam: Nothing to disclose \nStéphanie Franchi-Abella: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \nEvaluation of the Validity of Image-Defined Risk Fa ctors (IDRFs) in \nAbdominal Neuroblastoma \n*Z. Can Beyoğlu*, N. G. Akyel, E. Arslantaş, T. Banaz, M. Söyleyici, E. Ayaz, \nS. Akpınar Tekgündüz; Istanbul/TR \n(zekicanbeyoglu@gmail.com) \n \nPurpose or Learning Objective: The presence of image-defined risk factors \n(IDRFs) in neuroblastoma plays a crucial role in cl inical decision-making, \nparticularly when choosing between primary tumor re section and neoadjuvant \nchemotherapy. This study aims to evaluate how the p resence of IDRFs \ninfluences the surgical outcomes of patients with a bdominal neuroblastoma, \nspecifically focusing on complete tumor resection a nd the likelihood of \nrecurrence within one year. \nMethods or Background: This retrospective study involved 60 patients \ndiagnosed with abdominal neuroblastoma. Of these, 3 0 patients had tumors \nwith IDRFs present, while the remaining 30 patients  had no IDRFs identified. \nThe recurrence rates and surgical outcomes were ana lyzed over a one-year \nfollow-up period. Factors such as tumor residuals a nd overall recurrence rates \nwere examined to assess the impact of IDRFs on surg ical success and \nprognosis. \nResults or Findings: A total of 60 patients were included in the analysi s. \nAmong the 30 patients with IDRFs, 9 patients showed  no signs of recurrence \nwithin one year, while 12 patients experienced tumo r recurrence, and 9 had \nresidual tumor tissue post-surgery. In the group of  30 patients without IDRFs, \n22 patients did not experience recurrence, 5 patien ts had recurrences, and 3 \nhad residual tumor tissue after surgery. These find ings suggest that the \npresence of IDRFs significantly impacts the likelih ood of recurrence and \nsurgical success. \nConclusion: The study found that the presence of IDRFs in abdom inal \nneuroblastoma patients is an important predictor of  recurrence. Patients with \nIDRFs had a lower non-recurrence rate (30%) compare d to those without \nIDRFs (73%). Therefore, IDRF presence can be a key factor in determining \nsurgical approach and prognosis. \nLimitations: The study is limited by the small sample size and v ariability in \nimaging techniques used during follow-up, which cou ld affect the consistency \nof the results. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was approved by the \nEthics Committee of Başakşehir Çam and Sakura City Hospital, \nIstanbul/Turkey \nAuthor Disclosures:  \nMerve Söyleyici: Nothing to disclose \nSibel Akpınar Tekgündüz: Nothing to disclose \nEsra Arslantaş: Nothing to disclose \nNazli Gülsüm Akyel: Nothing to disclose \nTuba Banaz: Nothing to disclose \nZeki Can Beyoğlu: Nothing to disclose \nErcan Ayaz: Nothing to disclose \n \n \nValidation of a New Scoring System for Residual Tum or Assessment \nAfter Surgery in Pediatric Neuroblastoma: Prelimina ry Results \n*J. F. Schäfer*¹, J. Spogis¹, J. Fuchs¹, B. Hero², T. Simon², A. Eggert³,  \nM. Müller⁴, S. Warmann³, B. Timmermann⁵; ¹Tübingen/DE, ²Cologne/DE, \n³Berlin/DE, ⁴Heidelberg/DE, ⁵Essen/DE \n(juergen.schaefer@med.uni-tuebingen.de) \n \nPurpose or Learning Objective: The SIOPEN HR-NBL2 protocol for the \ntreatment of high-risk neuroblastoma (NB) is a mult inational trial with radiation \nrandomization for patients with macroscopic residua l tumors after induction \nchemotherapy and tumor resection. However, precise definitions of residual \ntumor on cross-sectional imaging are not available yet. This study aims to \nvalidate a newly developed scoring system for asses sing residual tumors \nproposed by the German Neuroblastoma registry. \nMethods or Background: Patients treated according to the GPOH NB \nguidelines and irradiated at the West German Proton  Therapy Center Essen \nwere retrospectively included if pre-/post-operativ e and current MRI were \navailable at the time of radiotherapy. The score is  based on MRI (tumor size \nand diffusion restriction), mIBG uptake, and the su rgical report, assigning a \nlesion score from 1 to 3 for each point in time. An onymized imaging data were \nuploaded to a browser-based imaging platform (mRay,  Germany) for multi-\nreader, multi-institutional evaluation. \nResults or Findings: A total of 15 patients (mean age 5.1y; SD 2.1y) wit h 45 \nMRIs were assessed by two experienced readers (R1/R 2). The mean \npreoperative tumor volume was 100 ml (range 3–344 m l). Image-defined risk \nfactors were identified in all patients except one.  Based on the surgical reports, \ncomplete macroscopic resection was achieved in 9 pa tients, while imaging \nanalysis by R1 and R2 confirmed complete resection in 4 cases. On a lesion-\nbased analysis, surgery identified 6 residual lesio ns in 6 patients, while R1 and \nR2 identified 25 lesions in 11 patients (mean size:  13 mm, range: 4–32 mm). \nThe inter-reader agreement for MRI scoring was exce llent (ICC 0.87; 95% CI: \n072–0.94). \n\n \n \nFriday \nAbstract-based Programme \n \n 174  \nConclusion: Regarding MRI findings, the newly developed scoring  system for \nresidual tumors in pediatric neuroblastoma is feasi ble. Further validation \nthrough multi-institutional, platform-based reading s is planned. \nLimitations: Preliminary data \nFunding for this study: No Funding. \nEthics committee - additional information: University of Essen, Germany \nAuthor Disclosures:  \nAngelika Eggert: Nothing to disclose \nBeate Timmermann: Nothing to disclose \nThortsen Simon: Nothing to disclose \nJörg Fuchs: Nothing to disclose \nJürgen F Schäfer: Nothing to disclose \nBarbara Hero: Nothing to disclose \nSteven Warmann: Nothing to disclose \nMichael Müller: CEO: providing the research platfor m \nJakob Spogis: Nothing to disclose \n \n \nImaging predictors of rupture in pediatric solid tu mors \n*G. G. Koodaly*, V. Smriti, A. D. Baheti, S. Kulkar ni, N. Shetty, K. B. Gala,  \nM. Ramadwar, S. Quereshi, G. Chinnaswamy; Navi Mumb ai/IN \n(ggkoodaly@gmail.com) \n \nPurpose or Learning Objective: Tumor rupture poses as a life-threatening \ncomplication in pediatric solid tumors, and require s a high-risk protocol \nmanagement. This study aims to identify the potenti al imaging predictors for \ntumor rupture. \nMethods or Background: The clinical data of children with pediatric solid \ntumor rupture at our institution from January 2021 to June 2024 were reviewed \nretrospectively. \nResults or Findings: Total of 22 cases, which comprised of 14 \nhepatoblastoma, 6 Wilms’ tumor, 1 neuroblastoma, an d 1 Ewing sarcoma were \nanalyzed. Patients were aged between 2 to 10 years (median 4.5 years). 17 \npatients were treatment naïve and 5 on chemotherapy . Patients commonly \npresented with abdominal pain, distension, nausea, vomiting, and signs of \nshock due to significant drop in hemoglobin. Imagin g via ultrasound and CT \nscans revealed tumors, with a notable correlation b etween tumor rupture and \nfactors such as high PRETEXT scores, tumor sizes ov er 10 cm, a greater \npercentage of necrotic component, intratumoral blee d and hyperdense ascitic \nfluid (>25 Hounsfield units). Among the hepatoblast oma cases, 71% had \nPRETEXT III and above. At diagnosis, 64% of the pat ients had hemoglobin \nlevels ≤ 8 g/L, with 18% at ≤ 6 g/L. Seven patients required angioembolization, \n2 underwent surgery and rest were treated conservat ively. Unfortunately, one \npatient developed tumor lysis syndrome and 3 succum bed to tumor rupture. \nConclusion: This study identifies the imaging predictors of tum or rupture in \npediatric solid tumors and associated risk factors,  such as chemotherapy, \nlarger size of tumor and other high-risk factors. \nLimitations: Retrospective study. Its prevalence in cohort is no t known. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Retrospective study \nAuthor Disclosures:  \nGenesis Giddo Koodaly: Nothing to disclose \nSuyash Kulkarni: Nothing to disclose \nSajid Quereshi: Nothing to disclose \nMukta Ramadwar: Nothing to disclose \nNitin Shetty: Nothing to disclose \nKunal Bharat Gala: Nothing to disclose \nAkshay Dwarka Baheti: Nothing to disclose \nVasundhara Smriti: Nothing to disclose \nGirish Chinnaswamy: Nothing to disclose \n \n \nThe Risk of Pediatric and Adolescent Hematologic Ma lignancies \nAssociated with Medical Imaging (RIC) \n*R. Smith-Bindman*¹, S. Albers², M. Kwan³, W. Bolch ⁴, E. Bowles⁵, C. Stewart¹, \nR. Greenlee⁶, J. Pole⁷, D. L. Miglioretti²; ¹San Francisco, CA/US,  \n²Davis, CA/US, ³Oakland, CA/US, ⁴Gainesville, FL/US, ⁵Seattle, WA/US, \n⁶Marshfield, WI/US, ⁷Brisbane/AU \n(rebecca.smith-bindman@ucsf.edu) \n \nPurpose or Learning Objective: Risks of hematologic malignancies \nassociated with medical imaging ionizing radiation exposure have not been \nevaluated in the U.S. or Canada. \nMethods or Background: This retrospective cohort study followed 3,724,622 \nchildren born at one of 6 U.S. healthcare systems o r in Ontario, Canada from \n1/1/1996 to 4/30/2016 from birth until the earliest  of a cancer diagnosis, death, \nemigration from Ontario, 6 months after disenrollme nt from healthcare system, \nage 21, or 12/31/2017. Active bone marrow radiation  doses from medical \nimaging examinations were estimated. Hazards ratios  (HR) and relative risks \n(RR) of hematologic malignancies associated with cu mulative radiation \nexposure were estimated. \nResults or Findings: A total of 2,961 hematologic malignancies were \ndiagnosed during 35,735,719 person-years of follow up, including lymphoid \nmalignancies (n=2,349, 79.3% of malignancies), myel oid or acute leukemia \n(myeloid, n=460, 15.5%); and histiocytic and dendri tic cell malignancies (H&D, \nn=129, 5.1%). Malignancy risk increased with cumula tive dose (p<0.0001); \ne.g., risk was 1.70 times higher among children wit h a cumulative dose of 15 to \n< 20 mGy vs. <1mGy (95%CI=1.27-2.28). The risk of m alignancy was 3.7 \ntimes higher (95%CI=2.82-4.71) for children with a cumulative exposure of 100 \nmGy vs. no exposure and was significantly elevated for cancer subtypes. RRs \ndecreased with increasing time since exposure and i ncrease with age at \nexposure and attained age. We estimate 27 excess he matologic malignancies \nby age 21 per 10,000 children with a cumulative exp osure of 30 mGy or higher \nvs. <1 mGy, equivalent to average dose of approxima tely 2 head CTs. \nConclusion: Children and adolescents who undergo radiation-base d medical \nimaging are at a small, but significant increased r isk of hematologic \nmalignancy. \nLimitations: While reverse causation is a potential limitation, analyses of \nclinical indications confirmed symptoms related to hematologic malignancy \nwere rare in included studies. \nFunding for this study: US National Institutes of Health, National Cancer \nInstitute R01CA185687, R50CA211115 \nEthics committee - additional information: The requirement for individual \ninformed consent was waived for the study \nAuthor Disclosures:  \nErin Bowles: Nothing to disclose \nMarilyn Kwan: Nothing to disclose \nJason Pole: Nothing to disclose \nRebecca Smith-Bindman: Nothing to disclose \nCarly Stewart: Nothing to disclose \nDiana L Miglioretti: Nothing to disclose \nWesley Bolch: Nothing to disclose \nSusan Albers: Nothing to disclose \nRobert Greenlee: Nothing to disclose \n \n \nUltra-low dose CT for non-accidental injury \nA. A. Mohammed, M. F. Mcentee, A. England, N. Moore , E. K. Mahon,  \nM. Maher, *R. Young*; Cork/IE \n \nPurpose or Learning Objective: The aim of this study is to compare two \nwhole-body CT protocols for SPA and to assess wheth er the ULD protocol can \nprovide diagnostic images of sufficient quality com pared to the standard dose \n(STD). \nMethods or Background: In this cross-sectional study, two sets of images o f \na newborn whole-body anthropomorphic phantom were a cquired using \ndifferent protocols, one with STD and the other wit h ULD protocol. The \neffective dose (ED) of both protocols was calculate d using the Monte Carlo \ndose simulation approach. The image quality arising  from both protocols was \nthen assessed at the ECR 2024 Congress using a four -section questionnaire. \nThe questionnaire included demographic information,  a comparison of the \nvisualization of different bony anatomical structur es, and confidence in \ndiagnosis using either protocol. The Wilcoxon signe d-rank test was used to \nevaluate the significant differences between STD an d ULD image quality \nscores. VGC analyser was used for image quality rat ing and comparison. \nResults or Findings: 46 participants were included in this study. For al l body \nparts, STD showed significantly higher image qualit y than ULD \n(AUCVGC=0.75). 76% of the participants were confide nt to use the STD \nprotocol for SPA diagnosis, whereas, 41% were confi dent to use the ULD \nprotocol. The percentage effective dose difference between protocols was \n93.5%(STD=0.56 mSv vs. ULD=0.04mSv) and most of the  participants \nunderestimated the dose reduction. \nConclusion: This study successfully compared the STD and ULD wh ole-body \nCT in phantom and shows ULD CT is a promising techn ique which may \ncompete with digital radiography for SPA diagnosis.  \nLimitations: The limitations of the study are: 1-the use of phan tom involves no \npatient movement experienced, and no pathologies. 2 -sampling bias was due \nto the observers as only people attending the ECR 2 024 could participate. 3-\nonly compares two whole-body CT protocols. \nFunding for this study: Taif University \nEthics committee - additional information: The ethical approval provided by \nthe University College Cork \nAuthor Disclosures:  \nMark F. Mcentee: Nothing to disclose \nNiamh Moore: Nothing to disclose \nRena Young: Nothing to disclose \nAhmed Abdulahad Mohammed: Nothing to disclose \nAndrew England: Nothing to disclose \nEimear Kate Mahon: Nothing to disclose \nMichael Maher: Nothing to disclose \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 175  \nDeep Learning-based Detection of Pediatric Bone Tum ors Using X-ray \nImaging \nS. Consalvo, A. Curto Vilalta, *A. W. Marka*, S. Br eden, B. Schlossmacher,  \nC. Eisfeld, D. Rückert, R. Von Eisenhart-Rothe, F. Hinterwimmer; Munich/DE \n(alexander.marka@gmail.com) \n \nPurpose or Learning Objective: To address the challenge in musculoskeletal \nradiology of early detection of bone tumours in chi ldren with x-ray imaging. \nWhile machine learning (ML) has shown proficiency i n differentiating tumour \nentities, a critical gap remains in initial tumour detection, particularly for non-\noncology-trained professionals and general practiti oners. Paediatric tumours \nare often incidentally discovered, underscoring the  need for more sophisticated \ntools. \nMethods or Background: This retrospective study utilized X-ray data from a  \ndiverse cohort of paediatric patients from our loca l musculoskeletal tumour \ndatabase. The dataset comprised 817 images (567 pat hological and 250 \nhealthy) from 511 patients, including ten benign, i ntermediate and malignant \ntumour entities. We employed the ResNet18 architect ure for classification, \nsupported by cross-validation techniques and excess ive data augmentation \nstrategies. Our methodology focused on enhancing th e ML system's ability to \ngeneralize across various clinical scenarios. \nResults or Findings: The ML model demonstrated high performance with an \naccuracy of 96.39%, a sensitivity of 96.0%, and a s pecificity of 96.0% into \n“tumour” and “no tumour”. The variance in cross-val idation splits was 0.05, \n0.10, and 0.13, respectively, indicating stable res ults across different test sets. \nThese metrics reflect the model's reliability and p otential effectiveness in \nclinical settings. \nConclusion: Current ML applications in orthopaedic oncology are  progressing \nyet remain insufficiently performant for widespread  clinical use. However, our \nfindings underscore the potential of ML tools in ai ding both young professionals \nand general practitioners. Additionally, future adv ancements should focus on \nmultimodal approaches that incorporate not only X-r ay data but also MRI and, \ncrucially, clinical data. Integrating these diverse  data sources will enhance the \nperformance and applicability of ML in diagnosing a nd managing paediatric \nbone tumours, offering a more holistic and effectiv e approach to patient care. \nLimitations: 90% Monocentric Data. \nFunding for this study: Nemetschek Innovation Foundation and Bavarian \nMinistry of Science and the Arts \nEthics committee - additional information: Klinikum rechts der Isar, \nTechnical University of Munich. \nAuthor Disclosures:  \nCarolin Eisfeld: Nothing to disclose \nAnna Curto Vilalta: Nothing to disclose \nSebastian Breden: Nothing to disclose \nAlexander Wolfgang Marka: Nothing to disclose \nFlorian Hinterwimmer: Nothing to disclose \nSarah Consalvo: Nothing to disclose \nDaniel Rückert: Nothing to disclose \nRüdiger Von Eisenhart-Rothe: Nothing to disclose \nBenjamin Schlossmacher: Nothing to disclose \n \n \n14:00-15:30 Research Stage 4 \nResearch Presentation Session: Neuro \nRPS 1511 \nInsights into brain tumours: from visible to \ninvisible and back again \n \nModerator \nC. Eraslan; Izmir/TR  \n(eraslancenk@gmail.com) \n \n \nImproved Brain Tumor visualization with 3T Stack-of -Stars Echo \nUnbalanced T1 Relaxation-Enhanced Steady-State MRI – A Two Center \nClinical Study \n*A. Toth*¹, R. Edelman², J. A. Chetta¹, J. Joyce¹, M. V. Spampinato¹, R. Zi³,  \nK. T. Block³, A. Varga-Szemes¹; ¹Charleston, SC/US,  ²Evanston, IL/US,  \n³New York, NY/US \n(adrienntoth706@gmail.com) \n \nPurpose or Learning Objective: The novel stack-of-stars echo unbalanced \nT1 relaxation-enhanced steady-state (SOS echo-uT1RE SS) sequence aims to \nprovide improved motion robustness and enhanced dar k blood contrast, and to \nimprove the visualization of small metastases and l ow enhancing lesions. This \nstudy compared the image quality and diagnostic uti lity of SOS echo-uT1RESS \nwith the widely used magnetization-prepared rapid a cquisition gradient-echo \n(MPRAGE) sequence in brain tumor imaging. \nMethods or Background: This two-center prospective study involved 25 \nadults with known brain tumors (n= 5 intra-axial pr imary brain tumors; n= 11 \nintra-axial brain metastases; n= 9 extra-axial brai n tumors). Each participant \nunderwent 3T contrast enhanced MRI of the brain wit h both standard \nMPRAGE and prototype SOS echo-uT1RESS sequences. Co ntrast-to-noise \nratio (CNR) and tumor-to-brain contrast were quanti tatively analyzed. Image \nquality, lesion conspicuity, and image artifacts we re scored on a 4-point Likert \nscale. Diagnostic performance and assessment of the  vascular and dural \ninvolvement were compared side-by-side by 2 readers . \nResults or Findings: There was no significant difference in CNR between \nMPRAGE and SOS echo-uT1RESS (27.0 ± 19.2 vs. 26.5 ± 14.9, respectively; \np = 0.84). SOS echo-uT1RESS demonstrated a 1.6-fold  improvement in tumor-\nto-brain contrast compared with MPRAGE (0.7 ± 0.4 v s. 0.4 ± 0.3, respectively; \np < 0.001). Image quality and artifacts were simila r for both sequences, while \nSOS echo-uT1RESS showed improved lesion conspicuity , diagnostic \nperformance and enhanced detection of vascular and dural invasion. \nConclusion: SOS echo-uT1RESS showed promising results for post- contrast \nevaluation of brain tumors on 3T MRI. This techniqu e enhanced lesions \nvisibility, achieving approximately 1.6-fold improv ement in tumor-to-brain \ncontrast compared to MPRAGE. It offered superior di agnostic performance and \nimproved detection of vascular and dural involvemen t. \nLimitations: The limitations of the study are the relatively sma ll patient cohort \nand that quantitative measurements were performed b y a single observer. \nFunding for this study: Funding was provided by NIH HHS United States \n(1R01CA263091 and 1R21CA273280). \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board (Pro00128013)). \nAuthor Disclosures:  \nKai Tobias Block: Nothing to disclose \nRobert Edelman: Nothing to disclose \nRuoxun Zi: Nothing to disclose \nAdrienn Toth: Nothing to disclose \nJennifer Joyce: Nothing to disclose \nJustin A Chetta: Nothing to disclose \nAkos Varga-Szemes: Nothing to disclose \nM. Vittoria Spampinato: Nothing to disclose \n \n \nMultiparametric MRI‑based radiomics with interpretable machine learning  \nfor predicting progesterone receptor expression in meningioma:  \nA multicenter study \n*G. Lin*, W. Chen, J. Ji; Lishui/CN \n \nPurpose or Learning Objective: This study aimed to develop and validate an \ninterpretable machine learning-based prediction mod el for preoperatively \npredicting progesterone receptor (PR) expression in  meningioma patients \nusing multiparametric magnetic resonance imaging (M RI). \nMethods or Background: The study retrospectively enrolled 739 patients wit h \npathologically confirmed meningioma from three medi cal centers, dividing them \ninto four cohorts: training (n = 294), internal tes t (n = 126), external test 1 (n = \n217), and external test 2 (n = 102). Radiomics char acteristics were derived \nfrom T2-weighted and contrast-enhanced T1-weighted MRI images, followed \nby feature selection. A machine learning-based comb ined model was \ndeveloped by incorporating radiomics scores (rad-sc ores) from the optimal \nradiomics model along with clinical predictors. The  Shapley additive \nexplanation (SHAP) method was employed to visually represent the process of \nmaking predictions. The prognostic value of the mod el was evaluated using \nKaplan-Meier survival analysis. \nResults or Findings: Among the 739 patients, 299 (40.5%) had negative PR  \nexpression confirmed by pathology. Twelve radiomics  features derived from \nmultiparametric MRI were selected to build the radi omics model. Tumor \nlocation and enhancement pattern were identified as  key clinical predictors and \nwere combined with rad-scores to create a combined model utilizing the \nextreme gradient boosting (XGBoost) algorithm. The combined model \ndemonstrated strong accuracy and robustness, with a rea under the curve \nvalues of 0.907, 0.827, 0.846, and 0.807 across tra ining, internal test, external \ntest 1, and external test 2 cohorts, respectively. The survival analysis indicated \nthat the combined model was able to effectively cat egorize patients based on \nrecurrence outcomes. \nConclusion: The XGBoost combined model, utilizing multiparametr ic MRI, \nshows promise for predicting PR expression in menin gioma patients. The \nSHAP visualization enhances the model’s clinical ap plicability. \nLimitations: As a retrospective study, it is susceptible to info rmation selection \nbias. \nFunding for this study: This work was supported by the Key Project of Joint  \nConstruction by Provincial and Ministerial Authorit ies (Grant No.WKJ-ZJ-2452 \nto Minjiang Chen), Medical and Health General Proje ct of Zhejiang Province \n(Grant No. 2023KY425 to Guihan Lin, Grant No. 2024K Y562 to Shuiwei Xia), \n\n \n \nFriday \nAbstract-based Programme \n \n 176  \nand Medical and Health Youth Innovation Project of Zhejiang Province (Grant \nNo. 2023RC115 to Weiyue Chen). \nEthics committee - additional information: All procedures performed in \nstudies involving human participants were in accord ance with the ethical \nstandards of the institutional and/or national rese arch committee and with the \n1964 Helsinki declaration and its later amendments or comparable ethical \nstandards. This study was approved by the Instituti onal Review Board and \nHuman Ethics Committee of the Fifth Affiliated Hosp ital of Wenzhou Medical \nUniversity (2024-336), the Sixth Affiliated Hospita l of Wenzhou Medical \nUniversity, and the Third Affiliated Hospital of We nzhou Medical University, \nwith the requirement for patient informed consent b eing waived due to its \nretrospective nature. All patients’ information was  anonymized prior to the \nanalysis. \nAuthor Disclosures:  \nJiansong Ji: Nothing to disclose \nWeiyue Chen: Nothing to disclose \nGuihan Lin: Nothing to disclose \n \n \nHistogram Analysis in Predicting the Intracranial M eningioma Grading \nBased on Amide Proton Transfer-Weighted Imaging \n*Y. H. Lee*, B-H. Kim, M. Kim, S-D. Kim; Ansan-Si/K R \n(younghen@korea.ac.kr) \n \nPurpose or Learning Objective: To determine whether amide proton transfer-\nweighted (APTW) histogram analysis is useful for pr edicting the grade of \nmeningioma \nMethods or Background: We retrospectively enrolled a total of 48 patients \n(M:F=16:32, mean age: 60.0 years; grade 1:grade 2/3 =36:12) with \npathologically proven intracranial meningioma who u nderwent mDIXON 3D-\nAPT sequence of the fast spin echo method in additi on to conventional 3T MR \nprotocols prior to surgical resection. From the rep resentative APTW images of \nthe tumor registered with gadolinium-enhanced T1 im ages, the following \nparameters of each histogram were obtained: every 5  intervals from 5th to 95th \npercentile, mean, median, maximum, minimum, standar d deviation, kurtosis \nand skewness. The diagnostic performance of each AP TW histogram \nparameter for differentiating grade 2/3 from grade 1 intracranial meningioma \nwere evaluated by drawing the receiver operating ch aracteristic (ROC) curves \nand calculating the cut-off values \nResults or Findings: Among all histogram parameters, only maximum, \nstandard deviation and 75th,80th,85th,90th, and 95t h percentiles for APTW \nsignal in grade 2/3 were significantly higher than those of grade 1 (p<0.05). \nAccording to the ROC curve comparison analysis, the  areas under the curves \nof 75th,80th,90th,95th percentile, maximum and stan dard deviation to \ndiscriminate grade 2/3 from grade 1 were 0.706,0.71 3,0.722,0.734, 0.738, \n0.722, and 0.718, respectively. \nConclusion: Histogram analysis of APTW imaging can be used in c linical \npractice for grading of intracranial meningiomas \nLimitations: 1.ROI-dependency, 2.small number of participants \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nYoung Hen Lee: Nothing to disclose \nBaek-Hyun Kim: Nothing to disclose \nSang-Dae Kim: Nothing to disclose \nMyungji Kim: Nothing to disclose \n \n \nPrognostic utility of intratumoral susceptibility s ignals in adult diffuse \ngliomas: a radiopathological study \n*J. I. Tudela Martínez*, V. Vázquez Sáez; Murcia/ES  \n \nPurpose or Learning Objective: Intratumoral susceptibility signals (ITSS) are \npromising radiological markers for assessing diffus e gliomas. This study \nevaluates the relationship between ITSS grading and  key radiological and \nhistopathological prognostic factors in adult diffu se gliomas. \nMethods or Background: Between January 1st, 2022, and April 30th, 2024, \nwe selected 99 patients diagnosed with adult diffus e glioma who met the \nfollowing criteria: age over 18 years, MRI scans al lowing ITSS quantification \nand confirmed pathological diagnosis with available  molecular testing. \nRadiological variables included tumor volume, subve ntricular zone involvement \nand relative cerebral blood volume (rCBV) on MRI pe rfusion. Histopathological \nfeatures examined were WHO-2021 grade, Ki-67 index,  mitotic count, \nnecrosis, microvascular proliferation, and key prog nostic mutations (IDH, p53, \nATRX, and CDKN2A/B). Spearman’s correlation and chi -square tests were \nused for quantitative and qualitative variables, re spectively. Multiple logistic \nregression models were developed to predict WHO tum or grade, categorized \nas low (1-2) or high (3-4), based on ITSS grade, tu mor volume, and rCBV. \nResults or Findings: ITSS grades 0-1 were more common in \noligodendrogliomas and astrocytomas, while grades 2 -3 were linked to \nglioblastomas (p<0,001). ITSS grade positively corr elated with rCBV, tumor \nvolume, WHO grade, mitotic count, and Ki-67 index ( p<0,001). Higher ITSS \ngrades also showed increased necrosis and microvasc ular proliferation \n(p<0,001). IDH mutations and 1p/19q co-deletions we re more prevalent in \ngrades 0-1 (p<0,001 and p=0,001, respectively), whi le CDKN2A/B alterations \ncorrelated with grades 2-3 (p=0,02). Regression mod els showed AUCs of \n0,937 and 0,960 for ITSS combined with rCBV and tum or volume, respectively \n(p=0,000). \nConclusion: ITSS represent valuable biomarkers for assesing dif fuse gliomas, \noffering diagnostic and prognostic insights that ca n guide clinical decision-\nmaking. Additionally, combining ITSS with MRI-rCBV and tumor volume \nenhances predictive capacity of these radiological parameters. \nLimitations: ITSS grading remains semi-quantitative; further stu dies should \nfocus on fully quantifying ITSS data. \nFunding for this study: None \nEthics committee - additional information: We consulted with the ethics and \nresearch committee regarding the need for approval for the study. They \nconfirmed that, due to its observational nature, su ch approval is not required. \nAuthor Disclosures:  \nVictoria Vázquez Sáez: Nothing to disclose \nJose Ignacio Tudela Martínez: Nothing to disclose \n \n \nRelationship between Whole-tumor MRI-based Fractal Analysis and \nMolecular Features in IDH-wildtype Glioblastoma \n*B. Zhang*, J. Zhou; Lanzhou/CN \n \nPurpose or Learning Objective: Molecular mechanisms and specific genes \ninvolved in the growth of Glioblastoma (GBM) are im portant factor in deciding \nthe treatment strategy. In this study, we aimed to non-invasively explore the \nrelationship between whole-tumor MRI-based fractal features and molecular \nfeatures of GBM. \nMethods or Background: The clinical and imaging data of 104 patients with \nIDH-wildtype GBM at our hospital between November 2 018 and June 2024 \nwere retrospectively analyzed. The molecular featur es of GBM were collected \nby molecular sequencing and immunohistochemical met hod, including MGMT \npromoter methylation, 1p/19q-codeleted, TERT promot er mutation, Ki67, and \nP53 status. The volume of interest of whole tumor w as manually segmented \nslice-by-slice using ITK-SNAP software. Fractal fea tures of whole-tumor in \ncontrast-enhanced T1-weighted imaging were extracte d using Image J \nsoftware. As many as 24 fractal features (fractal d imensions and lacunarity) \nwere generated within each volume of interest. Corr elation analyses were \nperformed using Spearman correlation analysis. Logi stic regression was used \nto build prediction models. \nResults or Findings: The L1 and L4 were positively correlated with 1p/19 q-\ncodeleted (correlation coefcient: 0.213 and 0.212).  The L5 was positively \ncorrelated with TERT promoter mutation (correlation  coefcient: 0.326). The L1, \nL4, and L9 were negatively correlated with TERT pro moter mutation \n(correlation coefcient: -0.251, -0.310, -0.196, res pectively). The L1, L3, and L4 \nwere positively correlated with Ki-67 proliferation  index (correlation coefcient: \n0.226, 0.200, 0.241, respectively). MGMT promoter m ethylation and P53 had \nno correlation with fractal features. The AUC of fr actal features predicting \n1p/19q-codeleted was 0.677 and predicting TERT prom oter mutation was \n0.755. \nConclusion: The fractal features were correlated with 1p/19q-co deleted, TERT \npromoter mutation, and Ki67 status in IDH-wildtype GBM. Fractal features can \nbe used as non-invasive quantitative parameters to predict the molecular \nfeatures of GBM. \nLimitations: Not \nFunding for this study: This study was supported by the National Natural \nScience Foundation of China (grant no. 82071872 and  82371914), the Science \nand Technology Program of Gansu Province (grant no.  21YF5FA123 and \n21JR11RA105), and the China International Medical F oundation (grant no. Z-\n2014-07-2101). \nEthics committee - additional information: This study was approved by the \nMedical Ethics Committee of the Second Hospital of Lanzhou University \n(approval number : 2020A-070) and informed consent was waived. \nAuthor Disclosures:  \nJunlin Zhou: Nothing to disclose \nBin Zhang: Nothing to disclose \n \n \nT1 Curves in the evaluation of radionecrosis or dis ease recurrence \n*C. Monopoli*, A. Romano, G. De Rosa, A. Romano, G.  Moltoni,  \nA. M. Ascolese, G. Capriotti, A. Bozzao; Rome/IT \n(cristiana.monopoli@uniroma1.it) \n \nPurpose or Learning Objective: Radiation treatment of brain metastases \ncreates diagnostic doubts in the differential diagn osis between disease \nprogression or radionecrosis induced by radiosurger y. MRI with the \nadministration of contrast medium doesn't offer the  possibility of reliably \ndistinguishing the two pathological entities. The a im of our study is to verify the \npresence of radionecrosis or disease recurrence thr ough the evaluation of T1 \nenhancement curves. \n\n \n \nFriday \nAbstract-based Programme \n \n 177  \nMethods or Background: 40 brain metastases undergone to radiosurgery \nwere evaluated (32 from lung, 4 from breast, 2 from  melanoma and 2 \ncolorectal). All patients underwent MRI examination  with dynamic T1 \nacquisitions with contrast medium. For each lesions  the T1 enhancement curve \nwas extracted by positioning a region of interest c orresponding to the solid \ncomponent, excluding the necrotic areas. All patien ts underwent a PET-DOPA \nstudy and the result of the examination was used as  the gold standard to \ndistinguish radionecrosis from disease progression.  The PET investigations \nidentifies three stages of the disease; radionecros is (rSUV<1.6); mixed picture \n(rSUV between 1.6 and 1.9); disease progression (rS UV>1.9) \nResults or Findings: 4 types of T1 enhancement curves have been identifi ed \n(A-D). Curve A showed constant growth over time; cu rve B showed faster \ngrowth in its initial portion and constant growth o ver time; curve C showed \nrapid initial growth and a final plateau; Curve D s howed rapid growth and rapid \nfinal washout. Of the 40 lesions, 13 showed uptake compatible with \nradionecrosis, 12 with a mixed picture and 15 with disease progression. \nCurves A and B corresponded to radionecrosis or mix ed in 90% of cases, \ncurves C and D corresponded to a progression of the  disease in 95% of cases. \nConclusion: T1 enhancement curves allows to distinguish a condi tion of \nradionecrosis from a progression of the disease. \nLimitations: Small enrolled population \nFunding for this study: Not funding received \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nCristiana Monopoli: Nothing to disclose \nAnna Maria Ascolese: Nothing to disclose \nAndrea Romano: Nothing to disclose  \nGabriella Capriotti: Nothing to disclose \nAlessandro Bozzao: Nothing to disclose \nAllegra Romano: Nothing to disclose \nGiulia Moltoni: Nothing to disclose \nGiulia De Rosa: Nothing to disclose \n \n \nPET/MRI in brain primary and secondary tumors treat ed with \nradiochemotherapy: a radiomic-based analysis of bra in Perfusion MRI \nand 11C-Methionine PET images acquired by a integra ted hybrid system \n*E. Masiello*, M. Barbera, F. Fallanca, S. Paola, A . Castellano, A. Falini,  \nN. E. Anzalone; Milan/IT \n(e.masiello@studenti.unisr.it) \n \nPurpose or Learning Objective: Perfusion-weighted MRI (PWI) and 11C-\nmethionine PET (MET-PET) provide valuable hemodynam ic and metabolic \ninsights for assessing brain tumors. This study aim ed to investigate the \ndiagnostic role of PWI and MET-PET, using radiomic analysis, in distinguishing \nprogression (PD), pseudoprogression (PsP), and radi onecrosis (RN) in patients \nwith brain tumors treated with radiotherapy (RT) or  radiochemotherapy. \nMethods or Background: Patients with primary and secondary brain \nneoplasms who developed post-treatment lesions of a t least 1 cm within the \nradiation field were retrospectively enrolled. All patients underwent \nsimultaneous PET/MRI examinations according to a st andardized protocol. \nRadiomics features were extracted from the 3D-segme ntation of parametric \nmaps, including relative cerebral blood volume (rCB V) from DSC, plasma \nvolume (Vp) and vascular permeability (Ktrans) from  DCE, relative cerebral \nblood flow (CBF) based on pseudo-Continuous Arteria l Spin Labeling (pCASL), \nand Standardized Uptake Value (SUV) from MET-PET. F or each lesion, \nimaging data were compared with outcomes based on R ANO criteria or \nhistological examination. \nResults or Findings: A semi-automatic 3D segmentation of 52 lesions (23 PD \nand 29 RN) was performed using PMod software (v. 3. 7) to extract 263 \nradiomic features. After feature selection, CBF Max imum gray level and SUV \nMaximum gray level were the most correlated (p < 0. 001). In terms of \naccuracy, the highest area under the curve (AUC) fo r PET features was SUV \nEntropy, while Vp Mean showed the highest AUC for P WI features. SUV \nEntropy achieved the highest sensitivity for detect ing PD (95.65%), while rCBV \nEntropy demonstrated the highest specificity (79.31 %). PET and PWI \nparameters exhibited similar overall accuracy, rang ing from 71.15% to 76.92%. \nConclusion: Radiomics features from MET-PET and PWI demonstrate  strong \npotential in accurately distinguishing PD and PsP f rom RN. Combining both \nmodalities using radiomics enhances overall diagnos tic accuracy. \nLimitations: Small and heterogeneous population \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective. \nAuthor Disclosures:  \nAndrea Falini: Nothing to disclose \nNicoletta Emanuela Anzalone: Nothing to disclose \nEdoardo Masiello: Nothing to disclose  \nScifo Paola: Nothing to disclose \nAntonella Castellano: Nothing to disclose \nMaurizio Barbera: Nothing to disclose \nFederico Fallanca: Nothing to disclose \n \nMultiparametric MRI-based clinical radiomics model for predicting TERTp \ngenotype and overall survival in oligodendrogliomas  \n*J. Zhao*¹, X. Ke¹, T. Gan¹, W. Hu¹, C. Xue², S. Li ³, Q. Zhou¹, J. Zhou¹; \n¹Lanzhou/CN, ²Qingdao/CN, ³Chengdu/CN \n(gtjlovezj@163.com) \n \nPurpose or Learning Objective: To test the hypothesis that combining \nfeatures from multiparametric MRI with clinically r elevant prognostic risk factors \nprovides a more accurate prediction of TERTp genoty pe and overall \nsurvival(OS)in patients with oligodendrogliomas(OGS ). \nMethods or Background: Preoperative multiparametric MRI sequences \n(T1WI, T2WI, and CE-T1-3D) from 135 patients with O GS (grades 2 and 3) \nwere collected and randomly divided into training ( n = 95) and validation (n = \n40) sets. Radiomics features were extracted, and th e least absolute shrinkage \nand selection operator regression was used to selec t the most relevant \nfeatures. Clinical relevant features identified thr ough univariate and multivariate \nlogistic regression analyses were incorporated to e stablish a clinical radiomics \nmodel. This model was used to develop a predictive TERTp genotype \nnomogram. Kaplan–Meier curves were used to assess O S differences \nbetween TERTp groups; the log-rank test determined significance. \nResults or Findings: The T2WI-based clinical radiomics model demonstrate d \nsuperior performance in predicting the TERTp genoty pe, with mean area under \nthe receiver operating characteristic curve (AUC) v alues of 0.90 (95% CI: 0.88, \n0.92; P = 0.0002) in the training set and 0.83 (95%  CI: 0.82, 0.85; P < 0.0001) \nin the validation set. The one year, two year and t hree year survival probability \nprediction models achieved AUC values of 0.85, 0.80 , and 0.79, respectively. \nConclusion: The multiparametric MRI-based clinical radiomics mo del provides \nthe most accurate prediction of the TERTp genotype.  Combined with clinical \nrelevant prognostic risk factors, the prognostic mo del offers precise prediction \nof OS in patients with OGS. \nLimitations: The limitation of this study was three-dimensional tumor \nsegmentation was manually performed, future researc h should explore \nautomated and efficient segmentation methods to red uce workload and \nsimplify clinical application. \nFunding for this study: Funding were provided by National Natural Science \nFoundation of China (82071872, 82371914), Science a nd Technology Program \nFunding Project of Gansu Province ( 21JR7RA404). \nEthics committee - additional information: The ethics committee notification \ncan be found under the number 2021A-348. \nAuthor Disclosures:  \nShenglin Li: Nothing to disclose \nWanjun Hu: Nothing to disclose \nJun Zhao: Nothing to disclose \nJunlin Zhou: Nothing to disclose \nCaiqiang Xue: Nothing to disclose  \nQing Zhou: Nothing to disclose \nXiaoai Ke: Nothing to disclose \nTiejun Gan: Nothing to disclose \n \n \nIntraoperative and postoperative MRI detection of i schemia in brain \ntumor surgery: a retrospective study on predictive factors and ischemic \nevolution \n*M. R. López De La Torre Carretero*, C. Mbongo, P. Corral Alonso,  \nD. A. Zambrano, Á. R. Cabrera Abud, C. D. Solano, J . M. Rodríguez Ortega, \nM. Macías de la Corte Hidalgo, M. Calvo Imirizaldu;  Pamplona/ES \n(mltorrec@unav.es) \n \nPurpose or Learning Objective: Intraoperative magnetic resonance imaging \n(iMRI) is an increasingly valuable tool in neurosur gical oncology, particularly for \nguiding tumour resection by real-time detection of residual tumour and \nmargins, but also acute ischemic complications usin g diffusion weighted \nimaging (DWI). However, some studies suggest iMRI m ay underestimate \nischemia compared to early postoperative MRI (epMRI ), highlighting the \nimportance of thorough postoperative evaluation. Ou r study aimed to compare \niMRI´s ability to detect ischemia against epMRI and  late postoperative MRI \n(lpMRI), and to identify predictive factors for pos toperative ischemia in patients \nundergoing brain tumour resection. \nMethods or Background: This retrospective study included 106 patients \nundergoing brain tumor resection at our centre. iMR I, epMRI (5-7 days post-\nsurgery), and lpMRI (30 days post-surgery) were per formed. Two radiologists \nanalysed imaging to detect intraoperative ischemia (IOI), early postoperative \nischemia (EPI), and late postoperative ischemia (LP I) using (DWI) and \nquantifying ischaemic volume on apparent diffusion coefficient (ADC) maps. \nVariables such as age, sex and tumour histology wer e recorded. Statistical \nanalysis was conducted (StataNow 18.5) using McNema r tests, Pearson’s chi-\nsquared test, and multivariate logistic regression.  \nResults or Findings:  McNemar test revealed that iMRI tends to \nunderestimate ischaemia compared to epMRI (p= 0.003 9). Only tumour type \nwas a significant predictor of EPI (p=0.041), with glioblastoma patients having \nlower probabilities of EPI compared to other tumour  types (p=0.041). EPI and \n\n \n \nFriday \nAbstract-based Programme \n \n 178  \nLPI were significantly associated (p <0.001), indic ating ischaemia detected in \nepMRI didn’t progress in lpMRI. \nConclusion: Our results suggest iMRI may underestimate the isch emia \ncompared to epMRI. Glioblastoma was associated with  a lower risk of \nischemia, highlighting the importance of personaliz ed surgical approach. The \nstrong relationship between EPI and LPI emphasizes the need for close follow-\nup to monitor late complications. \nLimitations: Retrospective study. \nClinical outcomes could be useful (further studies)  \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nCarmen Mbongo: Nothing to disclose \nJosé Marlon Rodríguez Ortega: Nothing to disclose \nPaula Corral Alonso: Nothing to disclose \nMarta Macías de la Corte Hidalgo: Nothing to disclo se \nManuel Rafael López De La Torre Carretero: Nothing to disclose \nMarta Calvo Imirizaldu: Nothing to disclose \nDaniel Alfonso Zambrano: Nothing to disclose \nÁlvaro Rafael Cabrera Abud: Nothing to disclose \nCarlos Delgado Solano: Nothing to disclose \n \n \nSupratentorial Changes in Patients with Cerebellopo ntine Angle Tumors: \nA Comprehensive Morphologic Analysis \n*A. Çolakoğlu*, B. Genç, K. Aslan, L. Incesu; Samsun/TR \n(ardacolakoglu@hotmail.com) \n \nPurpose or Learning Objective: Neuroplasticity that develops in intracranial \ntumors may guide the management of post-surgical or  radiotherapy treatment. \nTo our knowledge, there is no study investigating m orphometric changes in the \nbrain of patients with cerebellopontine angle tumor s. Our aim in this study is to \ninvestigate supratentorial morphometric changes in patients with \ncerebellopontine angle tumors. \nMethods or Background: The study included 29 patients with \ncerebellopontine angle tumors who had not yet recei ved any treatment, and 53 \nage- and sex-matched healthy controls. Voxel-based morphometry and \nsurface-based morphometry analyses were performed u sing CAT12, running \nunder SPM12, to examine gray matter volume changes and cortical thickness \nchanges in these patients. A general linear model w as used for statistical \nanalysis, and a p-value<0.05 with family-wise error  (FWE) correction was \nconsidered statistically significant. \nResults or Findings: Our VBM results showed an increase in gray matter \nvolume in the thalamus, ventral diencephalon, cingu late gyrus, precuneus, \ncuneus, superior parietal lobe, and parahippocampal  gyrus in these patients \n(p<0.05 FWE). Our SBM results revealed an increase in cortical thickness in \nthe right superior parietal and paracentral gyri in  patients with cerebellopontine \nangle tumors (p<0.05 FWE). \nConclusion: Our study is the first to demonstrate an increase i n gray matter \nvolume and cortical thickness in the supratentorial  region of patients with \ncerebellopontine angle tumors. These findings may b e associated with \nneuroplastic changes in these patients. \nLimitations: The limitations of the study are its single-center design, its \nretrospective nature, and the absence of neurocogni tive tests. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is approved by clinical \nstudies ethics committee of Ondokuz Mayıs Universit y. The reference number \nis 2024090641. \nAuthor Disclosures:  \nBariş Genç: Nothing to disclose  \nLütfi Incesu: Nothing to disclose \nArda Çolakoğlu: Nothing to disclose \nKerim Aslan: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n16:00-17:30 Research Stage 1 \nResearch Presentation Session: Chest \nRPS 1604 \nLung cancer screening and nodule \nassessment \n \nModerator \nD. Shaham; Jerusalem/IL  \n(dshaham@hadassah.org.il) \nAuthor Disclosures:  \nDorith Shaham: Consultant: AstraZeneca \n \n \nAutomated triage to a virtual lung nodule clinic us ing Veye Lung Nodule \nvolumetry could provide a cost saving compared to s tandard care \n*G. Dixon*¹, V. Dattani¹, M. Jordan², J. Rodrigues¹ ; ¹Bath/UK, ²Coventry/UK \n(gilesdixon@nhs.net) \n \nPurpose or Learning Objective: Pulmonary nodules are frequently identified \non non-screening CT chest scans. The management of pulmonary nodules in \nthe UK is governed by the British Thoracic Society (BTS) Guidance 2015. \nArtificial intelligence can provide automated nodul e identification and \nvolumetry. The clinical and cost effectiveness of i ncorporating automated \nnodule detection into patient pathways are currentl y unclear. \nMethods or Background: All CT scans including the thorax over one week in \na single UK centre during March 2022 were identifie d and their original reports \nreviewed. CT scans were categorised as to whether t he original report gave a \nBTS recommendation or not. Currently, patients with out a BTS \nrecommendation are assumed to undergo a full lung M DT discussion to \ndetermine the recommendation. A virtual nodule clin ic was modelled whereby \nall patients with >1 nodule of >80mm3 identified us ing Veye Lung Nodules who \ndid not receive a BTS recommendation in the origina l CT report were referred \nto a virtual nodule clinic. The cost of this approa ch was compared to current \npractice. \nResults or Findings: 80/357 scans had > 1 pulmonary nodule >80mm3 \nidentified using Veye Lung Nodules. 51/80 patients did not receive a BTS \nGuideline based recommendation in the original radi ologist report. A potential \ncost saving of £9.48 per case to review these cases  in a virtual lung nodule \nclinic was calculated. This represents a cost savin g of £483.48 during the study \nperiod (one week) and a potential annual cost savin g of between £25,141 and \n£36,854. \nConclusion: The study identified a potential cost saving of aut omated triage of \nlung nodules into a virtual lung nodule clinic. \nLimitations: Further work is ongoing to model longitudinal costs  of this cohort \nof patients. \nFunding for this study: No specific funding was used for this study \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nGiles Dixon: Speaker: Aidence \nVruti Dattani: Nothing to disclose \nMary Jordan: Nothing to disclose \nJonathan Rodrigues: Other: Aidence Speaker: Aidence  \n \n \nSOLACE - CT acquisition protocols in lung cancer sc reening across \nEurope \n*M. F. G. Konrad*¹, E. Nischwitz¹, V. Palm¹, O. Von  Stackelberg¹,  \nA. Baca-Stera², K. Błasińska², M. Adamek³, J. Chorostowska-Wynimko²,  \nH-U. Kauczor¹; ¹Heidelberg/DE, ²Warsaw/PL, ³Gdansk/ PL \n(mathis.konrad@med.uni-heidelberg.de) \n \nPurpose or Learning Objective: Exploration of the current status of \ninstitutional and technical factors of CT acquisiti on protocols applied in Europe \nthat may affect the radiation exposure of screening  participants in lung cancer \nscreening procedures. \nMethods or Background: To achieve an overview of the applied CT \nacquisition protocols in lung cancer screening acro ss Europe, data were \nacquired from leading investigators responsible for  the definition of CT \nacquisition protocols in the screening centres of t he SOLACE consortium. Data \nregarding institutional and technical factors of CT  acquisition protocols were \ncollected through a baseline survey. All data embod y the current status \nbetween June and October 2024. \nResults or Findings: Survey responses of 16 screening centres from 10 \nEuropean countries (Croatia, Czechia, Estonia, Fran ce, Greece, Hungary, \nIreland, Italy, Poland and Spain) were received. In  relation to institutional \nfactors, the CT acquisition protocols are establish ed and modified by personnel \n\n \n \nFriday \nAbstract-based Programme \n \n 179  \nof multiple professional roles (radiologists, radio graphers, medical physicists, \nmanufacturer personnel). In eight of 16 (50 %) inst itutions the protocol is \nestablished by a multiprofessional team. CT protoco ls are mostly modifiable \n(88 %). Regarding technical factors, the number of detector rows in the z-\ndirection ranges from 16 to 128. In 13 of 16 centre s (81 %) automatic exposure \ncontrol is applied. The applied reconstructed slice  thickness and increment lies \nin the range of 0.625 - 1.25 mm. \nConclusion: Lung cancer screening in Europe within the SOLACE c onsortium \nis widely implemented reaching the technical limits  of currently operated \ndevices. The definition of CT acquisition protocols  as a team effort and the \npossible modification of protocols are subjects of imaginable improvements. \nLimitations: The brief survey data stems only from institutions with research \nfocus. \nFunding for this study: This project is co-funded under the EU4Health \nProgramme 2021–2027 under grant agreement no. 10110 1187 \nEthics committee - additional information: Not a patient study \nAuthor Disclosures:  \nJoanna Chorostowska-Wynimko: Nothing to disclose \nAlicja Baca-Stera: Nothing to disclose \nKatarzyna Błasińska: Nothing to disclose \nHans-Ulrich Kauczor: Nothing to disclose \nOyunbileg Von Stackelberg: Nothing to disclose \nEmily Nischwitz: Nothing to disclose \nViktoria Palm: Nothing to disclose \nMathis Franz Georg Konrad: Other: This contribution  has been prepared on \nbehalf of the SOLACE consortium \nMariusz Adamek: Nothing to disclose \n \n \nVariations in participant positioning, scan directi on and scanogram angle \ninfluence organ-specific radiations doses in routin e low-dose chest CT \nfor lung cancer screening \n*L. D'Hondt*¹, C. Haentjens¹, A. Snoeckx², K. Bache r¹; ¹Ghent/BE, \n²Antwerp/BE \n(louisdho.dhondt@ugent.be) \n \nPurpose or Learning Objective: In lung cancer screening (LCS), non-ideal \npositioning, changes in scan direction or scanogram  angles are likely to occur \ndue to the high turnover of participants and large volume of scans performed. \nAdditionally, since we are dealing with healthy ind ividuals, careful management \nof radiation dose is crucial. This study aims to sy stematically simulate how \nparticipant positioning and scanning parameters aff ect organ-specific radiation \ndoses. \nMethods or Background: Using the Alderson Rando phantom, we performed \nCT scans under varying conditions on two scanners ( GE Revolution, Siemens \nSOMATOM Definition Flash) to establish automatic tu be current modulation \n(ATCM) variations. ImpactMC Monte Carlo software si mulated low-dose scans \nfor 32 patient-specific voxel models, calculating p ercentage dose differences to \nlungs, heart, thyroid, liver, and female breasts. D eviations included \nlateral/vertical mispositioning (2cm increments), s can direction changes \n(craniocaudally, caudocranially), and varying scano gram angles \n(anteroposterior, posteroanterior, lateral, combine d). \nResults or Findings: Vertical deviations caused linear increase in organ  \ndoses when positioned closer to the tube during sca nogram, with the GE \nscanner showing a 12-24% increase per 2cm deviation s. For the Siemens \nscanner this linear increase was less pronounced (3 -9% increase per 2cm \ndeviation). Lateral mispositioning increased organ doses by <10% for both \nscanners. Caudocranial scanning only increased thyr oid doses by 15.8% for \nGE, while Siemens in fact showed a 19.7% decrease i n thyroid dose and \nsignificant increases in lung (18.7%), heart (28.7% ), liver (32.6%), and breast \n(27.9%) doses. For both scanners, the highest doses  occurred with a \nposteroanterior scanogram. Siemens showed 72% dose reduction with dual \nscanogram, while GE showed minimally 34% reduction with an anteroposterior \nscanogram. \nConclusion: Non-ideal positioning and varying CT parameters can  \nsignificantly affect organ doses, potentially under estimating the anticipated \norgan-specific doses and related radiation-induced cancer risk in LCS. \nLimitations: Limitations are use of a phantom. \nFunding for this study: Funding was provided by the FWO “Kom op tegen \nKanker”-project for lung cancer screening research in Belgium. (Project \nnumber: G0B1922N). \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nKlaus Bacher: Nothing to disclose \nLouise D'Hondt: Nothing to disclose \nAnnemiek Snoeckx: Nothing to disclose \nClaudia Haentjens: Nothing to disclose \n \n \n \n \nPilot National Lung Cancer Screening in Poland – su mmary of the \nnorthern macroregion results - continuation of the Gdansk experience \n*J. Bidzinska*, J. Rymarowicz, K. Kobyłecka, K. Dzi adziuszko, E. Szurowska, \nW. Rzyman; Gdansk/PL \n(jbidzinska@gumed.edu.pl) \n \nPurpose or Learning Objective: The aim of he study was to evaluate the \nresults of the Pilot National Lung Cancer Screening  program in the northern \nmacroregion of Poland. \nMethods or Background: After primary prevention, lung cancer screening is \nthe strategy to reduce lung cancer-related mortalit y. A total of 3068 \nasymptomatic volunteers between 50 and 74 years of age with a smoking \nhistory of at least 20 pack years underwent screeni ng with the use of low-dose \ncomputed tomography of the chest. The radiological nodule management \nprotocol was designed for a 2-year follow-up. It wa s created specifically for the \nnational program and implemented nationwide. In the  northern macroregion, \nLDCT was performed in 4 radiological centers. All r adiologists were trained. \nThe program included an anti-smoking intervention. \nResults or Findings: At the time of submitting this abstract, 59 (1.92%)  \nparticipants of the Program were diagnosed with lun g cancer. In this group, 43 \nparticipants had resectable lung cancer, which is 7 3% of cases, including 22 \nwomen and 21 men. Few participants undergo diagnost ic workups. In addition, \nwe assessed (or the analysis is ongoing) the lung c ancer screening adherence \nrate, number and type of lung nodules, presence and  severity of emphysema, \nCAC, incidental findings, and effectiveness of smok ing cessation intervention. \nConclusion: Lung cancer screening with low-dose computed tomogr aphy of \nthe chest is an effective tool for reducing lung ca ncer-related mortality. LDCT is \nan effective tool for the identification of comorbi dities. It should be implemented \nas a National Lung Cancer Screening Strategy. \nLimitations: Quality control in one of the sites revealed that t he radiologist \nshould be re-trained periodically. \nFunding for this study: The program was financed by the Ministry of Health.  \nThe lung cancer prevention program was implemented in accordance with the \ncompetition regulations No. POWR.05.01.00-IP.00-010 /19 Operational \nProgram Knowledge Education and Development 2014-20 20 Priority axis V \nSupport for the health area Measure 5.1 Prevention programs. \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board of the Medical Universit y of Gdańsk \n(NKBBN/72/2020). During realisation of the program we asked all aprticipants \nfor the consent to use of data in the future analys es and studies. Only 8 \nparticipanst did not consent. \nAuthor Disclosures:  \nKatarzyna Dziadziuszko: Nothing to disclose \nKatarzyna Kobyłecka: Nothing to disclose \nJulia Rymarowicz: Nothing to disclose \nWitold Rzyman: Nothing to disclose  \nEdyta Szurowska: Nothing to disclose \nJoanna Bidzinska: Nothing to disclose \n \n \nImplications of 2D single slice vs. 3D whole-chest body composition \nphenotyping on outcome prediction: insights from th e National Lung \nScreening Trial \n*J. Jahn*, F. B. Pallasch, M. Jung, M. Reisert, F. Bamberg, J. Weiß; \nFreiburg/DE \n(johannes.jahn94@gmx.de) \n \nPurpose or Learning Objective: Body composition (BC) is linked to outcomes \nin cardiovascular (CV) disease and cancer. BC is ty pically estimated from a 3rd \nlumbar slice, but the role of standardized chest he ights and 3D volumes is \nunclear. Here, we used a fully automated deep learn ing network 1) to \ninvestigate the correlation between 2D single-slice  areas and 3D whole-chest \nvolumes and 2) to explore their association with mo rtality in heavy smokers. \nMethods or Background: Using baseline data from the National Lung \nScreening Trial, BC was estimated on chest CT as sk eletal muscle (SM), \nintramuscular (IMAT), and subcutaneous adipose tiss ue (SAT). Correlations \nbetween 2D thoracic vertebra slices and 3D chest vo lumes were explored. \nAssociations between BC measures and all-cause mort ality were assessed for \nboth approaches, secondarily for CV and lung cancer  mortality. Kaplan-Meier \ncurves (categories <15%; 15-85%; >85%) and Cox regr ession models adjusted \nfor demographics and CV risk factors were used. \nResults or Findings: Among 23,361 individuals (mean age 61.4±5, 41.6% \nfemale), 1,616 (6.9%) all-cause deaths occurred ove r a median follow-up of 6.5 \nyears. The highest correlation between 2D and 3D vo lumes was at T4 \nvertebra. For 3D volumes, lower SM, lower SAT, and higher IMAT were \nassociated with higher mortality in Kaplan-Meier cu rves (p<0.01). Associations \nbetween these groups remained robust after multivar iable adjustment \n(adjusted hazard ratio (aHR): 1.51, 95% CI 1.33-1.7 1, p<0.001; aHR: 1.21, \n95% CI: 1.04-1.40, p<0.05; aHR: 1.68, 95% CI: 1.36- 1.78, p<0.001, \nrespectively). Largely similar patterns were found for 2D measures, CV, and \nlung cancer mortality. \n\n \n \nFriday \nAbstract-based Programme \n \n 180  \nConclusion: BC measures independently predict mortality in heav y smokers \nbeyond traditional clinical risk factors. 2D measur es at T4 and 3D BC showed \nsimilar results, offering interchangeable use for i dentifying high-risk individuals \nin lung cancer screening, potentially enhancing per sonalized prevention. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: All National Lung Screening Trial \nparticipants provided written informed consent for the original trial and \nsecondary use of the data was approved by the local  IRB. \nAuthor Disclosures:  \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nMatthias Jung: Nothing to disclose \nFabian Bamberg: Speaker: Bayer Healthcare Research/ Grant Support: \nSiemens Healthineers Consultant: Bayer Healthcare S peaker: Siemens \nHealthineers Research/Grant Support: Bayer Healthca re \nFabian Bernhard Pallasch: Nothing to disclose \nJohannes Jahn: Nothing to disclose \n \n \nPhoton-Counting Detector CT Provides Superior Subso lid Nodule \nCharacterization Compared to Same-Day Energy-Integr ating Detector CT \n*J. Wang*, L. Song, W. Song; Beijing/CN \n(981891153@qq.com) \n \nPurpose or Learning Objective: To investigate the performance of photon-\ncounting detector (PCD) CT compared to conventional  energy-integrating \ndetector (EID) CT in identifying subsolid nodule (S SN) characteristics. \nMethods or Background: Participants with SSNs who underwent same-day \nEID CT and PCD CT between October 2023 and April 20 24 were prospectively \nincluded. The 1.0 mm EID CT images and, subsequentl y, 1.0, 0.4, and 0.2 mm \nPCD CT images were reviewed to assess image noise a nd subjective image \nquality on a 5-point Likert scale. SSN characterist ics, including lobulation, \nspiculation, pleural retraction, air cavities, intr a-nodular vessel sign, internal \nvascular changes, and heterogeneous solid component s, were evaluated. \nResults or Findings: Forty-eight participants (mean age: 56 ± 11 years; 16 \nmales) with 89 SSNs were included. PCD CT significa ntly reduced radiation \ndose when using matched scans (1.79 ± 0.39 vs. 2.17 ± 0.57 mSv, P < 0.001). \nCompared to 1.0 mm EID CT, 1.0 mm PCD CT images exh ibited significantly \nlower objective image noise and higher subjective i mage quality (all P < 0.001). \nCompared to EID CT, PCD CT demonstrated enhanced vi sualization of subtle \ncharacteristics, except for lobulation, with a 0.4 mm section thickness offering a \nfavourable balance between ultra-high resolution an d perceived image quality \nfor radiologists. \nConclusion: PCD CT facilitated radiation dose reduction and out performed \nconventional EID CT in terms of image quality and v isualization of SSN \ncharacteristics. \nLimitations: First, the primary focus was on SSNs with various i maging \ncharacteristics suggestive of malignancy; consequen tly, SSNs < 6 mm were \nnot included in the study. Second, quantitative SSN  analysis was not \nperformed because the study aimed to explore the qu alitative value of PCD CT \nin assessing SSN characteristics. Third, evaluation  of nodules on a mediastinal \nwindow was not performed in this study. Fourth, not  all SSNs yielded \nhistopathologic results. \nFunding for this study: This work has received funding by the National \nNatural Science Foundation of China (NSFC No. 82171 934) and the National \nHigh Level Hospital Clinical Research Funding (2022 -PUMCH-B-069). \nEthics committee - additional information: The Institutional Review Board of \nour hospital approved this prospective study [I-23P J1459]. \nAuthor Disclosures:  \nWei Song: Nothing to disclose \nLan Song: Nothing to disclose \nJinhua Wang: Nothing to disclose \n \n \nDiagnostic Outcomes of Korean Lung-RADS Category 2b  Nodules in \nLung Cancer Screening: A Single Tertiary Hospital E xperience \n*Y. Kim*, T. Ha, S. You, J. Sun; Suwon/KR \n(dbsk93@gmail.com) \n \nPurpose or Learning Objective: To evaluate the clinical outcomes of Korean \nLung-RADS category 2b nodules in lung cancer screen ing and to compare the \ndiagnostic performance between Korean Lung-RADS and  Lung-RADS. \nMethods or Background: This retrospective study included 2,908 participant s \nwho underwent low-dose chest CT scans at a single t ertiary hospital between \nSeptember 2019 and January 2024, as part of the Kor ean national lung cancer \nscreening program. The modified Korean Lung-RADS ad ded category 2b for \nnodules sized as category 3 or 4 but likely benign,  aiming to reduce false \npositives. All CT scans were interpreted using the Korean Lung-RADS \ncategorization (version 1.0 until 2021, version 1.1  from 2022, and version 2022 \nfrom 2024). To assess interobserver agreement, thre e radiologists \nindependently reassessed each 2b nodule, indicating  whether they would \nclassify it as category 2b. Diagnostic performance comparisons were \nconducted using McNemar and permutation tests, whil e interobserver \nagreement was analyzed using multirater Fleiss κ statistics. \nResults or Findings: 1,270 participants with confirmed diagnoses were fi nally \nincluded (mean age, 61.5 ± 5.2 years; 1,243 men). Lung cancer was identified \nin 28 participants (2.2%). Category 2b nodules were  found in 24 participants \n(1.9%), with one diagnosed as lung cancer (4.2%) an d the others benign \n(95.8%). Interobserver agreement was fair with a Fl eiss kappa value of 0.36 \n(95% CI 0.17-0.79). In terms of diagnostic performa nce, Korean Lung-RADS \nshowed significantly higher specificity (92.5%, 1,1 49 of 1,242) compared to \nLung-RADS (90.7%, 1,126 of 1,242; p < 0.001), while  sensitivity was \ncomparable between Korean Lung-RADS (89.36%, 25 of 28) and Lung-RADS \n(92.8%, 26 of 28; p = 1). \nConclusion: Korean Lung-RADS category 2b helped reduce the fals e-positive \nrate in lung cancer screening. However, significant  variability was observed \namong radiologists in classifying category 2b nodul es. \nLimitations: Retrospective study with small sample size. \nFunding for this study: No \nEthics committee - additional information: This retrospective study was \napproved by the institutional review board, which w aived the requirement for \nwritten informed consent (AJOUIRB-DB-2024-399). \nAuthor Disclosures:  \nYouna Kim: Nothing to disclose \nJoosung Sun: Nothing to disclose \nSeulgi You: Nothing to disclose \nTaeyang Ha: Nothing to disclose \n \n \nSafety of CT-guided core needle biopsy in patients with interstitial lung \nabnormalities (ILAs) \n*M. Balbi*¹, S. Capelli², A. Caroli², N. C. Culasso ¹, R. Senkeev¹, D. Morbidelli¹, \nG. Reboli¹, L. Righi³, A. Veltri¹; ¹Turin/IT, ²Rani ca/IT, ³Orbassano/IT \n(balbi.m@libero.it) \n \nPurpose or Learning Objective: To evaluate the safety of CT-guided core \nneedle biopsy (CNB) in patients with interstitial l ung abnormalities (ILAs). \nMethods or Background: Consecutive CT-guided pulmonary biopsies \nperformed at the San Luigi Gonzaga Hospital (Orbass ano, Italy) from February \n2010 to December 2023 (n=3251) were retrospectively  reviewed to identify \npatients with ILAs who underwent CNB for the assess ment of a pulmonary \nlesion (n=73, case group). A control group of 73 CN B patients with no evidence \nof ILAs, matched for age, sex, emphysema severity, and lesion depth and \ndimensions, was selected to compare the complicatio n rates. Logistic \nregression was performed to identify risk factors f or complications within the \ncase group, as well as to compare complication rate s between the case and \ncontrol groups. \nResults or Findings: Complications occurred in 21/73 cases (29%) and 24/ 73 \ncontrols (33%), including 2 major complications in cases (3%) and 3 in controls \n(4%). Considering the overall occurrence of complic ations, there were no \nstatistically significant differences between cases  and controls, as assessed by \nFisher’s exact test (p=0.72) and conditional logist ic regression adjusted for \nmatched variables (p=0.72). Among patient, procedur al, and lesion-related \ndata, three variables were found to be risk factors  for complications in the case \ngroup based on univariate, multivariate, and stepwi se logistic regression with \nbidirectional elimination, guided by the Akaike inf ormation criterion (AIC): \nlonger time of the needle within the lung (odds rat io (OR), 1.17; 95% \nconfidence interval (CI), 0.72–13.02; p = 0.126), n eedle traversal of ILAs (OR, \n7.04; 95% CI, 2.07–26.28; p = 0.002), and multiple pleural passes (OR, 8.06; \n95% CI, 1.26–70.46; p = 0.035). \nConclusion: CT-guided CNB in patients with ILAs showed safety c omparable \nto patients without ILAs. However, crossing ILAs du ring the procedure may \nincrease the complication risk. \nLimitations: -Retrospective, single-center study -Limited number  of cases \n-Lack of inter-reader agreement \nFunding for this study: The study did not receive any funding. \nEthics committee - additional information: The institutional review board \n(Comitato Etico Territoriale Interaziendale of the AOU Città della Salute e della \nScienza di Torino) approved this retrospective obse rvational case-control study \nand waived the requirement for written informed con sent. \nAuthor Disclosures:  \nSerena Capelli: Nothing to disclose \nLuisella Righi: Nothing to disclose \nAndrea Veltri: Nothing to disclose \nRouslan Senkeev: Nothing to disclose \nNoemi Cristina Culasso: Nothing to disclose \nDavide Morbidelli: Nothing to disclose \nAnna Caroli: Nothing to disclose \nMaurizio Balbi: Nothing to disclose \nGiulia Reboli: Nothing to disclose \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 181  \nDeep Learning-Based Image Domain Reconstruction Enh ances Image \nQuality and Pulmonary Nodule Detection in Ultralow- Dose CT with \nAdaptive Statistical Iterative Reconstruction-V \n*K. Ye*¹, B. Pan², J. Li³, Z. Pan², H. Yuan¹, N-J. Gong¹; ¹Beijing/CN, \n²Shagnhai/CN, ³Fujian/CN \n(yekaibysy@bjmu.edu.cn) \n \nPurpose or Learning Objective: To evaluate the image quality and lung \nnodule detectability of ultralow-dose CT (ULDCT) wi th adaptive statistical \niterative reconstruction-V (ASiR-V) post-processed using a deep learning \nimage reconstruction (DLIR)-based image domain comp ared to low-dose CT \n(LDCT) and ULDCT without DLIR. \nMethods or Background: A total of 210 patients undergoing lung cancer \nscreening underwent LDCT (mean ± SD, 0.81 ± 0.28 mSv) and ULDCT (0.17 ± \n0.03 mSv) scans. ULDCT images were reconstructed wi th ASiR-V (ULDCT-\nASiR-V) and post-processed using DLIR (ULDCT-DLIR).  The quality of the \nthree CT images was analyzed. Three radiologists de tected and measured \npulmonary nodules on all CT images, with LDCT resul ts serving as references. \nNodule conspicuity was assessed using a five-point Likert scale, followed by \nfurther statistical analyses. \nResults or Findings: A total of 463 nodules were detected using LDCT. Th e \nimage noise of ULDCT-DLIR decreased by 60% compared  to that of ULDCT-\nASiR-V and was lower than that of LDCT (p <0.001). The subjective image \nquality scores for ULDCT-DLIR (4.4[4,1, 4.6]) were also higher than those for \nULDCT-ASiR-V (3.6[3.1, 3.9]) (p <0.001). The overal l nodule detection rates \nfor ULDCT-ASiR-V and ULDCT-DLIR were 82.1% (380/463 ) and 87.0% \n(403/463), respectively (p <0.001). The percentage difference between \ndiameters >1 mm was 2.9% (ULDCT-ASiR-V vs. LDCT) an d 0.5% (ULDCT-\nDLIR vs. LDCT) (p =0.009). Scores of nodule imaging  sharpness on ULDCT-\nDLIR (4.0 ± 0.68) were significantly higher than those on ULDCT-ASiR-V (3.2 ± \n0.50) (p <0.001). \nConclusion: DLIR-based image domain improves image quality, nod ule \ndetection rate, nodule imaging sharpness, and nodul e measurement accuracy \nof ASiR-V on ULDCT. \nLimitations: Only ASiR-V algorithm was evaluated, and more diffe rent IR \nalgorithms should be included in the future. \nFunding for this study: With improved image quality and a higher nodule \ndetection rate, deep learning-based image domain re construction is beneficial \nin facilitating the use of ultralow-dose CT in lung  cancer screening. \nEthics committee - additional information: Peking university third hospital \nreview board \nAuthor Disclosures:  \nHuishu Yuan: Nothing to disclose \nKai Ye: Nothing to disclose \nBoyang Pan: Nothing to disclose \nJie Li: Nothing to disclose \nZhenglin Pan: Nothing to disclose \nNan-Jie Gong: Nothing to disclose \n \n \n16:00-17:30 Research Stage 2 \nResearch Presentation Session: \nMusculoskeletal \nRPS 1610 \nImaging of the peripheral joints: shoulder \nto foot \n \nModerator \nA. Serfaty; Cabo Frio, Rio de Janeiro/BR  \n(alineserfaty@gmail.com) \n \n \nRelationship Between Coraco-Glenoid Ligament Varian ts and SLAP \nLesions: Insights from a 3-T Arthro-MRI Study \n*M. Curti*, A. Cozzi, V. Chianca, S. Rizzo, F. Del Grande; Lugano/CH \n(curti.marco.33@gmail.com) \n \nPurpose or Learning Objective: The insertion site of the coraco-glenoid \nligament (CGL) of the shoulder seems to be involved  in superior labrum \nanterior and posterior (SLAP) lesions. Anatomical a nd arthro-magnetic \nresonance imaging (MRA) studies have reported five CGL variants. This study \naimed to evaluate inter-reader reliability in the i dentification of CGL presence \nand variants on 3-T MRA, also investigating potenti al associations between \nCGL variants and SLAP lesions. \nMethods or Background: In this retrospective cohort study, three board-\ncertified musculoskeletal radiologists (20, 10, and  5 years of experience) \nevaluated 1136 consecutive MRA examinations, 577 (5 0.2%) of patients with \nSLAP lesions. Inter-reader reliability in assessing  CGL presence and variants \nwas evaluated with Fleiss’ κ. After 1:1 patient matching according to age and \nsex, the association between the presence of SLAP l esions and CGL variants \nwas evaluated with binary logistic regression, calc ulating odds ratios (ORs) \nand their 95% confidence intervals (CI). \nResults or Findings: “Almost perfect” reliability was found for the iden tification \nof CGL variants (κ 0.832, 95% CI 0.813–0.851), the most frequent bein g type I \n(393/1136 patients, 34.6%), and type II (320/1136 p atients, 28.2%). 1:1 \nmatching resulted in two groups of 427 patients wit h and without SLAP lesions \n(256 males in each group; median age 52 years, IQR 44–59 years). Among \nthese 854 matched patients, compared to the type I CGL variant, all other CGL \nvariants had high ORs for the presence of SLAP lesi ons, ranging from OR 4.3 \n(95% CI 2.5–7.5) of type III to OR 39.9 (95% CI 24. 5–65.0) for type II. \nConclusion: 3-T MRA grants high reliability in identifying CGL presence and \nvariants. In an age and sex matched comparison, the  type II CGL variant was \nstrongly associated with the presence of SLAP lesio ns. \nLimitations: Single-center study \nFunding for this study: None \nEthics committee - additional information: Local etichs commitee \nAuthor Disclosures:  \nAndrea Cozzi: Nothing to disclose \nVito Chianca: Nothing to disclose \nStefania Rizzo: Nothing to disclose \nFilippo Del Grande: Nothing to disclose \nMarco Curti: Nothing to disclose \n \n \nCT Findings and the Time-course of Myositis Ossific ans in Short Rotator \nMuscles \n*A. Fujii*, M. Katsumata, C. Sato, S. Tsukahara, T.  Wada, M. Yamamoto,  \nH. Kondo, H. Oba, A. Yamamoto; Tokyo, Japan/JP \n \nPurpose or Learning Objective: Non-traumatic myositis ossificans (MO) in \nthe pelvic region in patients with neurological dis orders are classically reported \non radiographs. However, CT findings have been rare ly reported and the time-\ncourse has not been proven. The purpose is to inves tigate the patients’ \nbackground, CT findings and the time-course of MO i nvolving the short \nexternal rotator muscles (SERMs) of the hip. \nMethods or Background: Of patients hospitalized for acute cerebrovascular \ndisease between 2001 and 2023 in our hospital, 1,46 9 CT scans were \nperformed for 436 patients. Swelling, calcification  and ossification of SERMs, \nevaluated for each hip by two radiologists, were ju dged as calcification when \nthe average density of the lesion was above 100HU, and as ossification when \nbone marrow fat was confirmed. Medical charts were reviewed for their clinical \ncourse and neurological findings. \nResults or Findings: SERMs swelling was found in 22 cases, 29 hips (age \n61.0±3.0, 9 males, 13 females). Follow-up CT scans were examined in 6 \n(6/436, 1.4%) (aged 35–72, mean 54.7±12.7, 2 male, 4 female). All showed \nipsilateral limb paralysis due to cerebral hemorrha ge. Swelling was detected in \n8 hips of 6 patients, 4 unilaterally (right:left, 3 :1) and 2 bilaterally. Follow-up CT \nrevealed calcification in all swelling lesions in S ERMs . The post-onset time-\ncourse in 6 hips was 13–36 days for swelling, 23–53  days for calcification, and \n49–360 days for ossification. In 2 hips, calcificat ion disappeared on follow-up \nCT. \nConclusion: CT imaging illustrated the drastic changes in the m aturation \nprocess of SERMs-MO. Understanding the time-course of SERMs-MO is \nessential in the management, and frequency of follo w-up CT for distinguishing \nneoplasms and infectious conditions. \nLimitations: This was a retrospective study of a few patients. T he follow-up \ndurations varied between patients because examinati ons of the lesions were \nnot always performed. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by \nTeikyo University Medical Research Ethics Committee (23-030). \nAuthor Disclosures:  \nHiroshi Oba: Nothing to disclose \nSatoshi Tsukahara: Nothing to disclose \nMegumi Katsumata: Nothing to disclose \nAkio Fujii: Nothing to disclose \nAsako Yamamoto: Nothing to disclose \nChiaki Sato: Nothing to disclose \nMasayoshi Yamamoto: Nothing to disclose \nHiroshi Kondo: Nothing to disclose \nTakeshi Wada: Nothing to disclose \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 182  \nPredicting the Surgical Requirement for Bone Substi tutes in Distal \nRadius Fractures: Evaluation of Dual-Energy-Derived  Metrics \nL. D. Grünewald, V. Koch, *S. Mahmoudi*, S. Martin,  J-E. Scholtz, I. Yel,  \nC. Booz, T. Vogl; Frankfurt/DE \n(scherwin.mahmoudi@gmail.com) \n \nPurpose or Learning Objective: Distal radius fractures (DRF) are commonly \ndiagnosed in emergency departments. Depending on fr acture characteristics \nand patient factors, bone substitutes may be needed , a decision usually made \nduring surgery. However, preoperative preparations,  such as obtaining \ninformed consent, are necessary. This study aimed t o evaluate metrics from \nroutine CT scans as surrogates for bone texture to predict bone substitute use \nin surgery. \nMethods or Background: Distal radius scans of patients who underwent dual-\nenergy CT (DECT) between 01/2016 and 08/2021 were r etrospectively \nanalyzed. Cortical HU, trabecular HU, cortical thic kness, and DECT-based \nbone mineral density (BMD) were measured. Patient r ecords and follow-up \nimages were reviewed to determine bone substitute u se during surgery. \nReceiver-operating characteristic (ROC) analysis id entified AUC values for \nBMD, HU values, and cortical thickness, while logis tic regression assessed \ntheir association with bone substitute use. \nResults or Findings: A total of 263 patients (median age, 52 years; 132 \nwomen; 192 fractures) were included. ROC analysis s howed a higher AUC for \nDECT-derived BMD compared to cortical HU, trabecula r HU, and cortical \nthickness (0.87 vs. 0.62, 0.52, and 0.60, respectiv ely; P < .001). Logistic \nregression confirmed a significant association betw een lower DECT-derived \nBMD and the use of bone substitutes (Odds Ratio, 0. 94; P = .02), while cortical \nHU, trabecular HU, and cortical thickness were not significantly associated \n(P > .05). \nConclusion: Routine CT scans of the distal radius can predict t he use of bone \nsubstitutes in DRF surgical management, facilitatin g preoperative planning. \nDECT-derived BMD offers superior predictive perform ance compared to \ncortical HU, trabecular HU, and cortical thickness.  \nLimitations: Preselection bias, as patients received Xrays befor e CT. \nFunding for this study: No funding was received for this study \nEthics committee - additional information: Consent waived due to the \nretrospective nature of the study \nAuthor Disclosures:  \nSimon Martin: Nothing to disclose \nChristian Booz: Nothing to disclose \nIbrahim Yel: Nothing to disclose \nThomas Vogl: Nothing to disclose \nJan-Erik Scholtz: Nothing to disclose \nVitali Koch: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \n \n \nEvaluation of findings in recalcitrant tennis elbow , to rule out SMILE \nlesion (Symptomatic Micro-Instability In Lateral El bow) - a retrospective \nstudy: \n*S. Rajan*, J. S. Chatha, H. Mahajan; New Delhi/IN \n(drsriramrajan@gmail.com) \n \nPurpose or Learning Objective: Purpose: Recalcitrant lateral epicondylitis, \noften presents diagnostic challenges, particularly in ruling out symptomatic \nmicro-instability in the lateral elbow (SMILE lesio n). This study aims to evaluate \nthe radiological findings in a retrospective cohort , focusing on defining MRI \nfindings with arthroscopically confirmed SMILE lesi ons. \nMethods or Background: Methods: A total of 313 consecutive elbow MRIs \nreferred for tennis elbow of which 7 arthroscopical ly proven cases of SMILE \nlesions were identified as test cases, and key imag ing findings were \ndocumented as a checklist. The remaining 306 cases were randomized and \nanonymized and analyzed by two experienced radiolog ists (9 and 22 years of \nexperience). On MR, key findings included lateral u lnar collateral ligament \n(LUCL) tear, partial tear of the extensor carpi rad ialis brevis (ECRB), widened \nulnotrochlear joint or radiocapitellar posterior sh ift . Statistical significance of \nfindings was assessed using Chi-square tests for ca tegorical variables and \nFisher's exact test for small sample sizes. \nResults or Findings: Results: Of the 306 evaluated cases, 122 (25.8%) \nshowed partial LUCL tears, 12 (3.9%) had complete L UCL tears, 268 cases \n(87.5%). had deep surface partial tears of ECRB. St atistical analysis revealed \na significant association between the presence of L UCL tears (partial and \ncomplete) and abnormal radiocapitellar alignment (p  < 0.01). Abnormal \nulnotrochlear alignment was significantly associate d with the presence of \nECRB tears (p < 0.05). Intra-articular abnormalitie s included posterior radial \nsynovial plica in 22 cases (7.18%), synovitis in 3 cases (0.98%), and chondral \nlesions in the radial head in 39.78% of cases, with  minimal occurrence of \nlateral capitellar lesions (0.65%). \nConclusion: Conclusions: The study underscores a significant as sociation \nbetween MRI findings such as LUCL tears, ECRB injur ies, and joint \nmisalignments. \nLimitations: Arthroscopic proof was not available in the cases. MR \narthrography could possibly have helped. \nFunding for this study: None \nEthics committee - additional information: Retrospective study using HIPAA \nguidelines \nAuthor Disclosures:  \nJagneet Singh Chatha: Nothing to disclose \nSriram Rajan: Nothing to disclose \nHarsh Mahajan: Nothing to disclose \n \n \nDetecting and differentiating bone marrow edema in the knee using dual-\nenergy CT with water-calcium and water-hydroxyapati te material \ndecomposition \nW. Xiong¹, J. Han², G. Zhang², *T. Wang*², M. Cheng ¹, L. Hu¹, W. Li¹, Z. Ding¹, \nX. He¹; ¹Nanchang/CN, ²Shanghai/CN \n(tiantian.wang@cri-united-imaging.com) \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of \ndual-energy CT (DECT) with water-calcium and water- hydroxyapatite (HAP) \nmaterial decomposition for quantitatively detecting  and differentiating bone \nmarrow edema (BME) in knee. \nMethods or Background: This retrospective study included 26 patients who \nunderwent DECT and magnetic resonance imaging (MRI)  on the same day for \nthe knee. The DECT images were post-processed to ge nerate water-calcium \nand water-HAP material decomposed images, on which two readers \nindependently measured the water mass density of ed ema zone and \ncontralateral normal bone marrow by drawing ellipse  regions of interest. The \nMRI was served as the reference for presence and ex tent of BME. The edema \nextent was graded as mild or severe by visual asses sment. The diagnostic \nperformance of DECT was evaluated using the receive r operating \ncharacteristic (ROC) analysis. \nResults or Findings: Thirty-two bones were confirmed with BME at MRI, \ncaused by fractures (n=16), arthritis (n=10) and br uises (n=6). The mean water \nmass density of severe BME, mild BME and normal bon e marrow was 1073.9 \n± 15.2, 1052.9 ± 13.0, and 1012.5 ± 22.4 mg/cm3 (p <0.001), respectively, on \nwater-calcium images; and 1012.9 ± 15.6, 979.7 ± 23.8, and 959.1 ± 12.8 \nmg/cm3 (p <0.001), respectively, on water-HAP image s. In detecting the \npresence of BME, the area under the ROC curve (AUC)  of water-calcium and \nwater-HAP images was 0.976 and 0.885 (p = 0.001), r espectively, and the \naccuracy was 89.1% and 79.7%. In differentiating se vere and mild BME, the \nAUC of the two decomposed images was 0.836 and 0.89 6 (p = 0.318), \nrespectively, with accuracy of 70.3% and 87.5%. \nConclusion: Dual-energy CT with water-calcium decomposition sho wed \nsuperior diagnostic performance for quantitatively detecting BME in knee, while \nwater-HAP decomposition showed better performance i n differentiating severe \nand mild BME. \nLimitations: Not applicable. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study was a retrospective \nstudy. \nAuthor Disclosures:  \nWenfeng Li: Nothing to disclose \nZhichao Ding: Nothing to disclose \nTiantian Wang: Employee: Central Research Institute , United Imaging \nHealthcare \nXi He: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nWei Xiong: Nothing to disclose \nLeiming Hu: Nothing to disclose \nJintao Han: Nothing to disclose \nMaochu Cheng: Nothing to disclose \n \n \nSynthetic Multiplanar Reformation Generates Quantit atively Comparable \nKnee MR Images to Standard-Of-Care Acquisitions \nT. C. Arnold, L. Wang, L. Yao, A. Shankaranarayanan ,  \n*S. Pasumarthi Venkata*; Menlo Park, CA/US \n(srivathsa@subtlemedical.com) \n \nPurpose or Learning Objective: Knee MRI protocols acquire multiple \nacquisition planes for three main 2D sequences: T1w , T2w, and PD. However, \nprotocol sequence number is limited by time constra ints, meaning some \ncontrast and orientation combinations are not avail able. Here, we develop a \ndeep-learning-based method to synthesize 2D sequenc es using \ncomplementary contrast and acquisition plane inform ation from two input \nimages, and quantitatively compare synthesized and ground truth acquisitions \nMethods or Background: We retrospectively analyzed data from 50 knee MRI \nparticipants. Each contained three 2D sequences: ax ial T2 fat-sat, coronal T2 \nfat-sat, and coronal PD. The coronal T2 fat-sat ser ved as our target sequence \nand the remaining sequences were our network inputs . All images were \ncoregistered to the target image. Patients were div ided into training (N=34), \n\n \n \nFriday \nAbstract-based Programme \n \n 183  \nvalidation (N=8), and testing (N=8) sets. Quantitat ive metrics (PSNR, SSIM, \nand NMI) measured the similarity between the synthe sized coronal T2 and \nacquired images. \nResults or Findings: The algorithm demonstrated excellent quantitative \nperformance when comparing acquired and synthesized  coronal T2 fat-sat \nimages across all three metrics (SSIM: 0.86±0.13, P SNR: 29.98±5.93 dB, and \nNMI: 1.33±0.10). We also compared the synthesized i mages to a coronal \nreformat of the acquired axial T2 fat-sat (SSIM: 0. 68±0.22, PSNR: 23.33±5.21 \ndB, and NMI: 1.21±0.10) to illustrate the significant benefit of our algorithm \nover simple image reformation (paired t-test, SSIM:  p<0.005, PSNR & NMI: \np<0.001). Relative to simple reformats, the synthes ized image showed a 26% \nSSIM increase, 29% PSNR increase, and 10% NMI incre ase. Qualitatively, \npathological presentation in the synthesized image was similar to acquired \nsequences. \nConclusion: This study presents a deep-learning-based method to  synthesize \nhigh quality multiplanar reformations in knee MRI. Synthesized images \nexhibited strong quantitative similarity to acquire d images. \nLimitations: In future work, these methods should undergo qualit ative review \nby radiologists. \nFunding for this study: This study was funded by Subtle Medical. \nEthics committee - additional information: All data was retrospective and \nanonymized and therefore not considered human subje cts research. \nAuthor Disclosures:  \nSrivathsa Pasumarthi Venkata: Employee: Subtle Medi cal Inc. \nLanhong Yao: Employee: Subtle Medical \nLong Wang: Employee: Subtle Medical \nThomas Campbell Arnold: Employee: Subtle Medical \nAjit Shankaranarayanan: Employee: Subtle Medical \n \n \nA three-dimensional magnetic resonance imaging-base d scoring system \nto evaluate continuity, thickness and shape of the injured anterior \ncruciate ligament \n*N. Giannotti*¹, A. Liu², H. Gauffin², S. Filbay³, J. Kvist², M. Englund⁴; \n¹Sydney/AU, ²Linköping/SE, ³Melbourne/AU, ⁴Lund/SE \n(nicola.giannotti@sydney.edu.au) \n \nPurpose or Learning Objective: Anterior cruciate ligament (ACL) injuries are \ncommon, and ACL reconstruction (ACLR) is considered  vital for restoring knee \nstability. However, recent evidence evaluating non- surgical approaches to the \ninjured ACL challenges these notions. To date, no s tudies have used three-\ndimensional magnetic resonance imaging (3D-MRI) to systematically assess \nmorphological changes of the injured ACL at differe nt time-points in non-\nreconstructed patients. The aim of this study was t o develop and investigate \nthe reliability of a semi-quantitative scoring syst em to assess morphological \nchanges of the injured ACL using 3D-MRI. \nMethods or Background: Baseline and 4 follow-ups within 2 years 3T 3D-MRI \nscans were acquired with an isotropic proton-densit y fat-saturated sequence. \nScans were censored from further analysis after ACL R. 3D-MRI data were \nreconstructed on oblique coronal parallel to ACL, s agittal, and axial planes. \nUnblinded for time sequence, two readers independen tly scored the MRI data \nfrom all available time-points. The intra-rater rel iability was measured six \nmonths after the initial evaluation. Weighted kappa  (κw) was used to calculate \ninter- and intra-reader reliability. \nResults or Findings: 129 patients (46% female; 25 SD [7] years) with non -\nsurgically treated ACL injury from the NACOX study were included. The \nscoring system developed assessed the following ACL  3D-MRI features: \noverall structure, fiber continuity, thickness, and  shape. The injury location was \nscored at baseline. 2,430 gradings were collected i n total. The averaged inter-\nreader reliability (κw) was 0.738, 0.673, 0.710 and 0.539 for ligament s tructure, \nfiber continuity, thickness and shape, respectively . The intra-rater reliability \n(κw) ranged from 0.409 to 0.873. \nConclusion: We introduce a novel 3D-MRI-based semi-quantitative  scoring \nsystem to enable assessment of the acutely injured ACL and its morphological \nchanges over time with acceptable inter-reader reli ability. Future validation vs \nfindings from knee arthroscopy will be required. \nLimitations: No limitations were identified. \nFunding for this study: The NACOX study received funding from the Swedish \nResearch Council, Medical Faculty of Linköping Univ ersity, Swedish National \nCentre for Research in Sports, Medical Research Cou ncil of Southeast \nSweden (FORSS), ALF Region Östergötland \nEthics committee - additional information: Ethical approval for the study \nwas granted by the Swedish Ethical Review Authority , Dnr: 2016/44/31. \nAuthor Disclosures:  \nHåkan Gauffin: Nothing to disclose \nNicola Giannotti: Nothing to disclose \nJoanna Kvist: Nothing to disclose \nStephanie Filbay: Nothing to disclose  \nMartin Englund: Nothing to disclose \nAngie Liu: Nothing to disclose \n \n \nThe posterior cruciate ligament angle in the settin g of deficient anterior \ncruciate ligament deficient knees: the effect of ge nder, age, time from \ninjury and tibial slope \n*M. V. Bausano*¹, F. Di Maria², R. D'Ambrosi¹, L. M . Sconfienza¹, S. Fusco¹,  \nE. Abermann³, C. Fink³; ¹Milan/IT, ²Catania/IT, ³In nsbruck/AT \n(mariavittoriabausano@gmail.com) \n \nPurpose or Learning Objective: The aim was to assess the posterior cruciate \nligament (PCL) angle in anterior cruciate ligament deficient knees and correlate \nit with anatomic and demographic factors. \nMethods or Background: Patients included were initially noted to have an \nACL tear clinically as confirmed by MRI.For each pa tient were evaluated:PCL \nangle (PCLA), medial tibial slope (MTS), lateral ti bial slope (LTS), medial \nanterior tibial translation (MATT) and lateral ante rior tibial translation (LATT). \nAge, sex, and time from injury to MRI were manually  recorded.Age groups \nwere predefined dichotomizing age at its mean value  while for time interval \nfrom injury to MRI the cut-off of 90 days was selec ted to differentiate between \nchronic and acute lesions. Differences by groups we re assessed with t-test or \na Wilcoxon-Mann Whitney test, according to score di stribution. Spearman rank \ncorrelations were also estimated to explore correla tion among collected \nvariables. \nResults or Findings: A total of 193 patients were included in the study of \nwhich 91 females and 102 males with a mean age of 3 0.27±12.54.The mean \ntime from injury to MRI was 14.18±55.77 days.On ove rall population, mean \nPCLA resulted 128.72±10.33°, mean MTS 3.57±2.33, mean LTS 6.07±3.52, \nmean MATT and LATT were respectively 4.76±2.02 and 7.01 ±2.48 mm. In \n190 cases PCLA angle was≥to 105° and only in 3 was inferior.PCLA negatively \ncorrelated with medial and lateral anterior tibial translation(p<0.05).Female \nshowed a higher PCLA compared to male (130.55±10.23  vs 127.08±10.20; \np=0.019).Patients with chronic ACL injury showed a lower value of PCLA \ncompared to patients with acute injury(p=0.032) and  a superior grade of \nLATT(p=0.015). \nConclusion: In the setting of ACL lesions, PCLA has normal valu e in acute \ninjury and decrease over the time. PCLA is negative ly correlated with anterior \ntibial translation and female have higher PCLA comp ared to male. \nLimitations: First, the diagnosis was made exclusively by MRI an alysis, \nsecond the data excluded partial ACL injuries. \nFunding for this study: N/A \nEthics committee - additional information: After Institutional Review Board \napproval (ACL-L2104), two of the authors reviewed t he MR images of all \npatients with a clinical diagnosis of acute ACL inj ury. The present study was \nconducted following the Strengthening the Reporting  of Observational Studies \nin Epidemiology (STROBE) statement [3]. All procedu res were conducted in \naccordance with the standards highlighted in the 19 64 Helsinki Declaration and \nits later amendments. \nAuthor Disclosures:  \nElisabeth Abermann: Nothing to disclose \nFabrizio Di Maria: Nothing to disclose \nStefano Fusco: Nothing to disclose \nLuca Maria Sconfienza: Nothing to disclose \nRiccardo D'Ambrosi: Nothing to disclose \nMaria Vittoria Bausano: Nothing to disclose \nChristian Fink: Nothing to disclose \n \n \nDiminished Meniscal Height as an Indicator of Synov itis on Magnetic \nResonance Imaging of the Knee \nS. N. Yılmazer Zorlu, *N. A. Ahmady*, K. B. Karaca,  Z. Akkaya; Ankara/TR \n(nesarahmad.ahmadi123@gmail.com) \n \nPurpose or Learning Objective: To investigate the relationship between \nmeniscal height and adjacent synovitis on magnetic resonance images (MRI) \nof the knee, based on the hypothesis that, in contr ast to simple effusion, \nhypertrophic synovial tissue will cause compressive  distortion on neighboring \nmeniscus. \nMethods or Background: Contrast-enhanced knee MRIs from patients \n(≥18years), acquired during 2014-2024, at a single un iversity hospital were \nincluded in this cross-sectional, retrospective stu dy. Patients with history of \nknee surgery, acute trauma, local and systemic mali gnancy were excluded. On \npost-contrast fat-suppressed(fs)-T1-weighted images , 2 independent observers \nassessed the presence of synovitis (synovial enhanc ement ≥2mm) adjacent to \nthe medial meniscus body-posterior horn (MMB, MMPH)  and lateral meniscus \nbody-posterior horn (LMB, LMPH). Blinded to the syn ovitis status, meniscal \nheights were measured on coronal and sagittal fs-fl uid sensitive images where \nmenisci were intact and showed < grade 2 degenerati on. The relationship \nbetween meniscal height and synovitis was evaluated  using linear regression \nmodels adjusted for age, gender and body mass index  (BMI). Intra-class \ncorrelations (ICC) were used to test the reproducib ility of meniscal height \nmeasurements across 10 randomly selected patients. \nResults or Findings: A total of 129 patients (85 women, mean \nage=46.6±15years, mean BMI=24.6±5.1 kg/m2) (119 patients on 1.5T; 10 \npatients on 3.0T systems; ranges of slice thickness =3.5-6mm, gap=0.3-1mm, \n\n \n \nFriday \nAbstract-based Programme \n \n 184  \nFOV=12-16 cm) were included. Mean heights of MMB, M MPH, LMB, LMPH \nwere 5.5±1mm, 5.3±1.1mm, 5.9±0.9mm, 5.6+0.8mm, respectively. MMB, \nMMPH and LMPH were significantly thinner adjacent t o synovitis regions(β=-\n0.9, [95%CI= -1.34, -0.44], p<0.001; β=-1.26, [95%CI=-1.79, -0.73], p<0.001; \nβ=-0.7, [95%CI= -1.12, -0.29], p=0.001, respectively ). ICC ranged between \n0.96 - 0.99 (p<0.001), indicating excellent inter-r ater agreement. \nConclusion: Diminished meniscal height adjacent to high-signal areas on \nfluid-sensitive images may help distinguish effusio n from synovitis on clinical \nknee MRIs. \nLimitations: The BMI values were not available in 25 patients. \nFunding for this study: None \nEthics committee - additional information: Ethics Committee Approval \nNumber: I03-251-24 \nAuthor Disclosures:  \nKazım Burak Karaca: Nothing to disclose \nSezer Nil Yılmazer Zorlu: Nothing to disclose \nNesar Ahmad Ahmady: Nothing to disclose \nZehra Akkaya: Nothing to disclose \n \n \nThe precision of a novel method for automated CT-ba sed radio-\nstereometric analysis in evaluating tibial implant migration \n*M. Acke*¹, B. Keelson¹, L. H. W. Engseth², G. Van Gompel¹, F-D. Ohrn³,  \nJ. De Mey¹, A. Schulz², S. M. Röhrl², N. Buls¹; ¹Br ussels/BE, ²Oslo/NO, \n³Kristiansund/NO \n(manou.acke@gmail.com) \n \nPurpose or Learning Objective: The current gold-standard for implant \nmigration analysis is radiostereometric analysis (R SA), which is a time-\nconsuming and resource demanding method. We aim to evaluate the precision \nof a novel non-invasive method using automated CT-b ased radio-stereometric \nanalysis (CT-RSA) for tibial implant migration anal ysis in a porcine model. \nMethods or Background: A porcine knee cadaver with a tibial implant was \nexamined with marker-based RSA and computed tomogra phy (CT). RSA \nacquisitions (133kV, 6.3mAs) and CT images (120kV, 100mAs, 0.625mm, \n0.5s, FOV 200×200mm) were acquired in seven differe nt positions (P1,…,P7) \nof the cadaveric knee. To obtain enough statistical  power, the 7 positions were \ncompared to each other, resulting in 21 double exam inations \n(P1_P2,…,P6_P7) for each method. Post-processing of  the CT data was \nperformed by a novel in-house built automatic image  processing pipeline using \nSimpleITK. The reference standard was zero implant motion. Maximum-Total-\nPoint-Motion (MTPM) was calculated for RSA to ident ify the implant point with \nmaximum translation. Similarly, from the CT-RSA dat a we calculated Total-\nTranslation (TT) for 6 non-invasive, virtual landma rks on the implant, including \nthe centre-of-mass and 5 peripheral landmarks. In a ddition, we computed a \nvirtual implant mesh from the CT data, including it s displacement. For RSA and \nCT-RSA, displacement differences towards the refere nce-standard were \nassessed using a two-sample t-test. \nResults or Findings: The precision for MTPM using marker-based RSA was \n0.45mm (95%CI 0.19–0.70mm). TT calculation for virt ual landmarks using CT-\nRSA was more precise than RSA (p<0.001) with 0.15mm  (95%CI 0.12–\n0.18mm). The mesh data allowed to identify the poin t of maximum translation \nwith a displacement of 0.16mm (95%CI 0.13–0.19mm). \nConclusion: Compared to RSA, the novel automated CT-RSA shows \nimproved precision in analysing tibial implant migr ation for a porcine cadaver. \nLimitations: This is a porcine phantom-study, which limits gener alizability and \nresults may differ in humans. \nFunding for this study: Funding is provided by the UZBrussels, Belgium and \nCIRRO, Norway. \nEthics committee - additional information: The study was approved by the \nlocal research committee at Oslo University Hospita l on December 13, 2021. \nNo consent was necessary since exams were performed  on a porcine cadaver \nphantom. \nAuthor Disclosures:  \nStephan Maximilian Röhrl: Nothing to disclose \nManou Acke: Nothing to disclose \nJohan De Mey: Nothing to disclose \nLars Harald William Engseth: Nothing to disclose \nGert Van Gompel: Nothing to disclose \nAnselm Schulz: Nothing to disclose \nNico Buls: Nothing to disclose \nFrank-David Ohrn: Nothing to disclose \nBenyameen Keelson: Nothing to disclose \n \n \n \n \n \n \n \n \n \nHow Cone-Beam CT affects Clinical Management in Foo t Trauma \n*E. Mcdermott*, D. P. Moloney, S. Murphy, B. Gibney , P. J. Macmahon,  \nE. Kavanagh; Dublin/IE \n(edwardmcdermott@gmail.com) \n \nPurpose or Learning Objective: Foot trauma is a common presentation to the \nemergency department. In our institution Cone Beam Computed Tomography \n(CBCT) is routinely utilized in patients with suspe cted fractures with negative \nradiographs. Previous studies have investigated the  role of CBCT in the acute \ntrauma setting for wrist and ankle injuries, but no  large study has been \nperformed for foot trauma. Our study focuses on how  CBCT can improve the \ndiagnosis of foot fractures in a trauma setting and , with the input of orthopaedic \nsurgery, how it can alter the patient’s treatment p lan. \nMethods or Background: Patients with foot trauma who underwent CBCT \nbetween 2019 and 2023 were reviewed. In cases of di scordance between \nCBCT and X-rays, images were assessed with an ortho paedic surgeon. \nPatients were classified into three treatment optio ns based on X-rays and \nCBCT: No Treatment, Immobilisation, and Surgery. \nResults or Findings: 204 trauma patients had foot X-rays and CBCT. In 10 2 \ncases, CBCT identified 164 additional fractures, le ading to treatment changes \nin 62 cases (p < 0.01). 42 changed from ‘No Treatme nt’ to ‘Immobilisation,’ 3 \nfrom ‘No Treatment’ to Surgery,’ and 3 from ‘Immobi lisation’ to ‘Surgery.’ 12 \nwere downgraded from ‘Immobilisation’ to ‘No Treatm ent,’ and 2 from ‘Surgery’ \nto ‘Immobilisation’. \nConclusion: Our findings show that a large number of foot traum a goes un-\ndiagnosed on radiographs. The involvement of orthop aedic surgery in \nreviewing the radiographs and CBCT to assess how tr eatment plans may \nchange is a novel approach. We found that a signifi cant number of patients \nwould have their type of treatment altered with the  use of CBCT. Our findings \nshow the significant role CBCT can have when integr ated into trauma care. \nLimitations: Only a single centre was used for data collection. Reporting of \ninitial radiographs by non-musculoskeletal speciali sed radiologists. \nFunding for this study: None. \nEthics committee - additional information: Project approved by the Mater \nMisericordiae University Hospital's Clinical Audit and Effectiveness Committee. \nAuthor Disclosures:  \nPeter Joseph Macmahon: Nothing to disclose \nDarren Patrick Moloney: Nothing to disclose \nBrian Gibney: Nothing to disclose \nEdward Mcdermott: Nothing to disclose \nEoin Kavanagh: Nothing to disclose \nSophie Murphy: Nothing to disclose \n \n \n16:00-17:30 Research Stage 3 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1605 \nArtificial intelligence: real world results \nand large European projects \n \nModerator \nL. Martí-Bonmatí; Valencia/ES  \n(marti_lui@gva.es) \n \n \nAssessing the effectiveness of artificial intellige nce (AI) in prioritising CT \nHead interpretation: a stepped-wedge cluster-random ised trial (ACCEPT-\nAI) \n*K. Nash*¹, K. Vimalesvaran², R. Dharmadhikari³, M.  Hall⁴, A. Novak¹,  \nS. Ather¹, D. J. Lowe⁴, H. Shuaib², Accept-Ai Investigators²; ¹Oxford/UK,  \n²London/UK, ³Northumberland/UK, ⁴Glasgow/UK \n(katrinanash649@outlook.com) \n \nPurpose or Learning Objective: To evaluate the effectiveness of artificial \nintelligence (AI) in prioritisation of non-contrast  head computed tomography \nscan (NCCT). We evaluated; 1) whether there was a r eduction in report \nturnaround time (TAT) of prioritised NCCT, 2) the a ccuracy of the AI algorithm, \nand 3) the technical performance of the algorithm. \nMethods or Background: This large-scale multi-centre trial has been \nconducted across three emergency departments betwee n November 2023 to \nJuly 2024. Individuals above the age of 18 who pres ented to the emergency \ndepartment and underwent NCCT were included. Data c ollected included \ndemographic data, time from acquisition to report o f CT scan, referral, and \ndischarge, and death within 28 days. Findings were categorised into prioritised \n\n \n \nFriday \nAbstract-based Programme \n \n 185  \nfindings (intracranial haemorrhage, mass effect, fr acture), non-prioritised \nfindings (atrophy and infarct), and no prioritised findings. The study was \nconducted in 3 stages; pre-implementation, implemen tation and post-\nimplementation. Baseline data was collected during the pre-implementation \nphase, whilst the implementation phase enabled trai ning and integration of the \nAI tool into radiology workflow at each trust. Duri ng the post-implementation \nphase, AI results were visible to radiologists. Rad iology reports were coded to \nassess AI accuracy, with discrepant cases sent for ground truthing by two \nindependent radiologists. \nResults or Findings: 7,500 scans from the pre-implementation phase, and \n8,453 scans from the post-implementation phase have  been included. The \nmedian TAT (minutes) for prioritised scans in the p re-implementation was 34 \n(inter-quartile range 20-55) and 35 in the post-imp lementation phase (22-59). \nConclusion: We have successfully conducted a prospective trial of AI \nimplementation across multiple centres in the Unite d Kingdom. Initial results do \nnot show a difference in TAT for NCCT reports after  implementation of AI. \nLimitations: TAT has been calculated during preliminary analysis  of a larger \ndataset, therefore full results will be available f or presentation. \nFunding for this study: This study has been awarded funding from NHSx AI \nAward (Award reference: AI_Award02354). \nEthics committee - additional information: The study protocol was approved \nby the Research Ethics Committee (REC) of East Midl ands (Leicester Central), \nin May 2023 (REC 23/EM/0108) and was conducted in a ccordance with the \nprinciples of Good Clinical Practice. \nAuthor Disclosures:  \nHaris Shuaib: Nothing to disclose \nRahul Dharmadhikari: Nothing to disclose \nAccept-Ai Investigators: Nothing to disclose \nKatrina Nash: Nothing to disclose \nAlex Novak: Nothing to disclose \nSarim Ather: Shareholder: Share holder RAIQC Ltd \nKavitha Vimalesvaran: Nothing to disclose \nMark Hall: Nothing to disclose \nDavid J. Lowe: Nothing to disclose \n \n \nONCOPILOT: A Promptable CT Foundation Model For Sol id Tumor \nEvaluation \n*L. Machado*¹, H. Philippe², E. Ferreres², J. Khlau t², J. Gregory¹, M. Ronot¹,  \nD. Tordjman², P. Manceron², P. Herent²; ¹Clichy/FR,  ²Paris/FR \n(dr.machado.leo@gmail.com) \n \nPurpose or Learning Objective: Carcinogenesis leads to tumors with diverse \nshapes and behaviors. Although RECIST 1.1 remains t he standard for \nevaluation, its reliance on linear measurements and  high inter-reader variability \noften result in misclassification. Volumetric bioma rkers, such as total tumor \nburden, provide more information but need automated  segmentation. \nTraditional segmentation models struggle with compl ex lesions, have a narrow \nfocus, and lack interactivity. Foundation models wi th transformer architecture \naddress these issues through zero-shot learning and  visual prompts (point-\nclick, bounding box). We developed ONCOPILOT, a fou ndation model that \nenhances RECIST 1.1 measurements, enabling volumetr ic analysis while \nintegrating seamlessly into radiology workflows. \nMethods or Background: ONCOPILOT was trained on 7,500 CT scans, \nincluding normal anatomy and oncological cases. Its  segmentation \nperformance was assessed against nnUnet, a state-of -the-art baseline, using \nthe DICE coefficient. The evaluation also included comparisons with \nradiologists for RECIST 1.1 long-axis measurements,  annotation speed, and \ninter-reader variability. \nResults or Findings: ONCOPILOT outperformed state-of-the-art models, \nachieving a mean DICE of 0.78 post-editing versus 0 .70 for the baseline. Its \nRECIST error (7.4%, 1.1 mm) was not significantly d ifferent to radiologists' \n(8.6%, 1.3 mm). ONCOPILOT-assisted measurements wer e quicker than \nmanual (17.2 vs. 20.6 seconds, p < .05), with reduc ed inter-reader variability \n(1.7 vs. 2.4 mm, p < .05). \nConclusion: ONCOPILOT is among the first foundation model appli cations in \nradiology, acting as an interactive AI assistant fo r oncological evaluation. It \nimproves RECIST reproducibility, facilitates access  to volumetric biomarkers, \nand seamlessly integrates into radiology workflows,  reducing inter-reader \nvariability and measurement time. This approach off ers substantial potential to \nadvance oncology research and enhance clinical care . \nLimitations: ONCOPILOT showed reduced performance on small tumor s, \nparticularly lung lesions, which were overrepresent ed in the test set. Future \niterations should use a more balanced dataset and c over a broader range of \ntumor types. \nFunding for this study: This work was granted access to the HPC resources \nof IDRIS under the allocation 2024-AD011013489R2 ma de by GENCI. \nEthics committee - additional information: The data used is anonymous and \npublically available \n \n \n \nAuthor Disclosures:  \nJulien Khlaut: Employee: Raidium \nJules Gregory: Nothing to disclose \nPierre Manceron: Founder: Raidium \nMaxime Ronot: Nothing to disclose \nLéo Machado: Employee: Raidium \nPaul Herent: Founder: Raidium \nElodie Ferreres: Employee: Raidium \nHélène Philippe: Employee: Raidium \nDaniel Tordjman: Employee: Raidium \n \n \nEvaluating artificial intelligence for lung cancer detection on chest \nradiographs: multi-vendor comparison of diagnostic accuracy in a real-\nworld UK population \n*A. Maiter*, P. Metherall, J. Taylor, S. Matthews, K. Hocking, E. Burton,  \nE. Anderson, A. Swift, C. S. Johns; Sheffield/UK \n \nPurpose or Learning Objective: Automated detection of lung cancer on chest \nradiographs by AI could streamline diagnostic pathw ays. This study evaluated \nthe diagnostic accuracy of commercially available s oftware from seven vendors \nusing a large dataset of radiographs from a real-wo rld UK population. \nMethods or Background: Consecutive chest radiographs obtained at our \ntertiary UK centre were retrospectively identified.  Chest radiographs requested \nfrom primary care for adult patients, regardless of  indication, were eligible for \ninclusion. Software from the seven vendors evaluate d each radiograph \nindependently. The radiologist report for each radi ograph was also \ninterrogated. Diagnostic accuracy metrics were dete rmined by comparing the \nsoftware outputs and radiologist reports against th e diagnosis of lung cancer \nby multidisciplinary team decision. \nResults or Findings: 5,722 chest radiographs were included from 5,592 \npatients (median age 59 years, 54% female, 79% whit e, 1.6% prevalence of \nlung cancer). The software yielded the following (m ean±SD): sensitivity 46±9%, \nspecificity 95±4%, positive predictive value (PPV) 15±5%, negative predictive \nvalue (NPV) 99±0%, accuracy 70±4% and false positive per image rate (FPPI) \n0.05±0.04. Radiologist reports yielded the following: sensitivity 66%, specificity \n98%, PPV 36%, NPV 99%, accuracy 82% and FPPI 0.02. \nConclusion: All software demonstrated high specificity and NPV comparable \nto radiologists, and could add value by helping to exclude lung cancer. \nHowever, all software also showed lower sensitivity  and PPV than radiologists, \nsuggesting that they could increase the number of f alse negative and false \npositive results. While AI has potential to improve  the efficiency of “straight to \nCT” and other diagnostic pathways for suspected lun g cancer, the risks of \ndiagnostic errors require careful consideration. \nLimitations: This study used a retrospective dataset from a sing le centre. \nFurther testing of software performance in multi-ce ntre patient cohorts and \nevaluation of downstream impacts are essential prio r to clinical deployment. \nFunding for this study: This study was funded by the NHS South Yorkshire \nIntegrated Care System. \nEthics committee - additional information: This study received local \nresearch ethics committee approval (23/EM/0186). Th e need for dedicated \npatient consent was waived. \nAuthor Disclosures:  \nSuzanne Matthews: Nothing to disclose \nJonathan Taylor: Nothing to disclose \nKatie Hocking: Nothing to disclose \nEmily Burton: Nothing to disclose \nAndrew Swift: Nothing to disclose \nAhmed Maiter: Other: Bayer \nChris S Johns: Nothing to disclose \nElizabeth Anderson: Nothing to disclose \nPeter Metherall: Nothing to disclose \n \n \nEvaluation of AI-assisted Diagnosis of Clinically S ignificant Prostate \nCancer on MRI at Scale: Preliminary Findings from t he PI-CAI Consortium \n*J. J. Twilt*¹, A. Saha¹, J. S. Bosma¹, D. Yakar², M. Elschot³, J. Veltman⁴,  \nJ. Fütterer¹, H. E. Huisman¹, M. De Rooij¹; ¹Nijmeg en/NL, ²Groningen/NL, \n³Trondheim/NO, ⁴Almelo/NL \n(jasper.twilt@radboudumc.nl) \n \nPurpose or Learning Objective: To assess whether utilizing an internationally \nvalidated prostate artificial intelligence (AI) system enhances the accuracy of \nprostate MRI evaluations in diagnosing clinically s ignificant prostate cancer \n(csPCa; Gleason Grade ≥2), compared to non-assisted assessments, in a \ncomprehensive international reader study. \nMethods or Background: In this retrospective study, imaging and a prostate  \nAI system developed and benchmarked on 10,207 exami nations through an \ninternational confirmatory study (PI-CAI) were used . A total of 780 biparametric \nprostate MRI examinations (2015-2021) from men susp ected of csPCa, with \ndiagnostic-sufficient image quality and no prior cs PCa findings or treatment, \nwere included. Reference was established through hi stopathology and ≥3 \n\n \n \nFriday \nAbstract-based Programme \n \n 186  \nyears of follow-up. The AI was calibrated to genera te patient-level csPCa \nsuspicion scores (0-10) and associated lesion-detec tion maps using 420 \nexaminations from three Dutch centers. The remainin g 360 examinations \n(three Dutch, one Norwegian center) were used for o utcome analysis. Sixty-\none readers (53 centers, 17 countries) provided PI- RADS 3-5 annotations and \npatient-level csPCa suspicion scores (0-100) with a nd without AI assistance in \ntwo phases, separated by a 4-week washout period. M ulti-reader, multi-case \nanalysis compared diagnostic outcomes at a per-pati ent level. \nResults or Findings: Preliminary analysis of 22 readers (1-13 years of \nprostate MRI experience) demonstrated that AI assis tance improved csPCa \ndiagnosis, with AUROCs of 0.919 (95% CI: 0.896-0.94 2) compared to 0.870 \n(95% CI: 0.831-0.908) without assistance. Sensitivi ty and specificity at PI-\nRADS ≥3 improved from 94.1% to 96.5% and 46.6% to 49.1%, respectively. \nConclusion: AI assistance improves csPCa diagnosis and holds pr omise for \nimproving clinical outcomes. Further research is ne eded to confirm \ngeneralizability and assess workflow efficiency. \nLimitations: Preliminary findings limit the sample size and subg roup analysis \nconsidering reader expertise. Reuse of data may int roduce generalization bias, \nand AI was not used to guide histologic verificatio n. \nFunding for this study: EU Horizon 2020: ProCAncer-I (grant number \n952159), Health~Holland (grant number LSHM20103). \nEthics committee - additional information: Retrospective use of anonymous \npatient data was approved by institutional or regio nal review boards at each \ncontributing center (identifiers: REK 2017/576; CMO  2016-3045; IRB 2018-\n597; ZGT23-37), and was conducted in accordance wit h the principles of the \nDeclaration of Helsinki. Informed consent was waive d. \nAuthor Disclosures:  \nJurgen Fütterer: Grant Recipient: Siemens Healthine ers \nJoeran Sander Bosma: Nothing to disclose \nHenkjan En Huisman: Grant Recipient: Siemens Health ineers Grant Recipient: \nCanon \nAnindo Saha: Speaker: Guerbet \nMatthijs Elschot: Nothing to disclose \nDerya Yakar: Nothing to disclose \nJeroen Veltman: Nothing to disclose \nMaarten De Rooij: Nothing to disclose \nJasper Jonathan Twilt: Nothing to disclose \n \n \nThe PANORAMA study results: Pancreatic Cancer Diagn osis - \nRadiologists meet AI \n*M. Schuurmans*, N. Alves, P. Vendittelli, G. Litje ns, J. J. Hermans,  \nH. E. Huisman; Nijmegen/NL \n(megan.schuurmans@radboudumc.nl) \n \nPurpose or Learning Objective: The PANORAMA study transparently \nevaluates radiologists and AI in detecting pancreat ic ductal adenocarcinoma \n(PDAC) using contrast-enhanced computed tomography (CECT). \nMethods or Background: This retrospective study includes 3338 abdominal \nCECTs of patients without prior history of treatmen t or positive histopathology \nfindings of PDAC acquired between 2006 and 2021 fro m 5 centers \n(Netherlands, Norway, and Sweden). Of these, 2238 c ases (676 PDAC) are \npublicly available to develop and train AI algorith ms, and 100 and 1000 cases \nwere sequestered for AI tuning and testing, respect ively. The test set \ncomprises data from two external centres not presen t in the other cohorts. A \nsubset of 400 testing cases is used for the PANORAM A reader study. Both AI \nand radiologists indicate PDAC likelihood and local ization of lesions. \nAdditionally, radiologists provide a newly introduc ed PANC-RADS score to \nassess the urgency of expert referrals. AI is openl y developed and evaluated \nusing common metrics through the Grand Challenge pl atform. Patient-level \nperformance and lesion localization performance are  assessed using the area \nunder the receiver operating characteristic curve ( AUROC) and average \nprecision (AP), respectively. \nResults or Findings: The PANORAMA reader study results will be presented  \nfor the first time at ECR 2025. Currently, 57 radio logists (39 centers,13 \ncountries, 2-30 years of experience, median: 9 year s) participate in the study. \nThe baseline AI algorithm (nnU-Net with cross-entro py loss) achieves 0.9776 \nAUROC and 0.7226 AP in the tuning set. There are cu rrently 165 registered AI \nchallenge participants across 11 teams. \nConclusion: Transparently benchmarked AI can enable early PDAC detection \nat the expert-radiologist level. \nLimitations: While histopathology and follow-up were considered as the \nreference standard for the sequestered testing set,  we could not guarantee this \nlevel of evidence for all cases in the public train ing set. \nFunding for this study: This project has received funding from the European  \nUnion’s Horizon 2020 research and innovation progra m under grant agreement \nNo 101016851, project PANCAIM. \n \n \n \n \nEthics committee - additional information: Writen ethics committee approval \nwas obtained from every center participating in the  study. \nAuthor Disclosures:  \nNatalia Alves: Nothing to disclose \nHenkjan En Huisman: Nothing to disclose \nMegan Schuurmans: Nothing to disclose \nPierpaolo Vendittelli: Nothing to disclose \nJohn J Hermans: Nothing to disclose \nGeert Litjens: Nothing to disclose \n \n \nDevelopment and Implementation of a Web-Based Platf orm for Clinical \nValidation of Artificial Intelligence Models in Fiv e Cancer Types: An \nInitiative of the European CHAIMELEON Project \n*A. Galiana-Bordera*, J. Aquerreta-Escribano, P. M.  Martínez Gironés,  \nP. Lozano, G. Ribas, P. Jimenez, L. Cerda Alberich,  I. Blanquer,  \nL. Marti-Bonmati; Valencia/ES \n(adrian_galiana@iislafe.es) \n \nPurpose or Learning Objective: This study evaluates AI-based predictive \noncology models for prostate, lung, breast, and col orectal cancers through the \ncreation of an innovative web-platform. The goal is  to bridge the gap between \nAI development and clinical implementation, advanci ng AI adoption in \nhealthcare as part of the European CHAIMELEON proje ct. \nMethods or Background: We developed a web-platform with a microservices \narchitecture, including a REST API, ORTHANC PACS fo r image management, \nand Keycloak security. The frontend is a custom-bui lt application serving as a \ncontrol panel for users. It allows clinicians to op en a patient-specific DICOM \nviewer alongside clinical information, AI predictio ns, and validation buttons. \nThis interface facilitates a three-stage case revie w: standard evaluation, AI-\naided assessment, and final comparison with ground truth. AI models, \ndeveloped through an open challenge using anonymize d data, were integrated \ninto the platform. \nResults or Findings: The platform demonstrates efficiency, customizabili ty, \nand scalability, enhancing oncological study reprod ucibility. It creates an \nenvironment where medical images, clinical data, an d AI predictions coexist, \nadapting to various clinical and research settings.  Over 70 clinicians are using \nthe system to evaluate more than 2000 patients, wit h 60% reviewed in less \nthan a month. This rapid adoption highlights the pl atform's user-friendly design \nand potential for improving clinical workflow. \nConclusion: This platform represents a significant advancement in validating \nand adopting AI models in oncology. It provides a f oundation for integrating AI \ninto clinical environments, benefiting both clinici ans and patients. The study \nhighlights the importance of user-friendly interfac es and structured evaluation \nprocesses in bridging the gap between AI developmen t and clinical application. \nLimitations: The study's retrospective nature limits long-term o utcome \nassessment. Further research is needed on implement ation across diverse \nhospital infrastructures, the learning curve for clinicians, and ethical and \nregulatory aspects of AI integration in clinical pr actice. \nFunding for this study: This particular study did not receive direct fundin g. \nHowever, it was developed as part of the CHAIMELEON  project, which \nreceived funding from the European Union's Horizon 2020 research and \ninnovation program under Grant Agreement No. 952172 . \nEthics committee - additional information: Under the CHAIMELEON \nproject, all necessary ethical committee approvals have been obtained for the \nuse of medical images and the development of artifi cial intelligence models. \nThese approvals cover the collection, anonymization , and utilization of patient \ndata for research purposes. Furthermore, the web pl atform developed for \nevaluating these AI models has been registered and approved for use in \nclinical settings. This ensures compliance with dat a protection regulations and \nethical standards in medical research. The project adheres to strict protocols \nfor data handling and user access, maintaining pati ent privacy and data \nintegrity throughout the study. \nAuthor Disclosures:  \nPau Lozano: Nothing to disclose \nIgnacio Blanquer: Nothing to disclose \nLuis Marti-Bonmati: Nothing to disclose \nPaula Jimenez: Nothing to disclose \nJavier Aquerreta-Escribano: Nothing to disclose \nGloria Ribas: Nothing to disclose \nPedro Miguel Martínez Gironés: Nothing to disclose \nLeonor Cerda Alberich: Nothing to disclose \nAdrian Galiana-Bordera: Nothing to disclose \n \n \n \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 187  \nHospital patients' attitudes towards AI worldwide: Results from the \nCOMFORT study in 74 hospitals and 43 countries \n*F. Busch*¹, L. Hoffmann², L. Xu², L. Zhang³, L. Sa ba⁴, M. R. Makowski¹,  \nH. Aerts⁵, L. C. Adams¹, K. Bressem¹; ¹Munich/DE, ²Berlin/DE , ³Nanjing/CN, \n⁴Cagliari/IT, ⁵Boston, MA/US \n(felix.busch@tum.de) \n \nPurpose or Learning Objective: Too often, we see healthcare technology \nimplementations that focus only on the clinician's point of view and undervalue \nthe patient's perspective. Given the exponential ri se in artificial intelligence (AI) \napplications in healthcare, this international, mul ticentre, cross-sectional study \naimed to assess hospital patients' attitudes toward s AI in healthcare worldwide. \nMethods or Background: The present COMFORT study, involving 74 network \nhospitals in 43 countries, employed a quantitative 26-item instrument available \nin 26 languages on-site between February and Novemb er 2023. \nResults or Findings: 13806 patients from Europe (41.7%, n=5764/13806), \nAsia (25.2%, n=3473/13806), North America (16.5%, n =2284/13806), South \nAmerica (9.7%, n=1336/13806), Africa (5.3%, n=728/1 3806) and Oceania \n(1.6%, n=221/13806) were included. Overall, 57.6% o f respondents were \npositive about the use of AI in healthcare. Signifi cant differences in attitudes \nwere observed based on demographic characteristics,  health status and \ntechnological literacy. Female participants and tho se in poorer health had less \npositive attitudes towards the use of AI in medicin e. Conversely, higher levels \nof AI knowledge and frequent use of technological d evices were associated \nwith more positive attitudes. Notably, less than ha lf of the participants \nexpressed positive attitudes to all items related t o trust in AI, with the lowest \nlevel of trust being in the accuracy of AI in provi ding information about \ntreatment response. Patients showed a strong prefer ence for explainable AI \nand clinician-led decision-making, even if this mea nt a slight compromise in \naccuracy. \nConclusion: This large-scale, multinational study provides a co mprehensive \nperspective on patient attitudes towards AI in heal thcare across six continents. \nThe findings suggest the need for tailored AI imple mentation strategies that \naccount for patient demographics, health status, an d preferences for \nexplainable AI and physician oversight. All study d ata has been made publicly \navailable to encourage replication and further rese arch. \nLimitations: Non-probability sampling \nFunding for this study: This research is funded by the European Union \n(101079894). \nEthics committee - additional information: Ethical approval was obtained \nfrom Charité – Universitätsmedizin Berlin (EA4/213/ 22), which served as the \nlead institution. \nAuthor Disclosures:  \nLena Hoffmann: Nothing to disclose \nMarcus R. Makowski: Nothing to disclose \nLisa C. Adams: Nothing to disclose \nHugo Aerts: Nothing to disclose \nLina Xu: Nothing to disclose \nLongjiang Zhang: Nothing to disclose \nFelix Busch: Nothing to disclose \nKeno Bressem: Nothing to disclose \nLuca Saba: Nothing to disclose \n \n \nImplementing Artificial Intelligence with a Multi-A I Platform across 15 \ncentres: Experience and Strategy from an Internatio nal Healthcare \nOrganization \nD. Penha¹, A. Juhos¹, D. Tálos¹, L. Rosa², M. Santo s², E. Dias², R. Paroczai¹, \nR. Barone¹, *A. Roncacci*¹; ¹Amsterdam/NL, ²Lisboa/ PT \n \nPurpose or Learning Objective: Implementing artificial intelligence (AI) in \nradiology is challenging due to the variety of AI t ools and the complexity of IT \ninfrastructure and workflows. This presentation det ails a qualitative case study \nexamining the implementation of six AI solutions us ing a multi-AI platform \nacross 15 centers. \nMethods or Background: A longitudinal qualitative case study was conducted  \nin Portugal, in 15 radiology centers over two years  (May to October 2024), \nfocusing on the implementation of different AI tool s (Veye Lung Nodules, \nIcobrain DM, Transpara, Keros/Polaris, qXR, and ARV A) using a multi-AI \nplatform (Incepto). Data collected included 833 day s of work observations, 86 \nmeeting observations, and from Incepto dashboard. \nResults or Findings: The multi-country healthcare organization's AI team  \nmanaged the process adhering to a standard operatin g procedure covering \ninitiation, planning, implementation, and clinical/ operational phases. AI \ndeployment started with Icobrain DM (11 centers) an d Veye Lung Nodules (12 \ncenters), processing 1,699 exams and 49,278 respect ively. Transpara was \ndeployed across three centers, resulting in 5,791 s tudies over 19 months. \nKeros/Polaris in two centers, with 706 studies over  18 months. ARVA and qXR \nwere not implemented due to clinical decision. Over all, the AI implementation \nwas successful, with over 56,000 studies processed by four tools across 15 \ncenters. The study identified advantages of the mul ti-AI platform, including \nworkflow, cost-effectiveness, and improved patient care. However, \ndisadvantages such as complexity, integration chall enges, and potential \nvendor lock-in were also noted. \nConclusion: This case study provides valuable insights for heal thcare \norganizations considering AI implementation via mul ti-AI platform over stand-\nalone AI tools. \nLimitations: The implementation of a multi-AI platform in Radiol ogy lacks a \nclear blueprint to follow, creating the need to def ine new procedures and \nmetrics. These procedures suited our needs of but m ay not apply to other \nradiology institutions. \nFunding for this study: Not applicable \nEthics committee - additional information: Ethics committe and data/ legal \ndepartment involved and with full approval of the w hole project \nAuthor Disclosures:  \nDaniel Tálos: Nothing to disclose \nAlessandro Roncacci: Nothing to disclose \nLuis Rosa: Nothing to disclose  \nArpad Juhos: Nothing to disclose \nRobert Paroczai: Nothing to disclose \nDiana Penha: Nothing to disclose \nRocco Barone: Nothing to disclose \nMiguel Santos: Nothing to disclose \nEduardo Dias: Nothing to disclose \n \n \nFrom theory to practice: Re-identification Challeng e to test imaging data \nanonymization effectiveness \n*R. Catalán Flores*¹, I. Gómez-Rico¹, P. Jimenez¹, J. Gomes Carvalho²,  \nS. Mazzetti³, R. Martínez Martínez¹, M. França², D.  Regge³, L. Marti-Bonmati¹; \n¹Valencia/ES, ²Porto/PT, ³Torino/IT \n(rocio_catalan@iislafe.es) \n \nPurpose or Learning Objective: DICOM image de-identification is an \neffective measure to protect patient privacy and en sure compliance with the \nGDPR. However, no standardized methods guarantee th e irreversible \nanonymization of DICOM images or provide evidence o n the robustness of \nthese procedures. Organizations such as NEMA propos e de-identification \nprofiles for DICOM metadata, but the risk to data p rotection is assumed by the \nentity responsible for de-identification, as the da ngers of using these profiles \ncannot be accurately measured. Given this context, the Re-identification \nChallenge serves as a technical audit to assess the  robustness of DICOM \nimage de-identification methods. The objectives of this project are to validate \nthe robustness of these de-identification methods a nd to gain insight into \norganizing a challenge within the context of a Euro pean project. \nMethods or Background: The Re-identification Challenge consisted of a \nsingle phase in which the selected 68 participants were tasked to re-identify 38 \npseudonymized DICOM studies from multiple European hospitals (including \nSpain, Portugal and Italy), modalities and anatomic al regions. They were \npseudonymized locally using the de-identification p rofiles of ChAImeleon and \nProCancer-I European AI4HI projects. \nResults or Findings: Despite 74% of the participants delivered result, n one \nsucceeded in re-identifying the studies. Based on p articipants' reports of their \nattempts, vulnerabilities were discovered that allo wed them to narrow the \npopulation by obtaining their geographical region. \nConclusion: This challenge is a groundbreaking initiative that may pave the \nway for robust evaluations of de-identification met hods and may culminate in \nthe standardization of de-identification profiles i n the field of radiological \nimaging. \nLimitations: This challenge involved a limited sample of DICOM s tudies to \nensure GDPR compliance and protect patients’ privac y. Participant selection \nwas controlled through specific projects to prevent  unauthorized access and \nreduce the risk of data breaches, marking a pioneer ing effort in de-\nidentification research. \nFunding for this study: This challenge is part of the ChAImeleon project, \nspecifically of Work Package 10, titled 'Repository  Sustainability'. ChAImeleon \nhas received funding from the European Union's Hori zon 2020 research and \ninnovation programme under grant agreement No 95217 2. \nEthics committee - additional information: Ethical considerations have been \nparamount throughout the Challenge's development, g arnering approval from \nthe Ethics Committee of the involved data providers  institutions. Rigorous \nmeasures ensure privacy, security, and data legitim acy: patient consent forms \nwere obtained for the medical studies; legal expert s conducted a \ncomprehensive data protection impact assessment; an d the platform hosting \nthe studies adheres to stringent security protocols  and privacy policies. \nAuthor Disclosures:  \nJoão Gomes Carvalho: Nothing to disclose \nLuis Marti-Bonmati: Nothing to disclose \nRicard Martínez Martínez: Nothing to disclose \nDaniele Regge: Nothing to disclose \nRocío Catalán Flores: Nothing to disclose \nPaula Jimenez: Nothing to disclose \nManuela França: Nothing to disclose \nSimone Mazzetti: Nothing to disclose \nIgnacio Gómez-Rico: Nothing to disclose \n\n \n \nFriday \nAbstract-based Programme \n \n 188  \nEstablishing an applied framework to establish the trustworthiness of an \ninternational secure data environment for AI in CT imaging (AICT \nConsortium) \n*J. Kellas*¹, S. Van Wortswinkel², E. Casany Pujol³ , R. Lee¹, E. R. Ranschaert⁴; \n¹Oxford/UK, ²Borgerhout/BE, ³Barcelona/ES, ⁴Ghent/BE \n(john.kellas@nds.ox.ac.uk) \n \nPurpose or Learning Objective: The AICT consortium, funded by Horizon \nEurope (NetZeroAICT), consists of international cli nical sites across 3 \ncontinents and proposes a comprehensive trustworthi ness framework by \nsystematically integrating ethical, legal, sustaina bility and stakeholder \nengagement elements throughout research and develop ment. The goal is to \nensure acceptability and trust in our research, inn ovation pipeline and AI-driven \nradiology applications. \nMethods or Background: The NetZero AICT trustworthiness framework is \nbased on the European Commission's Trustworthy AI m odel, including \nfoundational elements—Lawfulness, Ethics, and Robus tness—structured with \nkey pillars: Human Agency, Technical Robustness, Pr ivacy, Transparency, \nDiversity, Societal Wellbeing, and Accountability. The project adheres to GDPR \ncompliance, local applicable privacy laws, and the EU AI Act, and integrates \nethics and privacy by design, broad and deep public  involvement, \nsustainability, innovation management, clinical val idation, and regulatory \ncompliance. \nResults or Findings: This is an interim report (year 1 of Horizon Europe  \nprogram). Here, we showcase the ‘ethics by design’ approach and an applied \nmodel of patient and public patient involvement and  engagement. A public \nadvisory group (PAG) has been formed with current m embership from 4 \ncountries and diverse backgrounds) with member repr esentation on project \nleadership groups (adopting and adapting a tiered m odel implemented by \nlarge-scale UK health data infrastructure projects (including OpenSAFELY). \nFeedback from the Public Advisory Group and the pro ject team strongly \nsupported the view that integrating ethics by desig n from the concept stage \nthrough to deployment has improved public confidenc e and ensured \ncompliance with ethical and legal standards. \nConclusion: By incorporating ethics by design, legal compliance , sustainability \nconsiderations and stakeholder engagement, the (Net Zero) AICT consortium \nestablished a reliable framework for radiology AI, promoting fairness and \ntrustworthiness among users and stakeholders. \nLimitations: This framework applies to healthcare imaging AI. \nFunding for this study: Horizon Europe and UK Research Innovation \nEthics committee - additional information: HRA number 22/HRA/2302 \nAuthor Disclosures:  \nRegent Lee: Shareholder: AI Sentia \nJohn Kellas: Consultant: This Equals \nSteven Van Wortswinkel: Employee: Ziekenhuis aan de  Stroom \nErik R. Ranschaert: Nothing to disclose \nErnest Casany Pujol: Employee: CMRAD \n \n \nTime impact of AI-Assisted knee MRI reading in a re al-world multi-center \nstudy within a radiology network \n*B. Rizk*¹, P. Cordelle², B. Dufour¹, N. Heracleous ¹, C. Thouly¹, P. Zille²,  \nF. Zanca³; ¹Sion/CH, ²Poitiers/FR, ³Leuven/BE \n(benrizk@gmx.net) \n \nPurpose or Learning Objective: We explored the impact on reading time of \nKEROS, a knee-MRI multifaceted AI algorithm, across  three distinct \ninterpretation workflows. \nMethods or Background: Clinical routine data was gathered from ten differe nt \ncenters during the daily workflow of eight radiolog ists, including four \nmusculoskeletal subspecialists (MSKs) and four gene ral knowledge \nradiologists (GENs), two of whom were junior (0 to 1 year of experience in \nprivate practice). We use a standardized report rel ying on voice recognition \nwithout secretariat formatting assistance. Data col lection was performed in \nthree phases: Phase 1, generated reports without AI  assistance; Phase 2, \nKEROS AI diagnosis was available before their inter pretation; Phase 3, AI \nfindings were pre-filling and auto-integrated into the structured report. \nReporting time was measured from report opening to validation, using the \nradiology information system (RIS). The Kruskal-Wal lis test (p<0.01) assessed \nsignificant differences, and time differences were calculated using weighted \nmeans. \nResults or Findings: In Phase 1, 431 exams were read with an average \nreading time of 10.8±9.5 to 26.3±9.3 minutes. In Phase 2 had 429 exams, with \nreading times between 12.8±8.2 and 27.5±9.2 minutes. Finally, Phase 3 \nincluded 425 exams with times ranging from 9.2±6.4 to 21.6±5.6 minutes. \nCases were nearly equally read by MSKs and GENs. Co mpared to Phase 1, \nPhase 3 showed an average time reduction of 2.1 min utes (p<0.01) (13%), \nprimarily driven by GENs, who saved up to 3.1 minut es (17%) (p<0.01). \n \n \n \nConclusion: We observe the average 13.4% (p<0.01) time reductio n after \nimplementing KEROS and pre-filled reporting (Phase 3 vs. Phase 1). \nGeneralists, the primary users of AI-assisted knee MRI readings, see an \naverage 17.4% (p<0.01) decrease in their reading ti me. In clinical practice, AI-\nassisted knee MRI reporting saves time for general radiologists, who benefit \nthe most from AI guidance. \nLimitations: No limitations \nFunding for this study: No funding \nEthics committee - additional information: We acquired the green light by \nthe ethical committee in Switzerland for our resear ch structure including data \nregistry and patient general consent for research. For the specific study \npresented here, only timing of radiologist reportin g data was used and not \npatient data. \nAuthor Disclosures:  \nPascal Zille: Nothing to disclose \nNatalie Heracleous: Nothing to disclose \nCyril Thouly: Nothing to disclose \nPhilippine Cordelle: Nothing to disclose \nBenoît Dufour: Nothing to disclose \nBenoît Rizk: Nothing to disclose \nFederica Zanca: Nothing to disclose \n \n \nPARROT: A Collaborative Polyglottal Annotated Radio logy Reports \nDatabase for Open Testing of Large Language Models \n*B. Le Guellec*¹, K. Bressem²; ¹Lille/FR, ²Munich/D E \n(bastien.le.guellec@gmail.com) \n \nPurpose or Learning Objective: To create a database of annotated radiology \nreports in diverse languages on which to test Large  Language Models (LLMs). \nMethods or Background: Large Language Models (LLMs) represent one of \nthe most important advancements in artificial intel ligence in recent years. In the \nfield of medicine, they hold the potential to tranf orm how physicians interact \nwith and interpret medical data. However, the curre nt research and \napplications of LLMs in medicine are predominantly focused on English-\nlanguage datasets. This narrow focus raises signifi cant concerns about the \nability of LLMs to generalize across the thousands of other languages. As a \nresult of difficulties accessing high-quality data from low-resources languages, \npatients may be excluded from the benefits of AI-dr iven advancements in \nhealthcare. To address this critical gap, we have l aunched the Polyglottal \nAnnotated Reports for Open Testing (PARROT) project . PARROT seeks to \ngather fictional medical reports from diverse lingu istic and cultural \nbackgrounds, manually annotated by experts for ICD- 10 codes and make them \nfreely accessible to the global research community.  PARROT is a completely \nopen-source initiative, inviting radiologists and m edical professionals from \naround the world to contribute. \nResults or Findings: 2648 annotated radiology reports from 75 radiologis ts \nfrom 20 countries in 13 languages have been collect ed. Most prevalent \nlanguages were Polish (808 reports), French (480 re ports) and Italian (285 \nreports). Contributions from the Global South inclu ded Ivory Coast, Mexico, \nMadagascar, Togo, Gabon, Argentina, Algeria, Turkey  and China. \nConclusion: The to the collaborative effort of 75 radiologists from 20 \ncountries, PARROT is the largest multilingual open database of radiology \nreports to date. \nLimitations: Annotation for this first iteration of PARROT is li mited to ICD-10 \ncodes. Most reports originated from European countr ies and European \nlanguages. \nFunding for this study: No specific funding \nEthics committee - additional information: None required \nAuthor Disclosures:  \nKeno Bressem: Nothing to disclose \nBastien Le Guellec: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nFriday \nAbstract-based Programme \n \n 189  \n16:00-17:30 Research Stage 4 \nResearch Presentation Session: Breast \nRPS 1602 \nInterventional radiology in breast cancer \ncare \n \nModerator \nI. Allajbeu; Cambridge/UK  \n(Ia359@cam.ac.uk;) \n \n \nDetection and characterization of cryoimmunologic r esponse induced by \nultrasound-guided cryoablation on early breast canc er: evaluation of \ncirculating markers \nF. Galati, *M. Pasculli*, C. Napoletano, V. Rizzo, R. Maroncelli, F. Cicciarelli, \nM. Nuti, C. Catalano, F. Pediconi; Rome/IT \n(marcella1.pasculli@gmail.com) \n \nPurpose or Learning Objective: Cryoablation is a minimally-invasive \nprocedure that uses cooling to induce necrosis of t he targeted tissue. All other \nminimally invasive techniques use hyperthermia, whi ch melts cell membranes \nand causes protein denaturation. In contrast, cryoa blation leaves tumor \nproteins and tumor-associated antigens intact, with  the potential to stimulate an \nanti-tumor immune response. Thus, the purpose of th is prospective pilot study \nwas to characterize the immune response induced by tumor cryoablation in \nblood samples from early breast cancer (BC) patient s. \nMethods or Background: We enrolled patients with early-stage BC, \nscheduled for breast surgery, not eligible for neo- adjuvant therapy, and with a \ncryo-feasible cancer location. Blood samples to ass ess immune response were \ntaken before (T0) and one week after cryoablation ( T1), and before (T2) and \none week after (T3) surgery. Analysis of T cell sub sets (CD3, CD8, CD4, \nCD137, and Tregs) and the inflammatory/damage molec ule HMGB1 were \nperformed by flow cytometry/ELISA. Circulating cyto kines were also analyzed \nusing the Luminex analysis. \nResults or Findings: From July 2022 to January 2023, ten patients underw ent \ncryoablation. Analysis of circulating markers of cr yo-immunological response \nrevealed a progressive release of HGBM1 after cryoa blation (T0-T1, p=0.04) \nuntil surgical resection of the primary tumor (T0-T 3, p=0.02). Cryoablation \nfollowed by surgery also induced a significant decr ease in CD137 T cell \nsubsets (total and CD4; p<0.01), and IL4 (T0-T3, p< 0.05). Finally, a significant \ndecrease in proliferative Treg cell subsets (Ki67+T regs; p<0.05) was observed. \nConclusion: In our pilot study, cryoablation induced the releas e of HMGB1, \nwhich acts to activate the primary phases of the im mune response, and the \ndecrease of the immunosuppressive Treg subset and t he pro-tumoral cytokine \nIL-4, probably released by CD4CD137 T cells. \nLimitations: Although the limited number of patients, cryoablati on was a \nvaluable method to enhance the anti-tumor response.  \nFunding for this study: The study has received funding from the Seed Grant \nfunding programme of the European Society of Radiol ogy (ESR) in \ncollaboration with the European Institute for Biome dical Imaging Research \n(EIBIR) kindly supported by an unrestricted, non-ex clusive grant from GE \nHealthcare. \nEthics committee - additional information: The study obtained the approval \nof the Institutional Review Board of “Sapienza” Uni versity of Rome (Ref.6528, \napproved 24.11.2021). \nAuthor Disclosures:  \nRoberto Maroncelli: Nothing to disclose \nVeronica Rizzo: Nothing to disclose \nFrancesca Galati: Nothing to disclose \nMarcella Pasculli: Nothing to disclose \nFederica Cicciarelli: Nothing to disclose \nMarianna Nuti: Nothing to disclose \nFederica Pediconi: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nChiara Napoletano: Nothing to disclose \n \n \n \n \n \n \n \n \n \nA convenient model based on mammography and magneti c resonance \nimaging for preoperative differentiation of breast phyllodes tumors and \nfibroadenomas \n*X. Ma*; Shanghai/CN \n(maxiaowen9397@163.com) \n \nPurpose or Learning Objective: To establish a fusion model based on \nmammography (MG) and magnetic resonance imaging (MR I) for the \npreoperative differentiation of breast phyllodes tu mors (PTs) and \nfibroadenomas (FAs). \nMethods or Background: The clinical data, MG images, and MR images of \npatients with breast FAs treated in our institution  from October 2019 to \nDecember 2020, as well as patients with PTs treated  from January 2011 to \nDecember 2020, were retrospectively collected. Univ ariate and multivariate \nlogistic regression analyses were conducted to sele ct independent factors and \nto construct a diagnostic model to differentiate PT s and FAs. The diagnostic \nperformance of the model was evaluated using the re ceiver operating \ncharacteristic (ROC) curve, calibration curve, and decision curve analysis \n(DCA). \nResults or Findings: A total of 147 patients with FAs and 138 patients w ith \nPTs were included in this study. The results of the  multivariate logistic \nregression analysis showed that patient age, maximu m diameter of mass, \ndensity on MG images, lobulation on MR images, and time-intensity curve \n(TIC) were independent factors contributing to the differential diagnosis. \nFinally, the fusion model showed satisfactory discr imination (area under the \ncurve (AUC) 0.90, 95% CI: 0.86-0.94) and calibratio n. DCA indicated good \nclinical benefit, as indicated by most values being  within threshold probabilities. \nConclusion: MG and MRI findings help differentiate between FAs and breast \nPTs preoperatively. The multimodal fusion model was  clinically efficacious and \nbeneficial and thus useful for accurate clinical di agnosis and treatment. \nLimitations: Our study is a retrospective and single-centre stud y, so it is \nnecessary to validate the accuracy of the research results through multicentre \nstudies. \nFunding for this study: None \nEthics committee - additional information: Fudan University Shanghai \nCancer Center \nAuthor Disclosures:  \nXiaowen Ma: Nothing to disclose \n \n \nCorrelation of radiological and pathological tumor sizes in early-stage \nbreast cancer based on molecular subtypes and accom panying in situ \ncarcinoma: A retrospective multicenter study \nD. E. Tekcan Sanli¹, *G. Esen*²; ¹Gaziantep/TR, ²Is tanbul/TR \n(gulesenicten@gmail.com) \n \nPurpose or Learning Objective: To compare the accuracy of radiological \ntumor sizes obtained by mammography (MMG), ultrason ography (US) and \nmagnetic resonance imaging (MRI) with pathological sizes and to determine \nwhether tumor size measurements change based on mol ecular subtypes and \nthe presence of accompanying ductal carcinoma in si tu (DCIS). \nMethods or Background: A total of 559 breast cancer patients diagnosed in \n11 university hospitals in Turkey between 2010 and 2022, underwent \npreoperative MMG, USG, and MRI, and did not receive  neoadjuvant \nchemotherapy (NAC) were included in the study. Tumo rs were divided into \nhistopathological (in-situ/invasive/ in-situ+invasi ve (mixed)) and molecular \n(Luminal A/B/HER2+/triple-) subgroups. Tumor sizes were measured on each \nmodality retrospectively and compared with the path ologic sizes reported in the \npostoperative pathology reports. Comparison was per formed based on \nhistological type (invasive/ in situ/ mixed), and m olecular subtypes. \nResults or Findings: The highest agreement in invasive tumors was obtain ed \nwith MRI (MRI:0.831, US:0.769, MMG:0.650). In the p resence of DCIS, the \nagreement was strong with MRI (r:0.770), moderate w ith MMG and US \n(r:0.517, r:0.593, respectively). In mixed tumors, agreement was strong with \nMRI (r:0.817), moderate with US (r:0.656), and low with MMG (r:0.499). Based \non molecular subtypes, highest correlation was obta ined with US and MRI in \nHER-2 (+) tumors (r:0.754, r:0.715, respectively), and with MRI in other \nsubtypes (Luminal A-B-triple (-)) with MRI (r:0.856 -0.815-0.858; respectively). \nThere was no statistically significant difference i n terms of other criteria. \nConclusion: This multicenter study shows that MRI is the most r eliable \nmethod in preoperative determination of tumor size for both invasive and in-situ \ntumors and all molecular subtypes. \nLimitations: In this retrospective study, the number of patients  was sufficient \nbut unbalanced when divided into subgroups. The pre sence of invasive lobular \ncarcinoma and axillary lymph nodes were not taken i nto account separately. \nFunding for this study: No \nEthics committee - additional information: Ethics committee approval is \nobtained from Acıbadem University. (No:2021-21/29) \nAuthor Disclosures:  \nGul Esen: Nothing to disclose \nDeniz Esin Tekcan Sanli: Nothing to disclose \n \n\n \n \nFriday \nAbstract-based Programme \n \n 190  \nBreast abnormalities identified on cross sectional imaging represent a \nsmall but significant subgroup of referrals to symp tomatic breast service \n*C. O'Brien*, M. R. Common, N. Hambly, N. Ní Mhuirc heartaigh, M. Bambrick, \nD. Duke, N. Healy, E. Downey; Dublin/IE \n \nPurpose or Learning Objective: Our retrospective study examines patients \nreferred to a symptomatic breast centre with breast  abnormality identified on \ncross-sectional imaging with special reference to f requency and outcomes. \nMethods or Background: All patients presenting to our symptomatic breast \ncentre over a 29-month period (Dec. 21-May 24) were  evaluated. Patients \nreferred following detection of breast abnormality on cross-sectional imaging \nwere identified. Results of subsequent breast imagi ng, image-guided biopsies \nand histopathology were analysed. \nResults or Findings: 13,336 consecutive referrals over a 29-month period  \nwere reviewed. 179 female/ 3 male patients, mean ag e 56.3 years (range 26 -\n89years), were referred with breast abnormality on cross-sectional imaging. \n158/182 (86.8%) were identified on CT, 9/182 (4,9%)  on MRI, 14/182 (7.7%) \non nuclear medicine, 1/182 (0.5%) on ultrasound. Cl inical breast examination \nwas normal in most patients. All patients underwent  breast imaging. 75/ 182 \n(41.2%) patients underwent image guided biopsy. His topathology results of \nbiopsies were benign in 26/75 (34.7 %), indetermina te (B3) in 5/74 (6.8%) and \nmalignant in 44/75 (58.6%). 2/5 patients with B3 hi stopathology underwent \nsurgical excision yielding final benign pathology, while 3/5 opted for lesion \nsurveillance. Overall cancer detection rate was 24. 2% (44/ 182). \nConclusion: While breast abnormalities detected on cross sectio nal imaging \nrepresent a small subgroup of referrals the cancer detection rate in this cohort \nis significant. With normal clinical examinations t hese patients are at risk of \ndelays along the traditional referral pathways base d on clinical suspicion of \nmalignancy. Direct referral of these patients to br east radiology ensures \nprioritisation based on lesion appearance on cross- sectional imaging. \nLimitations: Only cross-sectional breast abnormalities of patien ts referred to \nthe symptomatic breast service were reviewed. \nFunding for this study: Nil \nEthics committee - additional information: Approval obtained from the \nQuality and Safety Directorate team at our institut ion. \nAuthor Disclosures:  \nNiamh Hambly: Nothing to disclose \nDeirdre Duke: Nothing to disclose \nMarie Bambrick: Nothing to disclose \nMatthew R Common: Nothing to disclose \nEithne Downey: Nothing to disclose \nConor O'Brien: Nothing to disclose \nNeasa Ní Mhuircheartaigh: Nothing to disclose \nNuala Healy: Nothing to disclose \n \n \nMammography-based radiomic analysis in triple negat ive ductal invasive \nbreast cancer. Comparison between BRCA and not BRCA  mutated \npatients: Preliminary results \n*G. Sessa*, C. Beretta, C. Bozzola, F. Mogavero, C.  Parlato, L. Nocetti,  \nG. Ligabue, P. Torricelli, A. Pecchi; Modena/IT \n \nPurpose or Learning Objective: This study aims to evaluate the applicability \nof radiomics analysis to mammographic images of pat ients diagnosed with \ntriple negative breast cancer (TNBC) in order to id entify radiomics features that \ncan differentiate the mutational status of BRCA gen es. \nMethods or Background: This retrospective study included patients \nhistologically diagnosed with TNBC who performed a mammographic \nexamination between 2010 and 2021. Mammographic ima ges were reviewed \nand for each patient the tumor lesions were manuall y segmented in the \nmammographic projection where they were better dema rcable; a further \nelliptical ROI (region of interest) of standard siz e (100 mm2) was drawn in the \nmost homogeneous area of the controlateral healthy gland using the analogue \nmammographic projection of the same date or, if not  available, of the \ncorresponding bilateral mammographic investigation closer to the time of \ndiagnosis. Features from each ROI were extracted wi th Pyradiomics-3D. \nResults or Findings: The population included 50 patients and 51 lesions (12 \nBRCA+ patients and 13 lesions therein, 38 BRCAwildt ype patients and 38 \nlesions therein). The lesions included 37 nodules, 3 pathologic \nmicrocalcifications and 11 lesions appearing as nod ules with contextual \nmicrocalcifications. The segmentation was carried o ut on the following \nprojections: 24 LCC, 24 RCC, 21 LMLO, 2 LML, 21 RML O, 2RML. The first \npreliminary analysis demonstrated the feasibility o f the radiomics study in the \npopulation examined. Based on a previous study cond ucted on DCE-MRI of \nthe same target population, we expect to be able to  identify differences in \nradiomic patterns which represent the fenotipic exp ression of mutations \noccurring on a genetic level. \nConclusion: This study demonstrated the feasability of radiomic s analysis on \ndiagnostic mammograms of TNBC patients to build a p redictive model able to \ndiscriminate between carriers and non carriers of B RCA gene mutations. \nLimitations: Small population \nFunding for this study: No funding received for this study \nEthics committee - additional information: The research was approved by \nthe Area Vasta Emilia Nord Est Ethical Committee (4 63/2020/OSS/AOUMO) \nSIRER ID 236 - EMENDAMENTO SOSTANZIALE 1.0 (prot. A OU 0008681/22 \ndel 23/03/2022) and informed consent was obtained f rom all subjects. \nAuthor Disclosures:  \nGiulia Sessa: Nothing to disclose \nPietro Torricelli: Nothing to disclose \nAnnarita Pecchi: Nothing to disclose \nLuca Nocetti: Nothing to disclose \nGuido Ligabue: Nothing to disclose \nChiara Bozzola: Nothing to disclose \nCecilia Beretta: Nothing to disclose \nFrancesca Mogavero: Nothing to disclose \nChiara Parlato: Nothing to disclose \n \n \nCan the new combined CB SCORE reduce the number of breast false \npositive biopsies? Results from a monocentric study  \n*A. Portaluri*¹, F. M. Arico¹, X. Wang², T. Zhang²,  C. Sofia¹, E. Condorelli¹,  \nF. Catanzariti¹, M. A. Marino¹; ¹Messina/IT, ²Amste rdam/NL \n(antonio_3_ap@libero.it) \n \nPurpose or Learning Objective: To evaluate the performance of a new \ncombined score, CEUS-BI-RADS (CB) score, in differe ntiating between benign \nand malignant lesions and in reducing the number of  unnecessary breast \nbiopsies. \nMethods or Background: 331 women with a new breast lesion scheduled for \nUS-guided biopsy (sonographically assesed as BI-RAD S 4 a-c and 5) were \nenrolled. For each lesion, CEUS examination was per formed before the \nbiopsy. 2 single reader (1 high-experienced breast radiologist and 1 trainee) \nindependently assessed all CEUS studies with qualit ative analysis, in terms of \ntime and intensity of enhancement, enhancement patt ern and size increase \nafter contrast administration, assigning them a CEU S score from 0 to 3. \nMoreover a score from 1 to 4 was assigned for each BI-RADS category \nsonographically assesed as BI-RADS 4a, 4b, 4c and 5 . Finally a combined \nscore (CB score) was obtained. Descriptive statisti cs, area under the curve \n(AUC), receiver operating characteristic (ROC) anal ysis, sensitivity and \nspecificity were used to investigate the diagnostic  performance of the \ncombined approach. Inter-reader reliability was mea sured using Cohen’s \nkappa statistics. \nResults or Findings: 294 lesions have been found. CB score showed the \nhighest diagnostic performance compared to the CEUS  score alone (average \nAUCs = 0.935 vs 0. 890, p <0.0001), the highest sen sitivity (87.9% vs 91.3%) \nand specificity (75.2% vs 80.7%) . Moderate ( κ= 0. 45) agreement was found \nbetween the two readers. Finally, CB score would ha ve obviated up to 40% of \nunnecessary biopsies. \nConclusion: CB score allowed to improve the diagnostic performa nce of \nCEUS alone for lesions assesment, reducing the rate  of false positives \nbiopsies for both experienced and unexperienced rea der. \nLimitations: Deeper studies on the application of this score nee d to be carried \nout to improve the identification of small tumours that have minimal blood \nsupply. \nFunding for this study: None \nEthics committee - additional information: This is a single-center, \nretrospective study. Ethics committee approval was waived as all Contrast-\nEnhanced Ultrasound (CEUS) examinations were conduc ted as part of routine \nclinical care, following established protocols at o ur institution. These \nexaminations were clinically indicated prior to bio psy, ensuring that no \nadditional procedures were performed outside the St andard of Care (SOC). \nAuthor Disclosures:  \nCarmelo Sofia: Nothing to disclose \nElvira Condorelli: Nothing to disclose \nMaria Adele Marino: Nothing to disclose \nFrancesca Catanzariti: Nothing to disclose \nXin Wang: Nothing to disclose \nAntonio Portaluri: Nothing to disclose \nTianyu Zhang: Nothing to disclose \nFrancesco Marcello Arico: Nothing to disclose \n \n \nNon-surgical treatment of breast cancer: a comparis on of outcomes \nbetween Cryoablation with hormonal therapy versus C ryoablation alone \nand hormonal therapy alone in patients not eligible  for surgery \n*S. E. Baldi Giorgi*, F. Di Naro, G. Migliaro, F. P ugliese, T. Amadori, S. Vidali, \nG. Bicchierai, V. Miele, J. Nori; Florence/IT \n(sofiabaldi97@gmail.com) \n \nPurpose or Learning Objective: This study aims to evaluate the most \neffective non-surgical treatment for breast cancer in surgery-ineligible patients, \ncomparing ultrasound-guided Cryoablation combined w ith hormonal therapy \n(HT) versus Cryoablation alone and hormonal therapy  alone. \n\n \n \nFriday \nAbstract-based Programme \n \n 191  \nMethods or Background: A cohort of 64 patients (mean age 83.4 years) not-\nsuitable for surgery due to comorbidities and/or ad vanced age was enrolled, \nwith a total of 73 biopsy-confirmed malignant breas t lesions. All the lesions \nwere invasive ductal carcinomas (mean size 14.8 mm) , hormone-positive and \nHER2-negative, with no ultrasound-visible lymph nod e involvement. Patients \nwere divided into three groups: 36 lesions were tre ated with Cryoablation and \nHT, 19 with Cryoablation only and 18 with HT only. A locoregional staging was \nperformed at baseline with contrast-enhanced-mammog raphy (CEM) and \nultrasound, followed by CEM and ultrasound follow-u p at 12 months post-\ntreatment. Only patients completing the follow-up w ere included. Lesion size \nwas compared at baseline and 12 months after-treatm ent. Fisher's exact test \nwas used for group comparison. \nResults or Findings: Of the 73 lesions, 47 completed the 12-months follo w-\nup: 20 in the Cryoablation-with-HT group, 9 in the Cryoablation-only group, and \n18 in the HT-only group. Tumor size reduction was g reatest in the \nCryoablation-with-HT group (94%, mean reduction of 15.4 mm), followed by \nCryoablation-only (82%, mean reduction of 9.7 mm), and HT-only (43%, mean \nreduction of 4.6 mm). Tumors with absent or low CEM -enhancement, \nsuggesting residual-disease reduction, were most fr equent in the Cryoablation-\nwith-HT group (80%), followed by Cryoablation-only (77.8%), and HT-only \n(38.9%). Fisher’s test revealed a significant diffe rence between the \nCryoablation-with-HT and the HT-only groups (p<0.00 15), expressing the \nadded value of Cryoablation. \nConclusion: Cryoablation with hormonal-therapy significantly re duces tumor \nsize and residual disease more effectively than the rapy alone, making it a \npromising option for patients not-eligible for surg ery. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by \nComitato Etico Regione Toscana - Pediatrico (refere nce number: 165/2024). \nAuthor Disclosures:  \nGiulia Bicchierai: Nothing to disclose \nFrancesca Pugliese: Nothing to disclose \nJacopo Nori: Nothing to disclose \nSofia Vidali: Nothing to disclose \nVittorio Miele: Nothing to disclose \nSofia Elisabetta Baldi Giorgi: Nothing to disclose \nTommaso Amadori: Nothing to disclose \nGiuliano Migliaro: Nothing to disclose \nFederica Di Naro: Nothing to disclose \n \n \nVacuum-Assisted Biopsy in The Era of Low-Risk Ducta l Carcinoma in \nSitu (LR-DCIS) Active Surveillance: Real World Data  and Implications \n*H. L. Couto*¹, C. N. Valadares², B. F. De Paula Ri cardo¹, A. N. Soares¹,  \nP. H. Toppa¹, B. A. Coelho³, C. C. Pessoa⁴, N. Sharma⁵, E. Carvalho Pessoa⁴; \n¹Belo Horizonte/BR, ²São Paulo/BR, ³Montes Claros/B R, ⁴Botucatu/BR, \n⁵Leeds/UK \n(enriquecouto@hotmail.com) \n \nPurpose or Learning Objective: Evaluate vacuum assisted biopsy (VAB) as \ndiagnostic test of LR-DCIS in the context of real-w orld clinical practice. \nMethods or Background: Database analysis of 116 cancers [both invasive \nbreast cancers (IC) and ductal carcinoma in situ (D CIS)] diagnosed by VAB \nsubmitted to standard surgical treatment with compl ete histological data from \nVAB and surgery from 04/13/2017 to 11/28/2020. The VAB results were \nmatched to the surgical pathology, considered the g old standard. The \npathological diagnoses were grouped into malignanci es requiring immediate \nsurgical treatment [DCIS with high risk (HR-DCIS) o f progression to IC or IC] \nversus those eligible to active surveillance (LR-DC IS). HR-DCIS/IC were \nconsidered positive while LR-DCIS negative results.  VAB sensitivity, specificity, \npositive predictive value (PPV), negative predictiv e value (NPV), and accuracy \nwere obtained. \nResults or Findings: Median age was 56 (20-91); final median invasive tu mor \nsize (T) of 6mm (0,8 – 25) and 8mm (2 – 65) for DCI S; 65.52% were US-\nguided (70/116) and 44.48% (46/116) stereotactic gu ided; 42.24% (49/116) \nwere only masses, 26,72% (31/116) masses associated  with calcifications and \n31.03% (36/116) calcifications. Out of 116 malignan cy cases diagnosed by \nVAB, 15 (12.9%) resulted in LR- DCIS in the biopsy,  10 (8.6%) confirmed LR-\nDCIS in surgery, and 5 (4.3%) were upgraded to HR-D CIS/IC in surgery. VAB \nshowed 95.28% sensitivity and 100% specificity. The  positive predictive value \n(PPV) was 100%, and the negative predictive value ( NPV) was 66.67%. Of the \n5 false negatives (FN) LR-DCIS upgraded in surgery:  3 were HR-DCIS and 2 \nIC (pT1a-bpN0-luminal). \nConclusion: VAB, based in conventional pathology and \nimmunohistochemistry, had an elevated FNR LR-DCIS i n real world practice \nand, if applied, VAB LR-DCIS upgraded cases could b e treated by either \nhormone or radiation therapy isolated or combined c ounterbalanced by slight \nreduction of overtreatment. \nLimitations: Retrospective data base \nFunding for this study: None \nEthics committee - additional information: The study was approved by the \nEthics Committee of Santa Casa of Belo Horizonte un der the number \n25761019.8.0000.5138, and all methods were carried out in accordance with \nnational guidelines \nAuthor Disclosures:  \nNisha Sharma: Nothing to disclose \nBernardo F. De Paula Ricardo: Nothing to disclose \nCarla Carvalho Pessoa: Nothing to disclose \nBertha Andrade Coelho: Nothing to disclose \nCarolina Nazareth Valadares: Nothing to disclose \nHenrique Lima Couto: Nothing to disclose \nEduardo Carvalho Pessoa: Nothing to disclose \nPaola H. Toppa: Nothing to disclose \nAleida N. Soares: Nothing to disclose \n \n \nDoes touch imprint cytology prepared from core need le biopsy \nspecimens in breast lesions provide an immediate di agnosis? \nŞ. Kökten, H. Kılın, *N. Voyvoda*; Istanbul/TR \n(nuraykad@gmail.com) \n \nPurpose or Learning Objective: The touch imprint method is used during the \nfrozen study of sentinel lymph node samples in brea st cancer patients and \nprovides a rapid response with high accuracy and en sures the correct \nmanagement of patients during surgery. Similarly, i mprint preparations made \nfrom core biopsies can also be helpful in reaching rapid and accurate results. \nThe aim of this study was to determine the diagnost ic value and accuracy of \nthe imprint method. \nMethods or Background: Between January 2024 and March 2024, patients \nwho were referred to the breast imaging center of o ur hospital due to a mass in \nthe breast and planned for US-guided core needle bi opsy were included in the \nstudy. Touch imprint and core biopsy specimens were  retrospectively \nevaluated at different times by the same pathologis t. Pathological findings in \ntouch imprint evaluation were classified using the guideline. \nResults or Findings: 201 lesions of 178 patients with an average age of 48.05 \n(min: 16-max 82) were evaluated. Of the 201 lesions , 186 were breast lesions \nand 15 were axillary lymph nodes. Seven of 186 brea st lesions were excluded \nfrom the evaluation because touch imprint was defin ed as insufficient. The \nsensitivity of touch imprint was calculated as 87.5 0%, specificity as 89.16%, \nppv: 54.66%, npv as 97.95% and accuracy as 88.94%. If we exclude the \nlymphoma patient, the sensitivity of imprint for ly mph node was calculated as \n100%, specificity as 83.3%, ppv: 47.27%, npv as 100 % and accuracy as \n85.5%. When not excluded, the sensitivity of imprin t cytology was 88.89%, \nspecificity as 83.33%, ppv: 44.35%, npv as 98.05% a nd accuracy as 84.06%. \nConclusion: Imprint cytology prepared from core biopsies of bre ast lesions \ncan provide highly accurate and rapid diagnosis. Th us, treatment can be \nstarted without delay. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Local ethics committee approval \n(decision no: 2024/010.99/3/19) was obtained before  the study. \nAuthor Disclosures:  \nŞermin Kökten: Nothing to disclose \nNuray Voyvoda: Nothing to disclose \nHalil Kılın: Nothing to disclose \n \n \nDifferential efficacy of Cryoablation in breast can cer subtypes: \nultrasound-guided scar biopsy evaluation year post- treatment \n*F. Pugliese*, F. Di Naro, G. Migliaro, S. E. Baldi  Giorgi, T. Amadori,  \nD. De Benedetto, C. Bellini, V. Miele, J. Nori; Flo rence/IT \n(francescapugliese28@gmail.com) \n \nPurpose or Learning Objective: To assess the effectiveness of cryoablation \n(CR) in different subtypes of breast tumors, a coho rt of 39 biopsy-proven B5 \nlesions underwent ultrasound-guided scar biopsy eva luation one year post-\ntreatment. \nMethods or Background: From 2022-2023, the B5 lesions comprised 35 \ninvasive ductal carcinoma (IDC), 2 IDC-associated D CIS, and 2 invasive \nlobular carcinomas. All patients were deemed inoper able for advanced age \nand comorbidities, leading to their enrollment in a n annual follow-up and \nultrasound-guided scar biopsy. The study population  was stratified into three \nsubgroups: molecular subtype, dimensional cut-off, and growth index. \nResults or Findings: These lesions were hormone-responsive, with 19 \nclassified as Luminal A and 20 as Luminal B. The di mensional cutoff ranged to \n2.5 cm, with ice ball dimensions tailored to encomp ass a one-centimeter \nmargin around the lesions. Data analysis revealed n otable differences in the \nefficacy of cryoablation among the various subgroup s. When considering \nLuminal A lesions with ki67>20% the complete ablati on rate was 84.2% and for \nLuminal B it was 90.0%. Conversely, tumors with Ki6 7 expression ≤20% \nexhibited higher complete ablation rates, with Lumi nal A reaching 100% and \nLuminal B at 84.6%. Additionally, lesions ≤10 mm in size exhibited a higher \n\n \n \nFriday \nAbstract-based Programme \n \n 192  \ncomplete ablation rate of 100% compared to lesions >10 mm, which showed \nan 80.0% success rate. Histologically, CR was ineff ective in achieving \ncomplete ablation in DCIS cases, presenting a rate of 0%, but other subtypes \ndemonstrated a higher complete ablation rate at 91. 9%. However, none of \nthese differences were statistically significant. \nConclusion: Cryoablation emerges as a promising primary treatme nt option \nfor breast cancer is a safe and effective nonsurgic al alternative to breast-\nconserving surgery in select patients with unifocal  IDC low grade, hormone \nreceptor-positive, and ≤10 mm size \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study \nEthics committee - additional information: The study is retrospective \nAuthor Disclosures:  \nFrancesca Pugliese: Nothing to disclose \nDiego De Benedetto: Nothing to disclose \nJacopo Nori: Nothing to disclose \nVittorio Miele: Nothing to disclose \nSofia Elisabetta Baldi Giorgi: Nothing to disclose \nTommaso Amadori: Nothing to disclose \nChiara Bellini: Nothing to disclose \nGiuliano Migliaro: Nothing to disclose \nFederica Di Naro: Nothing to disclose \n \n \n \n \n \n \n \n\n \n \n 193  \n \n \n  \nSaturday, March 1 \n\n \n \nSaturday \nAbstract-based Programme \n \n 194  \n \n08:00-09:00 Research Stage 1 \nResearch Presentation Session: \nGenitourinary \nRPS 1707 \nAdvances in imaging techniques for the \ngenitourinary tract \n \nModerator \nA. Shinagare; Boston, MA/US  \n \n \nSynthesising and adapting multiple data sources to design \nenvironmentally sustainable and clinically effectiv e imaging pathways for \nvisible haematuria \nJ. B. John¹, K. O'Flynn², *S. Lambracos*³, B. Abdel qader⁴, S. Nalagatla⁴,  \nS. Khadouri⁵, T. W. R. Briggs³, W. K. Gray³, J. Mcgrath¹; ¹Exet er/UK, \n²Salford/UK, ³London/UK, ⁴Glasgow/UK, ⁵Leeds/UK \n(simon.lambracos@gmail.com) \n \nPurpose or Learning Objective: To design a risk-stratified imaging pathway \nthat reduces greenhouse gas (GHG) emissions using d iagnostic performance \nevidence for ultrasound and computerised tomography  urogram (CTU) in \ndetecting upper tract urothelial cancer (UTUC). \nMethods or Background: An audit of 15 UK hospitals’ first-line imaging for  \nvisible haematuria (VH) and non-visible haematuria (NVH), and use of one-\nstop cystoscopy and imaging, was performed. Urology  referral data from the \nIDENTIFY study (N = 10,896) were linked to national  Hospital Episode \nStatistics data to estimate absolute numbers of pat ients receiving ultrasound or \nCTU first-line across England annually. \nResults or Findings: Ultrasound was the first-line imaging choice for VH  and \nNVH in 53% and 93% of hospitals respectively; other  hospitals used CTU as \nfirst-line. One-stop assessment was performed in 44 % of audited cases. An \nestimated 127,701 ultrasound and 77,880 CTU were pe rformed across \nEngland annually, including an estimated 20% and 2%  of additional CTU for \npatients with VH and NVH respectively receiving ult rasound first due to \npersistent VH. Informed by these data, a risk-strat ified imaging pathway for \npatients with haematuria was developed. The pathway  comprised: one-stop \nclinic ultrasound first-line for all, additional CT U for high-risk cases (VH + \nage>65 + smoking history), patient-initiated follow  up direct to CTU for \npersistent VH if not performed initially. We estima te that this pathway could \nresults in 191,097 ultrasounds, 51,320 CTU and 73% of assessments \ncompleted with one-stop assessment, leading to an e stimated 269 tonnes \nCO2e reduction in net greenhouse gas emissions acro ss England for one year. \nAround 0.1% of patients would have UTUC missed with  first-line ultrasound on \nthis pathway. \nConclusion: Adoption of this evidence-based pathway will reduce  GHG \nemissions whilst delivering greater risk-stratified  imaging use, improving equity \nof healthcare access and reducing unwarranted use o f CTU. \nLimitations: N/A \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nJoseph B John: Nothing to disclose \nJohn Mcgrath: Nothing to disclose \nSinan Khadouri: Nothing to disclose \nBushra Abdelqader: Nothing to disclose \nSarika Nalagatla: Nothing to disclose \nSimon Lambracos: Nothing to disclose \nTim W R Briggs: Nothing to disclose \nKieran O'Flynn: Nothing to disclose \nWilliam K Gray: Nothing to disclose \n \n \nStudy on Optimization of Image Quality of Computed Tomography \nUrography Using Dual-layer Detector Spectral CT Vir tual Monoenergetic \nImaging Technique Combined with Contrast Split Bolu s Protocol \n*F. Zhao*, K. Li; Xi'an/CN \n(553785313@qq.com) \n \nPurpose or Learning Objective: To investigate the optimal keV of dual-\ndetector spectral CT virtual monoenergetic imaging (VMI) in computed \ntomography urography (CTU). \n \n \nMethods or Background: Hematuria patients who underwent dual-layer \ndetector spectral CTU were collected. In the experi mental group, 25mL of \ncontrast bolus was injected first, and then 20mL of  contrast bolus was injected \n15 minutes later, and cortico-excretory phase CT sc an was performed. In \ncontrol group, non-contrast phase CT scan was perfo rmed first, then 100mL of \ncontrast bolus was injected, and cortical, medullar y and excretory phases CT \nscans were performed later. The experimental group was reconstructed into \n40keV, 50keV, 60keV, 70keV VMI images and mixed ene rgy images, and the \nexcretory phase in control group was reconstructed into mixed energy image. \nThe differences of CT value, standard deviation (SD ), signal-to-noise ratio \n(SNR), contrast-to-noise ratio (CNR), subjective sc ore of the image quality \nwere compared. \nResults or Findings: There was no significant difference in SNR and CNR \nbetween the mixed energy images of the control grou p and the VMI 50keV \nimages of the experimental group. In the score of i mage display effect, the VMI \n50keV image of the experimental group had the highe st score. In the score of \nfilling degree of contrast bolus, VMI 40keV images in the experimental group \nhad the highest score.The effective dose in the exp erimental group was about \n19.96% of that in the control group, and the amount  of contrast bolus in the \nexperimental group was about 45% of that in the con trol group. \nConclusion: The dual-layer detector spectral CT VMI technique u sed in CTU \ncan improve the image quality while reducing the am ount of contrast bolus and \nradiation dose, and VMI 40～50keV is the best energy level for image display. \nLimitations: A single-center study with a small number of cases \nFunding for this study: This study was funded by the hospital by RMB 10,000  \nEthics committee - additional information: This study was approved by the \nEthics Committee \nAuthor Disclosures:  \nFanhui Zhao: Nothing to disclose \nKai Li: Nothing to disclose \n \n \nProspective evaluation of high-resolution diffusion -weighted imaging \naccelerated by deep-learning reconstruction in mult iparametric MRI of \nthe prostate \n*S. Ursprung*¹, J. Herrmann¹, E. Weiland², D. Nicke l², A. Lingg¹, S. Afat¹,  \nS. Gassenmaier¹; ¹Tübingen/DE, ²Erlangen/DE \n \nPurpose or Learning Objective: Prostate MRI is a gatekeeper for more \ninvasive investigations in prostate cancer diagnosi s. This study investigates the \npotential of deep-learning reconstruction of high-r esolution (HR) diffusion-\nweighted imaging (DWI) to improve image quality and  lesion detectability. \nMethods or Background: Prospective study comparing multiparametric MRI \naccording to PI-RADS 2.1 specifications with standa rd (DWI-Std) and DWI-HR \n(4-fold higher in-plane resolution) on a 3T MRI-sys tem. Two radiologists (7yr \nexperience) compared image quality of DWI-Std/DWI-H R qualitatively and \nquantitatively using the Prostate Imaging Quality S coring System (PI-QUALv2), \na 5-point Likert-scale assessing sharpness, noise, artefacts, overall impression \nand diagnostic confidence, and contrast-to-noise ra tio (CNR) of prostatic \nlesions. \nResults or Findings: 91 patients consented (17 exclusions for prostatect omy, \n2 radiotherapy, 3 incomplete imaging); 69 patients were included. Average \nacquisition time for DWI-Std/DWI-HR was 04:30min/05 :33min. DWI-DL showed \nhigher sharpness at all b-values and on ADC maps (p <0.001). This came at \nthe cost of higher noise on b1000 and ADC (p<0.001) , resulting in comparable \nlesion detectability (p=0.28) and PI-RADS scoring ( p=1). Readers favoured \nDWI-HR in 47% and DWI-Std in 33%. The PI-QUAL sub-s core of DWI-\nStd/DWI-HR was similar (p=0.37); only the reduced S NR/Contrast in DWI-HR \napproached significance (p=0.054). PI-QUAL was opti mal in 67%/61%, \nacceptable in 28%/33% and inadequate in 6%/6% when including DWI-\nStd/DWI-HR. The CNR of PI-RADS 3-5 lesions in the P Z was significantly \nhigher at b1000/on ADC maps from DWI-Std (median 10 .8 vs. 10.3, p=0.002 at \nb1000, 10.1 vs. 7.7, p=0.0002 for ADC). The CNR bet ween PZ/TZ was higher \nin DWI-Std (median 3.0 vs. 2.2; p=0.04). \nConclusion: We prospectively evaluated high-resolution DWI with  DL-\nreconstruction, showing improved sharpness at sligh tly reduced CNR and \nmaintained diagnostic performance in prostate MRI. \nLimitations: This study was conducted on scanners of a single ve ndor. \nConfirmatory histology was available for PI-RADS 3+  lesions only. \nFunding for this study: This study received no funding. \nEthics committee - additional information: Tubingen University Hospital IRB \nAuthor Disclosures:  \nJudith Herrmann: Nothing to disclose \nElisabeth Weiland: Employee: Siemens Healthineers \nSaif Afat: Speaker: Siemens Healthineers Research/G rant Support: Siemens \nHealthineers \nDominik Nickel: Employee: Siemens Healthineers \nAndreas Lingg: Nothing to disclose \nStephan Ursprung: Nothing to disclose \nSebastian Gassenmaier: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 195  \nUltra-Fast Biparametric MRI in Prostate Cancer Asse ssment: Diagnostic \nPerformance and Image Quality Compared to Conventio nal \nMultiparametric MRI \n*A-M. Pausch*, V. Filleböck, C. Elsner, N. Rupp, D.  Eberli, A. M. Hötker; \nZurich/CH \n \nPurpose or Learning Objective: To compare the diagnostic performance and \nimage quality of a deep-learning-assisted ultra-fas t biparametric MRI (bpMRI) \nwith the conventional multiparametric MRI (mpMRI) f or the diagnosis of \nclinically significant prostate cancer (csPCa). \nMethods or Background: This IRB-approved prospective single-center study \nenrolled 123 biopsy-naïve patients undergoing conve ntional mpMRI and \nadditionally ultra-fast bpMRI at 3T between 06/2023 -02/2024. Two radiologists \n(R1: 4 years and R2: 3 years of experience) indepen dently assigned PI-RADS \nscores (PI-RADS v2.1) and assessed image quality (m PI-QUAL score) in two \nblinded study readouts. Weighted Cohen’s Kappa ( κ) was calculated to \nevaluate inter-reader agreement. Diagnostic perform ance was analyzed using \nclinical data and histopathological results from cl inically indicated biopsies. \nResults or Findings: Inter-reader agreement was good for both mpMRI ( κ = \n0.83) and ultra-fast bpMRI (κ = 0.87). Both readers demonstrated high \nsensitivity (≥94%/≥91%, R1/R2) and NPV (≥96%/≥95%) for csPCa detection \nusing both protocols. The more experienced reader m ostly showed notably \nhigher specificity (≥77%/≥53%), PPV (≥62%/≥45%), and diagnostic accuracy \n(≥82%/≥65%) compared to the less experienced reader. There  was no \nsignificant difference in the diagnostic performanc e of correctly identifying \ncsPCa between both protocols (p>0.05). The ultra-fa st bpMRI protocol had \nsignificantly better image quality ratings (p<0.001 ) and achieved an 80% \nreduction in scan time compared to mpMRI. \nConclusion: Deep-learning-assisted ultra-fast bpMRI protocols o ffer a \npromising alternative to conventional mpMRI for dia gnosing csPCa in biopsy-\nnaïve patients with comparable inter-reader agreeme nt and diagnostic \nperformance at superior image quality. However, rea der experience remains \nessential for diagnostic performance. \nLimitations: The single-center design and the exclusion of some patients with \na PI-RADS ≥ 3 who did not undergo biopsy may limit the general izability of our \nfindings. \nFunding for this study: Holcim Stiftung zur Förderung der wissenschaftliche n \nFortbildung \nEthics committee - additional information: Cantonal Ethics Commission \nZurich \nAuthor Disclosures:  \nNiels Rupp: Nothing to disclose \nClara Elsner: Nothing to disclose \nDaniel Eberli: Nothing to disclose \nVivien Filleböck: Nothing to disclose \nAntonia-Maria Pausch: Nothing to disclose \nAndreas M. Hötker: Nothing to disclose \n \n \nIncreased dose of iodine contrast media does not in crease the odds of \ncontrast-associated acute kidney injury \n*F. B. Berglund*, P. Liss, R. Frithiof; Uppsala/SE \n(felix.berglund@uu.se) \n \nPurpose or Learning Objective: The aim of this study is to investigate if any \nof the two components of the g-I/eGFR ratio is inde pendently associated with \nthe development of contrast-associated acute kidney  injury (CA-AKI). \nMethods or Background: All patients admitted to the intensive care units \n(ICUs) of a tertiary hospital from January 2013 to February 2020 were \nretrospectively identified. Those who underwent iod ine contrast media (ICM)-\nenhanced CT exams were included in this nested case -control study. CA-AKI \nwas defined and staged based on the creatinine and urine output criteria set \nforth by the Kidney Disease Improving Global Outcom es guidelines. The two \ncomponents of the ratio, the dose of ICM (measured in grams of iodine) and \nrenal function estimated by plasma creatinine, were  analyzed separately in \nrelation to the odds of developing CA-AKI. \nResults or Findings: Among the 214 patients included in the analysis PC- AKI \noccurred in 42 of the patients (19.6%). Median age was 61.5 years (IQR 40-73) \nand 59.3% were of male sex. Renal function at the d ay of the CT-scan differed \nbetween those developing PC-AKI (eGFR 56.9, IQR 35- 87) and those that did \nnot (eGFR 81.6, IQR 58-96). However, the dose of IC M was not associated \nwith PC-AKI development (OR 1.31 (IQR 0.49-3.47), p =0.827). \nConclusion: In this case-control study, renal function at the d ay of the \nexamination but not the administered dose of iodine  contrast media was \nassociated with PC-AKI. This suggest that including  injected amount of iodine \ncontrast media as a variable to clinically predict risk of PC-AKI is futile. \nLimitations: This is a single center case-control study where on ly 42 critically \nill patients developed CA-AKI. This reduces the gen eralizability of the study as \nwell as its power. \nFunding for this study: The study was supported with funding from ALF from \nUppsala University Hospital, and the Swedish Resear ch Council (2014-02569 \nand 2014-07606). Funding bodies had no role in the design of the study, data \ncollection, interpretation, or in the writing of th e manuscript. \nEthics committee - additional information: This study was approved by the \nSwedish Ethical Review Authority (Dnr 2017/168 with  amendment 2020-\n00135). Declaration of Helsinki and its subsequent revisions were observed. \nAuthor Disclosures:  \nRobert Frithiof: Nothing to disclose \nPer Liss: Nothing to disclose \nFelix Björn Berglund: Nothing to disclose \n \n \nModified in-plane resolution while maintaining high  image quality T2-\nweighted prostate MRI \n*S. J. Riederer*¹, E. Borisch¹, A. Froemming¹, A. K awashima², N. Takahashi¹; \n¹Rochester, MN/US, ²Phoenix, AZ/US \n(riederer@mayo.edu) \n \nPurpose or Learning Objective: To determine if an axial T2-weighted spin-\necho (T2-WI) sequence with modified in-plane spatia l resolution could provide \nnon-inferior performance and reduced acquisition ti me vs. a standard PI-\nRADSv2.1-compatible sequence. \nMethods or Background: Both sequences used 3 mm slice thickness, 16 cm \nFOV, acceleration R=1.5, TR>3000, TE 150. The PI-RA DS-compatible \nsequence used 400×230 in-plane sampling, 0.4 mm × 0 .7 mm resolution \n(0.280 mm2 pixel area). The modified sequence used 320×280 in-plane \nsampling, 0.5 mm × 0.57 mm resolution (0.285 mm2 pi xel area). Although the \nincreased phase sampling of the modified sequence ( 280 vs. 230) required \nmore repetitions, the reduced frequency sampling (3 20 vs. 400) allowed lower \nbandwidth and reduced averaging. The two sequences were both applied to 62 \nconsecutive subjects identified for clinical prosta te MRI. The number of slices \nwas patient-specific but identical for the two sequ ences. Results were blindly \nreviewed by three experienced radiologists. Each se ries was assessed \nindividually for Image Quality (IQ) using a 0-3 sca le. For each subject the two \nseries were also compared for overall preference on  a five-point (-2, -1, 0, +1, \n+2) scale. \nResults or Findings: Scan time depended on slice count (29 to 45). Scan time \nrange for the PI-RADS-compatible sequence was 2:56- 5:04 (median 3:54) and \nfor the modified sequence 2:16-3:56 (median 3:01). Scan time reduction using \nthe modified sequence was 37 to 80 sec (median 53 s ec). 51/62=82.2% of the \nPI-RADS-compatible series and 56/62=90.3% of the mo dified sequence were \nevaluated as diagnostically interpretable (IQ=2,3).  Reviewer-averaged scores \nshowed a significant preference for the modified se quence (p<0.001). \nConclusion: Although not adherent to PI-RADSv2.1 guidelines, ac quisition \nwith essentially identical in-plane pixel area (0.2 8 mm2) allows 53 sec (23%) \nreduction in acquisition time and significantly imp roved reviewer preference vs. \na PI-RADS-adherent sequence. \nLimitations: Limited number of subjects \nFunding for this study: This work was supported by NIH. \nEthics committee - additional information: Approved by Institutional Review \nBoard (IRB) \nAuthor Disclosures:  \nAdam Froemming: Nothing to disclose \nStephen J. Riederer: Nothing to disclose \nEric Borisch: Nothing to disclose \nNaoki Takahashi: Nothing to disclose \nAkira Kawashima: Nothing to disclose \n \n \nCEST Imaging vs. DWI with and without CEST Imaging:  Capability for \nDistinguishing Malignant from Benign Prostatic Area s \n*T. Ueda*, H. Nagata, M. Nomura, T. Yoshikawa, D. T akenaka, Y. Ozawa,  \nY. Ohno; Toyoake/JP \n(yohno@fujita-hu.ac.jp) \n \nPurpose or Learning Objective: 3D Chemical exchange saturation transfer \n(CEST) imaging is recently developed to obtain CEST  information within entire \ntumor. The purpose of this study was to compare the  capability for \ndistinguishing malignant from benign prostatic area s among 3D CEST imaging, \ndiffusion weighted imaging (DWI) and combined both information. \nMethods or Background: Fifty-two suspected prostatic cancer patients \nunderwent DWI at b value as 0 and 1500 s/mm2 and 3D  CEST imaging, \nsurgical treatments and pathological examinations. According to the \npathological results, 154 areas were determined as malignant prostatic areas, \nand 154 out of 470 areas were computationally selec ted as benign prostatic \nareas. On each 3D CEST imaging, magnetization trans fer ratio asymmetry \n(MTRasym) at 3.5 ppm map was generated. Then, 308 R OIs were placed over \nmalignant or benign areas on each map, and MTRasym and ADC values were \ndetermined. Each index was compared between maligna nt and benign areas \nby Student’s t-test. ROC analysis was performed to compare diagnostic \nperformance among MTRasym, ADC and combined discrim inators. Finally, \nsensitivity, specificity and accuracy were compared  among all methods by \nMcNemar’s test. \n\n \n \nSaturday \nAbstract-based Programme \n \n 196  \nResults or Findings: MTRasym and ADC of malignant area had significant \ndifferences with those of benign area (MTRasym: p<0 .0001, ADC: p<0.0001). \nArea under the curves (AUC) of combined discriminat ors (AUC=0.86) was \nsignificantly better than that of MTRasym (AUC=0.81 , p=0.001) and ADC \n(AUC=0.76, p<0.0001). Specificity (SP) and accuracy  (AC) of combined \ndiscriminators (SP: 72.1%, AC: 78.6%) were signific antly higher than those of \nMTRasym (SP: 60.4%, p<0.0001; AC: 73.1%, p<0.0001) and ADC (SP: 64.2%, \np<0.0001; AC: 74.0%, p<0.0001). \nConclusion: 3D CEST imaging is considered at least as valuable as DWI and \ncan improve capability for differentiation of malig nant from benign prostatic \nareas with DWI. \nLimitations: LImited study population \nFunding for this study: Research grants from Canon Medical Systems \nEthics committee - additional information: Fujita Health University Hospital \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Smoking Re search Foundation \nResearch/Grant Support: Grant-in-Aid for Scientific  Research from the \nJapanese Ministry of Education, Culture, Sports, Sc ience and Technology \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grant-in-Aid  for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Canon Medica l Systems \nCorporation \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Smoking Res earch Foundation \nResearch/Grant Support: Canon Medical Systems Corpo ration \n \n \nThe comparison between virtual non-contrast and tru e non-contrast \nimaging of adrenal masses on photon-counting CT \n*X. Bai*¹, G. Zhang¹, X. Zhang¹, J. Zhang¹, L. Chen ¹, Q. Peng¹, Z. Lin²,  \nH. Sun¹, Z. Jin¹; ¹Beijing/CN, ²Shanghai/CN \n \nPurpose or Learning Objective: To investigate the differences in CT \nattenuation and radiomics features between virtual non-contrast (VNC) and \ntrue non-contrast (TNC) of adrenal masses on photon -counting CT (PCCT). \nMethods or Background: Patients with adrenal masses who underwent \nunenhanced and portal-venous-phase PCCT were includ ed. Image \nreconstructions of portal-venous phase were perform ed, including \nConventional VNC (VNCconv) and PureCalcium VNC (VNC pc) algorithms. For \ntwo dimensional (2D) measurements, we measured CT a ttenuation of adrenal \nmass at the largest slice on TNC, VNCconv, and VNCp c images, respectively. \nSemiautomatic segmentations of adrenal masses were performed to extract \nthree-dimensional (3D) CT attenuation and radiomics  features on TNC and \nVNC. The paired t-test, Bland–Altman plots and intr aclass correlation efficient \n(ICC) were used for statistical analyses. \nResults or Findings: The study consisted of 54 patients (27 female, mean  \nage 45.3 years) with 68 adrenal lesions. CT attenua tion on VNCconv and \nVNCpc was higher than that on TNC (Mean differences , 2D: 8.03 HU for \nVNCconv and 5.76 HU for VNCpc, 3D: 8.75 HU for VNCc onv and 6.89 HU for \nVNCpc). The proportion of lipid-rich adenomas (TNC < 10 HU) correctly \nclassified by VNCconv and VNCpc was 26.1% (6/23) an d 39.1% (9/23), \nrespectively. TNC, VNCconv, and VNCpc attenuation d id not differ significantly \nbetween 2D and 3D measurements (all P ＞ 0.05). The ICCs of first-order \nfeatures, shape features and texture features betwe en TNC and VNCconv \nwere 0.625, 0.820 and 0.591, respectively. \nConclusion: The VNC algorithms of PCCT overestimated CT attenua tion. CT \nattenuation at the largest slice can replace 3D att enuation. VNC and TNC have \nexcellent agreement on shape features and average a greement on first-order \nfeatures and texture features. \nLimitations: The limitations of the study are the study sample i s relatively \nsmall and the retrospective study may have a select ion bias. \nFunding for this study: Funding was provided by the National High-Level \nHospital Clinical Research Funding (2022-PUMCH-B-06 9, 2022-PUMCH-A-\n033 and 2022-PUMCH-A-035), CAMS Innovation Fund for  Medical Sciences \n(2022-I2M-C&T-B-019) and Beijing Municipal Natural Science Foundation \n(L232133). \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board (No. 23PJ1487). \nAuthor Disclosures:  \nJiahui Zhang: Nothing to disclose \nGumuyang Zhang: Nothing to disclose \nHao Sun: Nothing to disclose \nXin Bai: Nothing to disclose \nQianyu Peng: Nothing to disclose \nZijing Lin: Nothing to disclose \nLi Chen: Nothing to disclose \nZhengyu Jin: Nothing to disclose \nXiaoxiao Zhang: Nothing to disclose \n08:00-09:00 Research Stage 2 \nResearch Presentation Session: Cardiac \nRPS 1703 \nCardiac imaging: interactions with other \norgans and systemic diseases \n \nModerator \nP. Krumm; Tübingen/DE  \nAuthor Disclosures:  \nPatrick Krumm: Research Grant/Support: Spimed AI, S iemens Healthcare; \nSpeaker: Bayer Healthcare, Siemens Healthineers, Br acco \n \n \nAssociation between pericoronary fat attenuation in dex, fractional flow \nreserve and brain white matter hyperintensity: a ca se control study \n*J. Qin*, Y. Xu; Nanjing, Jiangsu/CN \n \nPurpose or Learning Objective: To explore the association between coronary \ncomputed tomography angiography (CCTA)- derived per icoronary fat \nattenuation index (pFAI), fractional flow reserve ( CT-FFR) and degree of white \nmatter hyperintensities (WMH). \nMethods or Background: Clinical, CCTA and brain magnetic resonance \nimaging (MRI) data of 561 participants were retrosp ectively analyzed. WMH \nwere assessed in periventricular (PVWMH) and deep ( DWMH) locations, and a \ntotal Fazekas score was calculated by summing the s cores for PVWMH and \nDWMH. The study cohort was classified into mild WMH  group (score 0-2) and \nmoderate-to-severe WMH group (score 3-6). Coronary artery disease (CAD) \nwas defined as one or more coronary arteries with d iameter stenosis of ≥50%. \nThe threshold value of CT-FFR was set as 0.80. Clin ical data, pFAI, CT-FFR \nand other coronary parameters were compared between  two groups, and \nindependent variables associated with moderate-to-s evere WMH were \nidentified using multiple logistic regression analy sis. \nResults or Findings: Compared with patients with mild WMH, those with \nmoderate-to-severe WMH showed larger volume of plaq ue (total, calcified, \nnoncalcified palque), higher plaque burden, longer plaque length, higher \nAgatston Score, higher value of pFAI and higher pro portion of CT-FFR≤ 0.80, \nhigher proportion of patients with CAD and aortic u lcers (all p<0.05). Multiple \nlogistic regression indicated that age [odds ratio (OR), 1.028; p=0.025], CAD \n(OR, 5.282; p=0.004), plaque burden (OR, 4.101; p=0 .004) , noncalcified \nplaque burden (OR, 2.850; p<0.001) and pFAI (OR, 1. 109; p<0.001) were \nindependently associated with moderate-to-severe WM H. \nConclusion: Besides the well-known factors including age, CAD a nd plaque \nburden, pFAI was also found to be associated with m oderate-to-severe WMH. \nLimitations: As a cross-sectional study, the progression of WMH was not \nanalyzed. Future research exploring the progression  of CCTA- derived \nparameters including CT-FFR and pFAI and the progre ssion of WMH would be \nmore valuable. \nFunding for this study: This work was supported by the grants from the \nNatural Scientific Foundation of China (Grant Nos. 82302163 for Yunfei Wang) \nand Young Scholars Fostering Fund fo the First Affi liated Hospital of Nanjing \nMedical University (Grant Nos. PY2022036 for Yunfei  Wang). \nEthics committee - additional information: The requirement of written \ninformed consent was waived due to the retrospectiv e nature. \nAuthor Disclosures:  \nJie Qin: Nothing to disclose \nYi Xu: Nothing to disclose \n \n \nQuantitative T1 mapping for the evaluation of the i ron overload in \nhereditary hemochromatosis \n*G. Prencipe*, I. Notarangelo, P. Mangano, L. Marin elli, L. Macarini,  \nG. Guglielmi, M. Gravina; Foggia/IT \n(gianluca.prencipe.12@gmail.com) \n \nPurpose or Learning Objective: The aim of this study was to evaluate the \nefficacy of cardiac T1 mapping MRI sequences in ass essing myocardial iron \noverload in patients with Hereditary Hemochromatosi s, in comparison to the \nmore commonly employed T2* sequences. \nMethods or Background: A total of 28 cardiac MRI scans, conducted \nbetween 2019 and 2023, were analyzed. All patients had Hereditary \nHemochromatosis and elevated serum ferritin levels.  The MRI scans were \nperformed using a Philips Achieva dStream 1.5T scan ner with cardiac gating \nand included T2* and native T1 mapping sequences, w ithout the use of \ncontrast agents. Quantitative T2* analysis was cond ucted by placing regions of \ninterest (ROIs) in the interventricular septum (IVS ), and these results were \ncompared with the corresponding native T1 mapping f indings. \n\n \n \nSaturday \nAbstract-based Programme \n \n 197  \nResults or Findings: Out of the 28 patients, 25 had T2* values over 20 m sec, \nindicating no significant iron deposition. The rema ining three patients exhibited \nT2* values between 15 and 20 msec, suggesting mild iron overload. None of \nthe patients demonstrated T2* values below 15 msec.  In the three patients with \nmild iron overload, T1 mapping showed corresponding ly low values, consistent \nwith their T2* measurements. However, 5 of the 25 p atients with normal T2* \nvalues had T1 values below the expected range (980- 1080 msec for the our \nscanner). \nConclusion: Native T1 values were reduced in patients with myoc ardial iron \naccumulation and correlated well with T2* measureme nts. T1 mapping \nprovides the added benefit of detecting early-stage  iron overload that might be \nmissed by T2* alone, making it a valuable tool for the early diagnosis of iron \ndeposition and for monitoring the effectiveness of chelation therapy \nLimitations: The study is retrospective. Single center study. Fe w patients with \nlow T2*. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective \nAuthor Disclosures:  \nLuca Marinelli: Nothing to disclose \nGiuseppe Guglielmi: Nothing to disclose \nIlenia Notarangelo: Nothing to disclose \nMatteo Gravina: Nothing to disclose \nGianluca Prencipe: Nothing to disclose \nPaola Mangano: Nothing to disclose \nLuca Macarini: Nothing to disclose \n \n \nSex-Specific Prognostic Value of Opportunistic Epic ardial Adipose \nTissue Quantification on Lung Cancer Screening Ches t CT \n*E. Norton*, J. M. Brendel, I. Hadzic, T. Mayrhofer , I. L. Langenbach,  \nM. C. Langenbach, M. T. Lu, H. Aerts, B. Foldyna; B oston, MA/US \n(eanorton@mgh.harvard.edu) \n \nPurpose or Learning Objective: To evaluate the prognostic value of \nepicardial adipose tissue (EAT) volume and density between women and men \n(who often present with different body fat distribu tion) eligible for lung cancer \nscreening, a group with an unmet need for better ri sk stratification. \nMethods or Background: Using a validated deep-learning algorithm, we \nmeasured EAT volume (indexed to body-surface-area; cm³/m²) and density \n(HU) on baseline non-contrast, low-dose chest CTs f rom the National Lung \nScreening Trial. Associations with 12-year all-caus e and cardiovascular \nmortality were assessed using multivariable Cox reg ression models, stratified \nby sex, and adjusted for age, race/ethnicity, smoki ng status (current vs. \nformer), pack-years, history of heart disease or st roke, diabetes mellitus, \nhypertension, educational status, and body-mass-ind ex. \nResults or Findings: Among 24,090 participants, 9,886 were women (41%; \n61±5 years), and 14,204 were men (59%; 62±5 years). Women presented \nlower EAT volumes and higher densities than men (65 .7±23.2cm³/m² and -\n77.2±5.1HU vs. 73.5±25.0cm³/m² and -78.0±5.2HU; p<0.001 for sex \ndifferences). A 10 cm³/m² EAT volume increase revea led similar prognostic \nvalues in women and men (all-cause mortality: aHR:1 .10 [95%-CI: 1.07–1.13] \nin women vs. aHR:1.11 [95%CI: 1.08–1.13] in men; ca rdiovascular mortality: \naHR:1.13 [95%-CI: 1.06–1.20] in women vs. aHR:1.16 [95%-CI: 1.11–1.21] in \nmen). EAT density demonstrated a stronger associati on with all-cause mortality \nin men than in women (aHR:1.41 [95%-CI: 1.28–1.56] in men vs. aHR:1.28 \n[95%-CI: 1.10–1.47] in women; p=0.010 for sex as in teraction term). The \nassociations were similar with cardiovascular death  (aHR:1.76 [95%-CI: 1.29–\n2.41] in women vs. aHR:1.78 [95%-CI: 1.45–2.19] in men). \nConclusion: Opportunistic EAT volume and density quantification  may \nimprove risk stratification in women and men eligib le for lung cancer screening, \nwith EAT density being a stronger predictor of all- cause death in men than \nwomen. \nLimitations: Exclusively heavy smokers were included, precluding  comparison \nwith non-smokers and lower-risk smokers. \nFunding for this study: Original data collection for NLST was funded by NIH . \nEthics committee - additional information: Insight IRB protocol #: \n2017P002400 \nAuthor Disclosures:  \nJan Michael Brendel: Nothing to disclose \nIsabel Luisa Langenbach: Nothing to disclose \nEmilia Norton: Nothing to disclose \nHugo Aerts: Nothing to disclose \nIbrahim Hadzic: Nothing to disclose \nThomas Mayrhofer: Nothing to disclose \nBorek Foldyna: Nothing to disclose \nMarcel Christian Langenbach: Nothing to disclose \nMichael T. Lu: Nothing to disclose \n \n \n \n \n \nCardiovascular Magnetic Resonance Features in Cirrh otic Patients with \nTransjugular Intrahepatic Portosystemic Shunt \n*J. Arenja*, D. G. Aydemir, L. Naimi, I. Molwitz, M . Sterneck, G. Adam,  \nP. Bannas, E. Tahir, J. Erley; Hamburg/DE \n \nPurpose or Learning Objective: A transjugular intrahepatic portosystemic \nshunt (TIPS) results in blood flow from the splanch nic to the venous circulation. \nWe aimed to evaluate if a TIPS impacts cardiovascul ar magnetic resonance \n(CMR) features in patients with liver cirrhosis. \nMethods or Background: In this retrospective, monocentric study, 60 patien ts \nwith liver cirrhosis received a CMR exam (3T, Ingen ia, Philips or Siemens). Left \nventricular (LV) mass, volumes (indexed to the body  mass index [BMI]), and \nejection fraction (EF) were analyzed. Strain was me asured by feature tracking \nusing Cvi42 (Circle Vascular Imaging). For statisti cs, mixed linear models were \nconducted. \nResults or Findings: 30 patients with TIPS (40% females, age 56±11 years , \nBMI 26.8±5.2 kg/m2, hepatic venous pressure gradien t (HVPG) pre-TIPS \n27.9±6.7 and post-TIPS 11.2±4.1 mmHg) were compared to 30 matched \npatients without TIPS (47% females, age 53±14 years , BMI 26.1±5.1 kg/m2). \nPatients with TIPS showed a -14.8% attenuated LV gl obal radial strain [95% \nconfidence interval (CI): -28.1 to -1.5%] compared to patients without TIPS (p = \n0.037). LV mass, volumes, and longitudinal/circumfe rential strain were not \nsignificantly different between the groups. In pati ents with TIPS, LV end-\ndiastolic mass (regression coefficient = 1.5 ml/m2 [95%-CI: 0.1 to 3%], p = \n0.040) and LV global longitudinal strain (0.6% [95% -CI: 0.1 to 1.2%], p = 0.035) \nwere associated with HPVG post-TIPS. \nConclusion: Cirrhotic patients with TIPS show attenuated LV rad ial strain \ncompared to patients without TIPS, possibly reflect ing subclinical LV \ndysfunction due to TIPS-induced increase in preload . A higher post-TIPS \nHPVG is associated with increased LV end-diastolic mass and attenuated LV \nglobal longitudinal strain, indicating LV hypertrop hy and subclinical dysfunction \nwith persistent portal hypertension. \nLimitations: This is a retrospective study with a small sample s ize, warranting \nvalidation in a larger cohort. \nFunding for this study: Not applicable. \nEthics committee - additional information: The study was approved by the \nlocal ethics committee of the medical association i n Hamburg. \nAuthor Disclosures:  \nGerhard Adam: Nothing to disclose \nLieda Naimi: Nothing to disclose \nIsabel Molwitz: Nothing to disclose \nMartina Sterneck: Nothing to disclose \nPeter Bannas: Nothing to disclose \nJennifer Erley: Nothing to disclose \nJennis Arenja: Nothing to disclose \nDestina Gizem Aydemir: Nothing to disclose \nEnver Tahir: Nothing to disclose \n \n \nArrhythmic Burden, Myocardial Markers, and Long-ter m Survival in \nDistinct Cardiac Magnetic Resonance Subsets of Syst emic Sclerosis \n*E. Moliterno*¹, L. Giarletta¹, G. Rovere¹, S. L. B osello¹, G. De Luca²,  \nA. Tonutti³, M. A. D'Agostino¹, L. Natale¹, R. Mara no¹; ¹Rome/IT, ²Milan/IT, \n³Rozzano/IT \n(eleonora.moliterno@gmail.com) \n \nPurpose or Learning Objective: Cardiac involvement in Systemic Sclerosis \n(SSc) is widely recognized as heterogeneous and, wh en clinically evident, it is \nassociated with a poor prognosis. Recently, 5 cardi ac magnetic resonance \n(CMR) phenotypes in SSc have been identified (Knigh t DS et al. European \nHeart Journal 2023). These phenotypes do not align with the existing clinical \nsubgroup classifications or autoantibody statuses, yet each has a distinct 5-\nyear prognosis. Our objective is to test the long-t erm prognostic significance of \nthis classification system in an external cohort an d to compare ECG Holter \nmonitor parameters, NT-proBNP and troponin T levels  and 10-year survival \noutcomes across these groups. \nMethods or Background: CMR assessments were conducted in 3 Italian \ntertiary centers on 143 consecutive SSc patients wh o presented symptoms of \ndyspnea, palpitations or chest pain. Based on the C MR findings, patients were \ncategorized into 5 distinct groups: dilated right h earts with right ventricular \nfailure (RVF); biventricular failure with dilatatio n and dysfunction (BVF), normal \nfunction with average cavity (NF-AC), small cavity (NF-SC), and large cavity \n(NF-LC). \nResults or Findings: The distributions for NF-AC, NF-SC, NF-LC, BVF, and  \nRVF were 46.2%, 22.4%, 14.0%, 14.0%, and 3.5%, resp ectively. Proportions \nof male patients and pulmonary function tests showe d statistically significant \ndifferences across the subsets. Troponin T and NT-p roBNP values were similar \nacross all subsets. The NF-LC and RVF groups exhibi ted Left Bundle Branch \nBlock (LBB) and ventricular ectopic beats (VEB) mor e frequently compared to \nother groups. There was a variation in 10-year surv ival rates across the \ngroups, with patients in the RVF, NF-LC, and BVF ca tegories showing poorer \nprognosis. \n\n \n \nSaturday \nAbstract-based Programme \n \n 198  \nConclusion: This data confirms the prognostic value of the prop osed CMR \nsubsets in an another European SSc cohort, highligh ting that subsets with \npoorer prognosis are associated with a higher arrhy thmic burden. \nLimitations: No \nFunding for this study: None \nEthics committee - additional information: No additional information \nAuthor Disclosures:  \nSilvia Laura Bosello: Nothing to disclose \nGiacomo De Luca: Nothing to disclose \nMaria Antonietta D'Agostino: Nothing to disclose \nEleonora Moliterno: Nothing to disclose \nGiuseppe Rovere: Nothing to disclose \nAntonio Tonutti: Nothing to disclose \nLuigi Natale: Nothing to disclose \nRiccardo Marano: Nothing to disclose \nLorenzo Giarletta: Nothing to disclose \n \n \nEvaluating the Relationship Between Systemic Inflam mation Index (SII), \nSystemic Inflammation Response Index (SIRI), and Co ronary Calcium \n(Ca) Scoring in Atherosclerotic Cardiovascular Dise ase (ASCVD) \nI. T. Rakıcı, *K. F. Kaldırımoğlu*, A. S. Mahmutoğlu; Istanbul/TR \n(kemal.kaldirimoglu@gmail.com) \n \nPurpose or Learning Objective: Atherosclerosis is the main cause of \ncoronary artery disease, with inflammation as a key  factor. Studies link SII (SII \n= PlateletxNeutrophil/ Lymphocyte) and SIRI (SIRI =  NeutrophilxMonocyte/ \nLymphocyte) to inflamatory diseases and cardiovascu lar outcomes. This study \nexamines the relationship between these markers, Ca  scoring, and ASCVD. \nMethods or Background: Images of 460 patients who underwent coronary \nCTA and calcium scoring were analyzed and divided i nto five groups: normal \ncalcium score, calcium score 1-100, calcium score > 100, calcium score of 0 \nwith hypodense plaques, and calcium score of 0 with  myocardial bridging. The \nSIRI and SII were calculated to assess significant differences and establish \ncutoff values. \nResults or Findings: In the group with a calcium score greater than 100,  the \nSIRI value was significantly higher than in all oth er groups, indicating an \ninflammatory process in ASCVD. A notable difference  in SIRI values was \nobserved between calcium scores of 1-100 and above 100, suggesting a \ncorrelation between higher scores and elevated infl ammatory markers. In \ncases with a calcium score of 0 and myocardial brid ging, no significant \ndifferences in inflammatory markers were found, sug gesting no increase in the \ninflammatory process. The SII value showed no signi ficant differences (p>0.05) \namong the other groups, except that it was signific antly higher (p<0.05) in the \ncalcium score > 100 and 1-100 groups compared to th e calcium score 0 group \nwith myocardial bridging. No significant difference s in SIRI and SII values were \nobserved in cases with a calcium score of 0 and hyp odense plaques, indicating \nSIRI and SII values were insufficient to detect inf lammation in these cases. \nConclusion: SIRI effectively discriminates between calcium scor e levels, \nwhereas SII is not a reliable indicator. Future res earch is needed to determine \nclinically relevant cutoff values in this area. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by the \nUniversity of Health Sciences Istanbul Training and  Research Hospital Clinical \nResearch Ethics Committee (Decision number: 91). \nAuthor Disclosures:  \nAbdullah Soydan Mahmutoğlu: Nothing to disclose \nIbrahim Taşkın Rakıcı: Nothing to disclose \nKemal Furkan Kaldırımoğlu: Nothing to disclose \n \n \nMyocardial strain trajectories by cardiac magnetic resonance and its \nrelationship with myocardial fibrosis in Duchenne m uscular dystrophy \nassociated cardiomyopathy \n*H. Xu*, X. Ting; Cheng du/CN \n(xuhuayan2022.09_@scu.edu.cn) \n \nPurpose or Learning Objective: This study aims to explore the relationship \nbetween Myocardial strain trajectories and myocardi al fibrosis by cardiac \nmagnetic resonance (CMR) in Duchenne Muscular Dystr ophy (DMD). \nMethods or Background: DMD is a severe, X-linked genetic disorder \ncharacterized by progressive degeneration of skelet al muscle due to mutations \nin the dystrophin gene. And cardiomyopathy is becom ing the leading cause of \ndeath. Cardiac magnetic resonance (CMR) and late ga dolinium enhancement \n(LGE) are important tools in recognizing myocardial  involvement, myocardial \nstrain assessed by CMR feature tracking imaging (FT ) may demonstrate early \nfunctional changes. Trajectories is a new method us ed to explore the dynamic \nchanges of disease progression. Dynamic changes of myocardial strain using \ntrajectories and myocardial tissue characteristics has not been investigated.  \n \nWe obtained myocardial strain parameters of more th an 2 times by CMR FT \nmethod on 110 DMD patients. Global circumferential strain data was used to \nconstruct trajectory model. Late gadolinium enhance ment progression was \nmeasured. Group-based trajectory modeling (GBTM) wa s performed to detect \nthe trajectories of FT. Data t tests or one-way ANO VA adjusting for multiple \ncomparisons. \nResults or Findings: Three circumferential strain FT trajectory groups w ere \nidentified as: up-down FT trajectory group(N=14), F T trajectory group(N=32), \nthe steady FT trajectory group (N= 64). The occurre nce rate of adverse cardiac \nevent in the three trajectory analysis groups were significantly \ndifferent(P=0.03). \nConclusion: This study is the first to analyze the myocardial s train trajectory of \nDMD patients. It provides further evidence of the c orrelation between FT and \nprogression of myocardial fibrosis. \nLimitations: The relatively small number of patients assigned to  the 3 groups \nmay result in insufficient statistical power. \nFunding for this study: National funding of china \nEthics committee - additional information: Ethics committee of west china \nsecond university hospital \nAuthor Disclosures:  \nXu Ting: Nothing to disclose \nHuayan Xu: Nothing to disclose \n \n \n08:00-09:00 Research Stage 3 \nResearch Presentation Session: Neuro \nRPS 1711 \nFoetal and paediatric neuroimaging: \nunveiling the youthful brain \n \nModerator \nD. Zlatareva; Sofia/BG  \n(dorazlat@yahoo.com) \n \n \nEven Low Levels of Prenatal Alcohol Exposure Can In duce Structural \nAlterations in Fetal Temporal Gyrification \n*P. Kienast*, M. Stuempflen, J. Tischer, A. Taymour tash, G. Langs, D. Prayer, \nG. Kasprian; Vienna/AT \n(patric.kienast@meduniwien.ac.at) \n \nPurpose or Learning Objective: Around 10-20% of women in Europe \nconsume alcohol during pregnancy, leading to negati ve neurodevelopmental \noutcomes such as fetal alcohol spectrum disorders ( FASD). Although the \nharmful impact of alcohol on neuronal development d uring pregnancy is well-\nestablished, studies on its effects during the pren atal phase remain limited. \nThis study evaluates prenatal changes in fetal gyri fication resulting from \nmaternal alcohol consumption using automated surfac e measurements from \nfetal MRI data. \nMethods or Background: This study involved 500 fetal MRI examinations \nwhere mothers were asked about their alcohol consum ption habits using tools \nlike PRAMS and TACE. The brains were denoised, moti on-corrected, and \nsegmented for 3D reconstruction. Gyrification indic es were calculated, and \nasymmetry of the cerebral cortex was quantified. Ne urotypical alcohol-affected \ncases were compared with healthy control cases. \nResults or Findings: The study finally included 22 alcohol-exposed fetus es \n(mean gestational-age [GW] 27.61 ± 3.94 weeks) and 22 non-alcohol-exposed \ncontrol fetuses (mean GW 27.57 ± 3.94 weeks), match ed in a 1:1 ratio. Of the \nmothers in the alcohol-exposed group, 17 consumed o nly small amounts of \nalcohol (<14 grams per week). The typical asymmetry  observed in the temporal \nlobes of the fetal brain was significantly diminish ed (p=0.48, 95% CI -2.24 to -\n0.01) in fetuses exposed to alcohol. \nConclusion: Alcohol consumption during pregnancy impacts the de velopment \nof the fetal brain, with alterations observable in temporal brain asymmetry \nthrough fetal MRI. These findings align with studie s linking prenatal brain \nasymmetry to language development in childhood, whi ch is frequently impaired \nin children diagnosed with FASD. These structural c hanges are found even in \ncases of small alcohol consumption. \nLimitations: A potential limitation of the study is the relative ly small sample \nsize, which may limit the ability to detect more su btle effects of low-level \nalcohol exposure on fetal brain development. \nFunding for this study: The sponsor is the Medical University of Vienna. \nEthics committee - additional information: This study was approved by the \nlocal IRB. \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 199  \nAuthor Disclosures:  \nGeorg Langs: Nothing to disclose \nAthena Taymourtash: Nothing to disclose \nMarlene Stuempflen: Nothing to disclose \nJohannes Tischer: Nothing to disclose \nDaniela Prayer: Nothing to disclose \nPatric Kienast: Nothing to disclose \nGregor Kasprian: Nothing to disclose \n \n \nIntracranial Volumes of Preterm Born Infants and Ag e-matched Fetuses: \nA Comparison of In Utero versus Ex Utero Conditions  \n*G. Abaci*, G. Biechele, S. Schläger, F. Obereisenb uchner, K. Förster,  \nA. Flemmer, J. Ricke, A. Hilgendorff, S. Stöcklein;  Munich/DE \n(gizem.abaci@med.uni-muenchen.de) \n \nPurpose or Learning Objective: The aim of our study was to compare \nintracranial volume (ICV), brain volume (BV), and t he volumes of the inner \ncerebral fluid (iCSF) and outer cerebral fluid spac es (oCSF) of preterm brorn \ninfants with age-matched fetuses and to assess the impact of postnatal \ntherapies including ventilation. \nMethods or Background: Preterm born infants and fetuses were matched for \ngestational age (± 4 days), ranging from 32 to 39 weeks' gestation (GW). 26 \npreterm infants and 26 age-matched fetuses without reported brain pathology \nunderwent MRI. Anatomical T2-weighted brain images were analyzed by \nmanual segmentation. \nResults or Findings: Preterm infants were characterized by reduced ICV ( p= \n0.005), BV (p= 0.007) and oCSF volumes (p= 0.074) w hen compared to age-\nmatched fetuses undergoing prenatal imaging, wherea s ICV-corrected BV \n(BV/ICV, p= 0.385), oCSF (oCSF/ICV, p= 0.568) and t otal CSF (total CSF/ICV, \np= 0.274) did not differ significantly between the two groups. When considering \npostnatal therapies, more days of noninvasive posit ive pressure ventilation \nwere associated with higher corrected values for oC SF (oCSF/ICV, p= 0.0392), \nand total CSF (total CSF/ICV, p= 0.0282) as well as  lower corrected values for \nBV (BV/ICV, p= 0.0282). \nConclusion: Preterm infants showed reduced ICV, BV and oCSF vol umes \ncompared to age-matched fetuses in utero. CSF volum es in preterm infants \nwere impacted by postnatal therapy, potentially ind iciating that intrathoracic \npressure might influence venous return and CSF reab sorption into the venous \nsystem. In conclusion, brain and CSF volumes are im pacted by prematurity \nand associated therapeutic strategies. Potential me chanisms underlying these \nefffects and their implications for ex utero versus  in utero brain development \nneed to be further investigated. \nLimitations: Retrospective design \nFunding for this study: None \nEthics committee - additional information: Lmu 207-33 \nAuthor Disclosures:  \nAnne Hilgendorff: Nothing to disclose  \nAndreas Flemmer: Nothing to disclose \nGizem Abaci: Nothing to disclose \nSarah Schläger: Nothing to disclose \nKai Förster: Nothing to disclose \nSophia Stöcklein: Nothing to disclose \nFlorian Obereisenbuchner: Nothing to disclose \nGloria Biechele: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \nMachine Learning Analysis in Diffusion Kurtosis Ima ging for \nDiscriminating Pediatric Posterior Fossa Tumors \n*I. P. Voicu*, C. D'Orazio, E. Piccirilli, F. Dotta , P. Toma, G. S. Colafati; \nRome/IT \n(paul.voicu@hotmail.it) \n \nPurpose or Learning Objective: Differentiating pediatric posterior fossa (PF) \ntumors medulloblastoma (MB), ependymoma (EP) and pi locytic astrocytoma \n(PA) remains relevant, because of prognostic implic ations. Diffusion kurtosis \nimaging (DKI) has not been investigated for pediatr ic PF tumors. Whole-tumor \nbased (VOI) segmentations may improve repeatability  compared to \nconventional region-of-interest (ROI) approaches. O ur purpose was to \ncompare repeatability between ROI and VOI measureme nts and assess DKI \naccuracy in discriminating among pediatric PF tumor s with machine learning \n(ML) techniques. \nMethods or Background: We retrospectively analyzed 34 children (M, F, \nmean age 7.48 years) with PF tumors who underwent p reoperative MRI on a 3 \nTesla magnet, including DKI. For each patient, two neuroradiologists \nsegmented the whole solid tumor, the ROI of area of  maximum tumor diameter \nand a small 5 mm ROI. The automated analysis pipeli ne included inter- \n \n \n \nobserver variability (coefficient of variation- COV ), and machine learning (ML) \nanalyses. We estimated DKI accuracy with MANOVA ana lysis. We applied \nSMOTE to balance the dataset and performed a Random  Forest (RF) machine \nlearning classification analysis based on all DKI m etrics from the SMOTE \ndataset ( 70/30 for the training and testing cohort ). \nResults or Findings: Tumor histology included medulloblastoma (15), pilo cytic \nastrocytoma (14) and ependymoma (5). VOI-based meas urements presented \nlower variability than ROI-based measurements. DKI- derived metrics \ndiscriminated accurately between PF tumors. SMOTE g enerated a balanced \ndataset with 45 instances (34 original and 11 synth etic, 10 EP and 1 PA). ML \nanalysis yielded accuracy of 0.928, correctly predi cting all but one lesion in the \ntesting set. \nConclusion: VOI-based measurements presented improved repeatabi lity \ncompared to ROI-based measurements. ML techniques b ased on DKI-derived \nmetrics are useful for discrimination of pediatric PF tumors. \nLimitations: The study was approved by Institutional Review Boar d (IRB) of \nBambino Gesù Children’s Hos-pital (RAP-2024-0001). T \nFunding for this study: None \nEthics committee - additional information: The study was approved by \nInstitutional Review Board (IRB) of Bambino Gesù Ch ildren’s Hos-pital (RAP-\n2024-0001). \nAuthor Disclosures:  \nClaudia D'Orazio: Nothing to disclose \nPaolo Toma: Nothing to disclose \nFrancesco Dotta: Nothing to disclose \nIoan Paul Voicu: Nothing to disclose \nGiovanna Stefania Colafati: Nothing to disclose \nEleonora Piccirilli: Nothing to disclose \n \n \nUltra-low dose computed tomography as an alternativ e to radiographic \nshunt series in the diagnosis of mechanical ventric uloperitoneal shunt \ncomplications – an ex vivo phantom study \n*B. Yildirim*, R. J. Serger, S. Zensen, H. Styczen,  M. Schüßler, M. Forsting,  \nC. Deuschl, M. Opitz, D. Bos; Essen/DE \n(berk.yildirim@uk-essen.de) \n \nPurpose or Learning Objective: The standard modality for the diagnosis of \nventriculoperitoneal (VP) shunt failure is the radi ographic shunt series (RSS). \nHowever, ultra-low dose computed tomography (ULD-CT ) offers lower \nradiation exposure compared to RSS. The aim of this  study was to compare \nthe radiation doses of RSS and ULD-CT on photon-cou nting CT (PCCT) in the \ndiagnosis of mechanical shunt failure in human phan tom models and to \ndemonstrate the diagnostic performance of ULD-CT. \nMethods or Background: VP shunts with different mechanical complications \nwere placed on human phantom models corresponding t o ages of 1, 5, 10 and \n30 years. RSS and ULD-CT on PCCT scans based on top ograms with tube \ncurrents ranging from 10 and 55 mAs (Sn100 kV) were  performed on each \nphantom. Effective doses of RSS in pediatric phanto ms were estimated by \nusing the conversion factors of Seidenbusch (2006, 2008 and 2009). Effective \ndoses of ULD-CT were estimated according to ICRP 10 3. \nResults or Findings: ULD-CT demonstrated lower effective doses compared \nto RSS for phantoms representing ages of 5, 10 and 30 years, while \nsuccessfully detecting mechanical VP shunt complica tions in all cases. \nHowever, higher effective doses were assessed for U LD-CT scans of the 1-\nyear phantom in comparison to RSS. The effective do ses for radiographic RSS \nand ULD-CT (using the lowest dose by utilizing 10 m As topograms), \nrespectively, were as follows: 1 year: 0.056 vs. 0. 104 mSv; 5 year: 0.186 vs. \n0.092 mSv; 10 year: 0.240 vs. 0.082 mSv; 30 year: 0 .641 vs. 0.050 mSv. \nConclusion: ULD-CT is a potentially superior alternative to rad iographic shunt \nseries for the detection of mechanical VP shunt com plications in patients aged \n5 years and above, which is particularly relevant i n children due to reduction of \nthe radiation risk. \nLimitations: Phantom study. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Ethics committee approval was \nnot required. \nAuthor Disclosures:  \nMichael Forsting: Nothing to disclose \nMaximilian Schüßler: Nothing to disclose \nHanna Styczen: Nothing to disclose \nDenise Bos: Nothing to disclose \nCornelius Deuschl: Nothing to disclose \nSebastian Zensen: Nothing to disclose \nRaya Juliane Serger: Nothing to disclose \nBerk Yildirim: Nothing to disclose \nMarcel Opitz: Nothing to disclose \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 200  \nLiving in a gold mining community: Assessing 3rd tr imester estimated \nfoetal weight, prevalence of multimorbidity across body systems, and a \n6-year trend in birth outcomes \n*A. D. Piersson*¹, E. K. Effah², B. Brusah³, S. T. Quartei⁴, H. Mumuni⁵,  \nA. Akanchimayoro³, L. Jones⁶, K. Dzefi-Tettey⁴, M. R. Asamani⁷; ¹York/UK, \n²Obuasi/GH, ³Cape Coast/GH, ⁴Accra/GH, ⁵Tamale/GH, ⁶Virginia, VA/US, \n⁷Michigan, MI/US \n(a.piersson@yorksj.ac.uk) \n \nPurpose or Learning Objective: In this study, we assessed the following – \n3rd trimester estimated foetal weight (EFW), preval ence of multimorbidity \nacross body systems, and a 6-year trend in birth ou tcomes in a mining \ncommunity, exposed to Galamsey, a term used to defi ne illegal small-scale \nand artisanal gold mining. \nMethods or Background: We compared 3rd trimester estimated foetal weight \non ultrasound systems between pregnant women ( ≥ 18 years old) with \nsingleton pregnancies living in a mining (n = 181) and non-mining communities \n(n = 260). Then, we compared health problems across  body systems in \nanother two groups (mining, n = 507; non-mining, n = 127). Further, we \nanalyzed a 6-year trend in birth outcomes in the mi ning community located in \nObuasi, Ashanti Region in Ghana. \nResults or Findings: Overall, women living in the mining area showed \nsignificantly higher 3rd-trimester EFW than those l iving in a non-mining \ncommunity (mining: 2.95+/-0.66 kg; non-mining: 2.72 +/-0.04 kg; p = 0.0002). \nWomen living in the mining community showed a highe r prevalence of \nmorbidities in 8 body regions/conditions than those  living in the non-mining \ncommunity. A sub-analysis of the prevalence of the 8 body regions/conditions \namong women in the two communities showed significa nt difference (p = \n0.016). A 6-year trend in birth outcomes showed inc reasing rates of stillbirth \nvarying between 1.69% and 2.90%, a relatively low r ecord of congenital \nanomalies, and a relatively high record of newborn complications. \nConclusion: Our preliminary findings suggest a potential link b etween \nmaternal exposure to mining-related environmental f actors and adverse \nperinatal outcomes, including increased foetal grow th, a higher prevalence of \nmaternal morbidities, rising stillbirth rates, and increased newborn \ncomplications. Further study is warranted to invest igate the underlying \nmechanisms and long-term health implications for bo th mother and child. \nLimitations: Our study did not provide quantitative maternal blo od or \nenvironmental measurements of heavy metal exposures . \nFunding for this study: None \nEthics committee - additional information: Ethical Clearance - \nUCCIRB/EXT/2022/30 \nAuthor Disclosures:  \nLashell Jones: Nothing to disclose \nKlenam Dzefi-Tettey: Nothing to disclose \nBenjamin Brusah: Nothing to disclose  \nAlbert Dayor Piersson: Nothing to disclose \nAugustine Akanchimayoro: Nothing to disclose \nEmmanuel Kofi Effah: Nothing to disclose \nMercedes Rowe Asamani: Nothing to disclose \nHanifatu Mumuni: Nothing to disclose \nSarah Teiko Quartei: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n08:00-09:00 Research Stage 4 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1705 \nHuman and machine factors in artificial \nintelligence \n \nModerator \nE. Kotter; Freiburg/DE  \n(elmar.kotter@uniklinik-freiburg.de) \nAuthor Disclosures:  \nElmar Kotter: Advisory Board: contextflow; Author: thieme, springer; Share \nHolder: Siemens, Bayer; Speaker: siemens healthinee rs, bayer \n \n \nVisual acuity among participants of the European Co ngress of Radiology \n2024: should visual assessment be recommended for r adiologists? \n*T. Van Nijnatten*¹, M. Smidt¹, J. E. Wildberger¹, M. Fuchsjäger², F. J. Gilbert³, \nF. Pediconi⁴, R. G. H. Beets-Tan⁵, F. Van Den Biggelaar¹, C. Catalano⁴; \n¹Maastricht/NL, ²Graz/AT, ³Cambridge/UK, ⁴Rome/IT, ⁵Amsterdam/NL \n(Thiemovn@gmail.com) \n \nPurpose or Learning Objective: Currently there is no recommendation \nregarding visual assessment for radiologists. The a im was to evaluate visual \nacuity among participants of the European Congress of Radiology (ECR) 2024. \nMethods or Background: Participants of ECR 2024 organized by the \nEuropean Society of Radiology (February 28th-March 3rd 2024; Vienna, \nAustria) were asked to participate in an on-site vi sual assessment. Medical \nethical approval was obtained. Each participant sig ned written informed \nconsent. The assessment consisted of vision chart r eading (Sloan ETDRS \nVision Chart) at 66 cm. Afterwards, auto-refraction  was performed to determine \nrefractive error. Finally, participants re-read a d ifferent vision chart, using on-\nsite glasses to correct for refractive error. A Log MAR score of 0.0 (i.e. Snellen \nequivalent 1.0/100%) was considered an adequate vis ual acuity for radiology \nreporting. \nResults or Findings: 321 participants completed on-site visual assessmen t \n(41% (132/321) male and 59% (189/321) female). Repo rted professions were \n114 consultant or board certified radiologists (36% ), 121 radiology residents \n(38%), 24 PhD students in radiology (7%), 37 medica l students (12%), 11 \nradiographers (3%), 3 medical physicists (1%) and 1 1 others (3%). Mean age \nwas 30 years (range: 18-69). Of the 57% (182/321) p articipants who wore \nglasses/contact lenses, 171 (94%) wore glasses/cont act lenses during image \ninterpretation tasks. Among all participants, 24 pa rticipants (7.5%) did not \nachieve logMAR score of 0.0 when reading the first vision chart. After auto-\nrefraction measurements, 11 out of these 24 partici pants improved to a \nlogMAR score of at least 0.0 using on-site glasses to correct for the refractive \nerror. \nConclusion: A considerable percentage of radiologists has accur ate visual \nacuity at a radiology reporting distance of 66 cm. Yet, 7.5% of the participants \nof the on-site visual assessment did not achieve an  adequate vision score. \nVisual assessment could be considered among radiolo gists. \nLimitations: N/a \nFunding for this study: N/a \nEthics committee - additional information: Metc 2023-0249. \nAuthor Disclosures:  \nMarjolein Smidt: Nothing to disclose \nThiemo Van Nijnatten: Nothing to disclose \nMichael Fuchsjäger: Nothing to disclose \nFiona J. Gilbert: Nothing to disclose \nJoachim E. Wildberger: Nothing to disclose \nRegina G. H. Beets-Tan: Nothing to disclose \nFrank Van Den Biggelaar: Nothing to disclose \nFederica Pediconi: Nothing to disclose \nCarlo Catalano: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 201  \nExploring how AI influences human gaze behaviour du ring \nmammography reading \nA. Taib¹, G. Partridge¹, P. Phillips², J. James¹, * Y. Chen*¹; ¹Nottingham/UK, \n²Lancaster/UK \n(yan.chen@nottingham.ac.uk) \n \nPurpose or Learning Objective: Most studies assess artificial intelligence’s \n(AI) diagnostic performance in mammography, but few  examine its impact on \nhuman reader behaviour and decision-making. The aim  was to investigate the \ninfluence of AI prompts on human performance, visua l search patterns, reader \nconfidence and look for any interaction with reader s individual personality traits \nwhen reading standard 2D screening mammograms. \nMethods or Background: In this paired reader study, eight readers working in \nthe UK breast screening programme evaluated a set o f 60 anonymised \nmammograms with and without AI (Lunit Insight MMG).  Cases with false \nnegative and false positive AI prompts were incorpo rated into the test set \ncontaining a mix of normal, benign and malignant ca ses. Readers initially \nassessed the mammograms without AI while their visu al search behaviour was \nmonitored using eye-tracking equipment (SmartEyePro ). After a six-week \nwashout period, readers reviewed the same cases wit h the addition of AI with \neye tracking. For each read, clinical opinion was r ecorded using a scale (1-\nnormal, 2-benign, 3-indeterminate, 4-suspicious, 5- malignant) and entered \nonto the Personal Performance in Mammographic Scree ning (PERFORMS) \nwebsite. Each reader completed a specially designed  psychological \nquestionnaire. \nResults or Findings: A paired analysis at breast level, using pathologic al data \nas the ‘ground-truth’, determined how correct and i ncorrect AI prompts \ninfluenced diagnostic accuracy, gaze behaviour and reader confidence. The \neffect of different reader personality traits was a lso correlated with these \noutcomes. \nConclusion: There is little evidence exploring how AI influence s a reader’s \nvisual search patterns during mammography interpret ation. This pilot study \nprovides an insight into changes in reader behaviou r when using AI and will \nhelp guide further studies and recommendations on h ow radiologists should \ninteract with AI when interpreting screening mammog raphy. \nLimitations: The limited sample of human readers may lead to typ e two error. \nFunding for this study: By Lunit. \nEthics committee - additional information: The study was approved by the \ninstitutional review board. \nAuthor Disclosures:  \nGeorge Partridge: Nothing to disclose \nPeter Phillips: Nothing to disclose \nAdnan Taib: Nothing to disclose \nJonathan James: Nothing to disclose \nYan Chen: Nothing to disclose \n \n \nNarrow AI as Double Edged Sword: effects of using A I for fracture \ndetection on distributing attention among focal and  peripheral tasks \n*F. Mol*¹, D. Pourhassan Gilkalaye¹, M. H. Rezazade  Mehrizi¹, W. Grootjans²; \n¹Amsterdam/NL, ²Leiden/NL \n(f.p.j.mol@gmail.com) \n \nPurpose or Learning Objective: Exploring the impact of using AI on \ndistributing attention among focal tasks (diagnosis ) versus peripheral tasks \n(detection of incidental findings) in shoulder radi ographs by radiographers. \nMethods or Background: 17 radiographers evaluated 255 shoulder \nradiographs from 15 outpatient trauma patients, wit h fracture detection as \nprimary task. To assess impact of AI assistance, 17 0 cases were analysed \nusing commercially available AI software for fractu re detection (Gleamer). Both \nfracture (204) and non-fracture (51) cases were inc luded. Additionally, 102 \ncases had incidental findings (e.g., pulmonary nodu les, bone cysts, rotator cuff \ncalcifications). Eye-tracking (Tobii 5) and mouse-t racking (in-house) software \nwere used to measure attention distribution. Data a re presented as mean ± \nstandard deviation, and statistical differences wer e assessed using Wilcoxon \nsigned-rank test, with significance defined as p<0. 05. \nResults or Findings: Participants spent more time on cases with AI \nassistance (n=170), averaging 47±28.24 seconds, com pared to 33 ± 28.24 \nseconds without AI (p < .001). Mouse clicks were 3. 85 ± 10.24 without AI, and \n12.09 ± 10.24 with AI (p < .001). Eye-tracking data indicated greater attention \nto peripheral tasks with AI assistance (p = 0.024),  while fracture detection was \nhigher without AI (p = 0.038). Radiographers mentio ned 31% of incidental \nfindings. \nConclusion: Use of narrow AI tools can increase sensitivity tow ards peripheral \ntasks, possibly resulting from enhanced cognitive a vailability by delegating the \nfocal task to AI. Similarly, measured decrease in f racture detection with AI \nindicates decrease in attention towards the focal t ask. This positions narrow AI \nas a “double-edged sword”: while automation can fre e up cognitive resources, \nit can lead to over-reliance and reduced attention to focal tasks. \nLimitations: Real-world clinical environment may not be simulate d completely \nin experimental setting. Eye and mouse tracking may  not capture all attention \ndistribution aspects. \nFunding for this study: N.a. \nEthics committee - additional information: N.a. \nAuthor Disclosures:  \nDorsa Pourhassan Gilkalaye: Nothing to disclose \nFerdinand Mol: Nothing to disclose \nWillem Grootjans: Nothing to disclose \nMohammad Hosein Rezazade Mehrizi: Nothing to disclo se \n \n \nColour map recommendations for MR relaxometry \n*B. D. Wichtmann*¹, M. Fuderer², N. Desouza³, F. Cr ameri⁴, V. Gulani⁵,  \nN. Sollmann⁶, S. Weingärtner⁷, S. Mandija², X. Golay³; ¹Bonn/DE, ²Utrecht/NL, \n³London/UK, ⁴Bern/CH, ⁵Ann Arbor, MI/US, ⁶Ulm/DE, ⁷Delft/NL \n(barbara.wichtmann@ukbonn.de) \n \nPurpose or Learning Objective: Quantitative imaging data may be colour \ncoded and represented as a colour-map. However, com monly used schemes \n(e.g. rainbow, jet) lack perceptual uniformity, hav e the brightest colour mid-\nrange and are not usable by colour-blind individual s. Furthermore, lack of \nstandardization of colour-maps, makes comparisons a cross studies and \ninstitutions difficult and misleading. This work de scribes recently published \nrecommendations for standardisation of MR relaxomet ry colour-maps (Fuderer, \nMRM 2024) in order to promote their adoption and dr ive the process for other \nbiomarkers. \nMethods or Background: Recommendations were generated in 4 Delphi \nrounds. A multidisciplinary committee devised quest ions on key colour-map \nfeatures, including the colour choice for T1/T2 map s, even colour gradient \ncontrast, high overall colour and lightness contras t, intuitive and constant \ngradient magnitude, and recognizability. Questions were circulated to the \nISMRM quantitative imaging group and European subsp ecialist society \nrepresentatives. Respondents received feedback afte r each round to aid \nconsensus. Responses on a 9-point Likert scale were  summarised to Agree, \nNeutral, Disagree categories. 75% consensus was the  threshold for items \nreaching recommendation. The proposed colour maps w ere based on previous \nproposals (Griswold, ISMRM 2018) but modified for p erceptual linearity and \nreadability by colour-blind people. \nResults or Findings: 58 experts responded to Round 1; 48 (45% medical, \n47% physicists) completed all 4 rounds. There was c onsensus that the \nlogarithm-processed Lipari colour-map for T1 and th e logarithm-processed \nNavia colour-map for T2 were suitable. Colour bars were deemed mandatory \nas was a specific value indicating “invalidity”. Th ere was no consensus on \nwhether to fix ranges by anatomy. \nConclusion: The logarithm-processed Lipari colour map for displ aying T1 and \nR1 values and the logarithm-processed Navia colour- map for displaying T2, \nR2, T2* and R2* are recommended for use in scientif ic reports. \nLimitations: Future work will focus on range recommendations. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: No patient-sensitive data were \nprocessed in this study. \nAuthor Disclosures:  \nNandita Desouza: Nothing to disclose \nVikas Gulani: Grant Recipient: I get research suppo rt from Siemens \nHealthineers. I have intellectual property licensed  by Siemens Healthineers. \nFabio Crameri: Nothing to disclose \nMiha Fuderer: Grant Recipient: NWO grant number 179 86, which has partly \nbeen funded by the company Philips \nSebastian Weingärtner: Nothing to disclose \nBarbara Daria Wichtmann: Speaker: I have given scie ntific presentations for \nPhilips GmbH, Lilly Deutschland, and Bender group/b .e.imaging GmbH on \nunrelated topics for which I received monetary comp ensation. \nStefano Mandija: Nothing to disclose \nXavier Golay: Consultant: Bioxydyn Founder: Gold St andard Phantoms: \nFounder, shareholder and employee \nNico Sollmann: Nothing to disclose \n \n \nEvaluating the Impact of Explainable AI on Anchorin g and Automation \nBiases in Mammography Interpretation \n*F. Pesapane*, L. Nicosia, S. Carriero, L. Mariano,  A. C. Bozzini, A. Latronico, \nL. Meneghetti, F. Abbate, E. Cassano; Milan/IT \n(Filippopesapane@gmail.com) \n \nPurpose or Learning Objective: This study investigates how AI support \ninfluences diagnostic biases (anchoring and automat ion) among radiologists \nwith varying experience levels in breast imaging. I t evaluates whether \nexplainable AI (XAI), using a heatmap, reduces thes e biases and improves \ndiagnostic accuracy. \nMethods or Background: Six radiologists (2 low experience: 0-5 years, 2 \nmedium: 5-10 years, 2 high: >10 years) participated . Each assessed 200 \nmammograms across two phases: (1) AI BIRADS score p resented before \ndiagnosis (anchoring phase), and (2) AI score prese nted after an independent \ndiagnosis (automation phase). A crossover design wa s used, with a 30% AI \n\n \n \nSaturday \nAbstract-based Programme \n \n 202  \nerror rate. Two AI conditions were tested: standard  (score only) and \nexplainable (score with heatmap). Diagnostic change s, accuracy, and bias \nfrequency were recorded, with subgroup analyses bas ed on experience. \nResults or Findings: In the anchoring phase, radiologists altered their \ndiagnoses in 180/400 cases (45%) when AI was incorr ect; XAI reduced this to \n100/400 (25%). In the automation phase, 220/400 cor rect diagnoses (55%) \nchanged after AI input; XAI reduced this to 120/400  (30%). Low-experience \nradiologists showed higher susceptibility, particul arly in automation (260/400, \n65% change rate). XAI improved accuracy in this gro up by 80/400 cases \n(20%). Experienced radiologists demonstrated minima l bias reduction with XAI, \nindicating experience as a moderating factor. \nConclusion: XAI reduces anchoring and automation biases, especi ally for less \nexperienced radiologists when AI errors are present . Tailored AI solutions with \nexplainability are crucial for unbiased decision-ma king in breast imaging. \nLimitations: This study involved a small sample size, which may limit \ngeneralizability. Future studies should expand the sample and explore the \nlong-term impact of XAI on diagnostic confidence. \nFunding for this study: N/A \nEthics committee - additional information: Code of approval: UID 4810 \nAuthor Disclosures:  \nLuciano Mariano: Nothing to disclose \nLuca Nicosia: Nothing to disclose \nAntuono Latronico: Nothing to disclose \nAnna Carla Bozzini: Nothing to disclose \nSerena Carriero: Nothing to disclose \nFrancesca Abbate: Nothing to disclose \nFilippo Pesapane: Nothing to disclose \nEnrico Cassano: Nothing to disclose \nLorenza Meneghetti: Nothing to disclose \n \n \nThe diagnostic performance of an AI model in prosta te cancer detection \ndecreased significantly in reduced scan-quality of biparametric MRIs, \nwhile radiologists’ performances did not decrease \n*E. H. P. Pooch*¹, G. Agrotis¹, A. Dehghanpour², R.  G. H. Beets-Tan¹,  \nT. Janssen¹, I. G. Schoots¹; ¹Amsterdam/NL, ²Rome/I T \n \nPurpose or Learning Objective: To assess the diagnostic performance of \nartificial intelligence (AI) model and radiologists  in detecting Grade Group (GG) \n≧2 disease in prostate cancer suspected men, on diag nostic biparametric MRI \n(bpMRI) scans, considering variations in scan quali ty as assessed by PI-QUAL \nscores. \nMethods or Background: A nnU-Net GG≧2 cancer segmentation model used \n1500 bpMRI scans for training (cohort PI-CAI) and 8 9 scans for external \nvalidation (cohort PROMIS). The external cohort ana lysis included PI-\nRADSv2.1 assessment by two readers (R), while one a ssigned PI-QUALv1 \n(MRI-quality) scores. The outcome measurement was G G≧2 cancer, based on \nbiopsies. MRI-positive scans were defined as PI-RAD S 3-5 scores. The \nmodel’s and radiologists' diagnostic performance (A UCs) were compared. \nResults or Findings: Overall, the trained model (AUC=0.888) achieved an \nAUC of 0.652(0.525-0.760) during external validatio n. At reduced scan quality \n(PI-QUAL 1-3), the model’s AUC dropped to 0.552(0.3 50-0.747), while at high-\nquality scans (PI-QUAL 4-5), the model’s AUC improv ed to 0.720(0.556-\n0.855). In contrast, the AUCs of R1 and R2 were 0.7 33(0.631-0.829) and \n0.711(0.614-0.803), respectively, showing a signifi cant difference to the AI \nmodel in the reduced-quality group (p<0.04), but no t significant in the high-\nquality group (p>0.99). The readers’ AUCs did not d rop at reduced scan quality \n(AUCs 0.723(0.576-0.875) and 0.727(0.576-0.862) and  did not improve at \nhigh-quality scans (0.743(0.616-0.848) and 0.695(0. 562-0.812)), respectively. \nConclusion: The diagnostic performance of the AI model differed  significantly \nbetween reduced- and high-quality scans. In contras t, radiologists maintained \nconsistent diagnostic accuracy. To ensure optimal p erformance, consistently \nhigh-quality MRI scans are required for a successfu l implementation of AI in \nclinical practice. \nLimitations: Limited sample size and only one radiologist provid ed PI-QUAL \nscores, which may limit the findings’ generalizabil ity. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was made using public \ndata. \nAuthor Disclosures:  \nGeorgios Agrotis: Nothing to disclose \nAilin Dehghanpour: Nothing to disclose \nEduardo H. P. Pooch: Nothing to disclose \nTomas Janssen: Nothing to disclose \nRegina G. H. Beets-Tan: Nothing to disclose \nIvo Gerardus Schoots: Nothing to disclose \n \n \n \n \n \nVariability of classification labels is an importan t barrier to effective \ncomparison of artificial intelligence software betw een vendors \n*A. Maiter*, E. Hesketh, P. Metherall, J. Taylor, S . Alabed, K. Dwivedi,  \nW. Tindale, A. Swift, C. S. Johns; Sheffield/UK \n \nPurpose or Learning Objective: Comparing the performance of AI software \nbetween different vendors is important for guiding procurement and \ndeployment decisions. This requires consistency in how software outputs are \npresented. We assessed the number, nature and termi nology of classification \nlabels provided by commercially available software from seven vendors for the \nevaluation of chest radiographs. \nMethods or Background: The classification labels provided by the software \nfrom each vendor were appraised qualitatively and u sing descriptive statistics. \nSynonymous labels were reconciled by merging. Label s were categorised \naccording to their intended purpose. Where relevant , label terminology was \ncompared with the 2024 and 2008 Fleischner Society Glossary of Terms for \nThoracic Imaging. \nResults or Findings: The median number of labels per vendor was 17 (IQR 7 \nto 100). Most labels were for the detection of path ology (median 88%, IQR \n74% to 94%); these varied from non-specific signs ( e.g. ‘bronchovascular \nmarkings’) to specific diagnoses (e.g. ‘sarcoidosis ’). In some cases, individual \nvendors provided multiple labels with overlapping m eanings (e.g. \n‘consolidation’, ‘air bronchogram’, ‘air space opac ification’ and ‘alveolar pattern \nopacity’). Fewer labels were for the detection of d evices (median 12%, IQR 0% \nto 23%); these also ranged from non-specific (e.g. ‘catheter’) to more specific \nwith a decision on adequacy (e.g. ‘suboptimal nasog astric tube’). The median \nconcordance of terminology with the Fleischner Soci ety Glossary was 58% \n(IQR 50% to 72%). \nConclusion: We identified considerable variability in the label s provided by \nsoftware, including inconsistent adherence to estab lished terminology. This \nrepresents a barrier to effective comparison of per formance between vendors \nand potentially limits the clinical utility of soft ware outputs. Our study highlights \nthe need for better harmonisation of output labels across the AI field. \nLimitations: The interpretation of classification labels can be subjective and \nmay differ between assessors. \nFunding for this study: This study was funded by the NHS South Yorkshire \nIntegrated Care System. \nEthics committee - additional information: This study did not involve patient \ndata, and no ethics committee approval was required . \nAuthor Disclosures:  \nJonathan Taylor: Nothing to disclose \nAndrew Swift: Nothing to disclose \nAhmed Maiter: Other: Bayer \nChris S Johns: Nothing to disclose \nSamer Alabed: Nothing to disclose \nPeter Metherall: Nothing to disclose \nEleanor Hesketh: Nothing to disclose \nKrit Dwivedi: Nothing to disclose \nWendy Tindale: Nothing to disclose \n \n \nRethinking Radiology Reports: The Perspective of Re ferring Physicians \n*P. Reschke*, L. D. Gruenewald, V. Koch, E. Höhne, T. J. Vogl, J. Gotta; \nFrankfurt/DE \n \nPurpose or Learning Objective: High report quality and completeness are \nessential for efficient patient management. However , the clarity and \ncomprehensiveness of radiology reports are often a point of contention among \nreferring physicians. This study aims to assess ref erring physicians’ \nperspectives on the quality and utility of radiolog y reports in clinical practice. \nMethods or Background: A prospective, anonymous online survey was \nconducted from June 2023 to June 2024, targeting pr acticing physicians in \nGermany, including internists, general practitioner s, surgeons. \nResults or Findings: A total of 149 participants were included: 40% inte rnists, \n35,8% general practitioners, 24,2% surgeons. The av erage satisfaction score \nfor radiology report completeness was 34.4 (±42.3) on a scale from -100 to \n+100. The primary reasons for incomplete reports we re a lack of clinical \ncontext (33.3%), missing prior imaging (18.6%), ina ppropriate imaging \ntechniques (13.8%) and unclear clinical questions ( 11,6%). Nearly half of the \nrespondents (48.9%) preferred concise reports, whil e 35.7% opted for medium-\nlength, and only 15.4% favored detailed reports. A majority of participants \npreferred semi- or fully structured reporting forma ts (92.5%), with free-text \nbeing rarely chosen (7.5%), showing no significant differences across \nspecialties (p=0.08). Most participants (84.1%) fou nd imaging in \ninterdisciplinary case conferences valuable for und erstanding reports, with \n35.9% rating them as “very helpful.” \nConclusion: Referring physicians strongly prefer structured rep orting and \nconcise radiology reports. Integrating imaging into  interdisciplinary meetings \ncan further improve report comprehension. \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 203  \nLimitations: As the survey was conducted exclusively in Germany,  the results \nmay not be directly applicable to other healthcare systems or international \nsettings, where clinical practices and communicatio n standards may differ. The \nstudy focused on a limited number of specialties (i nternists, general \npractitioners, surgeons), potentially overlooking t he perspectives of other key \nstakeholders such as neurologists, oncologists, or emergency medicine \nphysicians. \nFunding for this study: None. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nLeon D. Gruenewald: Nothing to disclose \nVitali Koch: Nothing to disclose \nPhilipp Reschke: Nothing to disclose \nThomas J. Vogl: Nothing to disclose \nJennifer Gotta: Nothing to disclose \nElena Höhne: Nothing to disclose \n \n \n08:00-09:00 Room G1 \nResearch Presentation Session: \nRadiographers \nRPS 1714 \nAdvancements in MR: safety, patient care, \nand technological innovation \n \nModerators \nS. B. Grover; Greater Noida/IN  \n(shabnamgrover@yahoo.com) \nC. Tsiotsios; Limassol/CY \n(crtsiotsios@gmail.com) \n \n \nMRI safety education across Europe: perspectives on  teaching and \nassessment \nJ. Mcnulty¹, A. Cradock¹, A. Mcgee¹, *A. De Bock*²,  J. L. Portelli³, A. England⁴; \n¹Dublin/IE, ²Brussels/BE, ³Msida/MT, ⁴Utrecht/NL \n \nPurpose or Learning Objective: Little has been published on the details of \ncurricular content for radiographers in MRI safety despite the EFRS MRSO role \ndescriptor benchmark document being published in 20 21. This survey aimed to \ncapture the current status quo in terms of MRI safe ty education for \nradiographers across Europe in terms of content, pe dagogical and assessment \napproaches, challenges, and opportunities. \nMethods or Background: An online survey of educational institutions (EIs) \ninvolved in the delivery of MRI safety education to  radiographers was \ndesigned. The survey was distributed via the EFRS t o all member EIs. \nAdditionally, it was shared with all member nationa l societies and to committee \nand working group members for onward distribution t o EIs. \nResults or Findings: Responses were received from 69 EIs. 58 EIs confirm ed \nthey deliver MRI safety education to radiographers (23 to undergraduate (UG) \nmedical imaging (MI) students, 23 to UG combined MI  and radiotherapy (RT) \nstudents, 5 to UG RT students, and 22 to Masters st udents. The most common \nMRI safety teaching methods for postgraduates (PGs)  were in-person lectures \n(70.8%), small group discussions (61.9%), review of  MRI safety scenarios \n(61.9%), and clinical placements (61.9%). The MRI s afety contact hours for \nPGs averaged just 6 (range: 0-30 hours). For PGs, a t the ‘advanced level’, the \nmost infrequently taught topics were ‘Digital data safety’ (27.8%) and ‘Special \nMRI systems’ (33.3%). 65.2% of EIs were either ‘ver y satisfied’ or ‘satisfied’ \nwith their PG MRI safety education provision. \nConclusion: Significant heterogeneity in content and levels of delivery of MRI \nsafety education for radiographers was evident and highlights the potential for \na common MRSO curriculum. \nLimitations: This was a convenience sample of EIs across Europe accessed \nthrough the EFRS network and thus may not be repres entative of the \npopulation as a whole. \nFunding for this study: The ECSO-MRI Project is co-funded by the \nErasmus+ Programme of the European Union (KA220: 20 22-1-BE02-KA220-\nHED-000085873). \nEthics committee - additional information: Due to the nature of this study \nand the target population, the University College D ublin Human Research \nEthics Committee – Sciences granted a formal Ethics  Exemption (Reference: \nLS-LR-24-203-McNulty). \n \n \n \nAuthor Disclosures:  \nJonathan Loui Portelli: Nothing to disclose  \nAndrea Cradock: Nothing to disclose  \nAnke De Bock: Nothing to disclose  \nJonathan Mcnulty: Nothing to disclose  \nAllison Mcgee: Nothing to disclose \nAndrew England: Nothing to disclose \n \n \nVetting MR referrals, are radiographers as good as radiologists? \n*E. Kjelle*, I. Ø. Brandsæter, J. Porthun, B. M. Ho fmann; Gjøvik/NO \n(elin.kjelle@usn.no) \n \nPurpose or Learning Objective: This study compares radiographers' and \nradiologists' vetting of justification for MRI refe rrals for low back pain and \nheadaches. \nMethods or Background: 360 lower back MR referrals and 353 brain MR \nreferrals in adult patients were collected from pri vate imaging centers in \nNorway. Four experienced radiologists and four expe rienced radiographers \nvetted the referrals using the Choosing Wisely reco mmendations for patients \nwith low back pain and uncomplicated headaches. The  assessors had three \nalternatives: 1)Justified, 2)Unjustified, and 3)Nee d more information. Data was \nanalyzed using descriptive statistics, chi-square t est to compare groups, and \nGwen's AC2 for inter-rater agreement analysis. Sign ificant level was p<0.05 \nResults or Findings: On average, in brain MR, the radiographers rated 52 % \nof the referrals as justified, compared to 53% amon g the radiologists. The \nunjustified rate was 32% among radiographers and 41 % among radiologists. In \nthe category need more information, the rate was 15 % for radiographers and \n6% for radiologists. The difference was statistical ly significant, p<0.001. The \ninterrater agreement was moderate 0.45 (95% CI:0.38 -0.52) among the \nradiographers and good 0.71 (95% CI:0.68-0.75) amon g the radiologists. In \nlower back MRI, the justified rate was 65% for radi ographers and 59% among \nradiologists. The unjustified rate was lower among the radiographers (18%) \nthan among radiologists (26%). While the need more information rate was \nhigher among radiographers. The difference was stat istically significant, \np<0.001. The interrater variability was moderate in  both groups, with 0.60 (95% \nCI:0.53-0.66) among the radiographers and 0.56 (95%  CI:0.49-0.62) among \nthe radiologists. \nConclusion: Radiographers need more referral vetting training t o avoid \nperforming unjustified imaging. In this study, radi ographers often required more \ninformation or rated a referral as justified compar ed to radiologists. \nLimitations: The general clinical practices and experience could  influence the \nassessors' vetting; thus, these results are not nec essarily generalizable. \nFunding for this study: This study was funded by the Norwegian Research \nCouncil (Project number 302503). \nEthics committee - additional information: Regional Committees for Medical \nand Health Research Ethics ref.no. 378396 \nAuthor Disclosures:  \nIngrid Øfsti Brandsæter: Nothing to disclose \nBjørn Morten Hofmann: Nothing to disclose \nElin Kjelle: Nothing to disclose \nJan Porthun: Nothing to disclose \n \n \nAssessing MRI referrals' appropriateness for low ba ck pain post a \nradiology-initiated intervention \n*C. C. Chilanga*¹, M. Heggelund¹, E. Kjelle²; ¹Dram men/NO, ²Gjøvik/NO \n(catherine.chilanga@usn.no) \n \nPurpose or Learning Objective: To evaluate a pilot intervention to reduce \nlow-value Magnetic Resonance Imaging (MRI) referral s for Low Back Pain \n(LBP). \nMethods or Background: The study evaluated MRI referrals for LBP before \nand after an intervention involving information cam paigns and return letters to \nclinicians whose referrals were declined. Four radi ologists and two \nradiographers assessed the referrals based on quali ty and justification. \nJustification was classified as justified, unjustif ied, or requiring more \ninformation. A point system rated quality on an 8-p oint scale, with scores \nabove 5.5 marked as \"good\" and below 2.5 as \"poor.\"  Pre- and post-\nintervention variations were analysed using mixed m odel in Stata (Release 18). \nA p-value <.05 was considered significant. \nResults or Findings: A total n= 300 referrals (150 pre- and post-interve ntion) \nwere assessed. Post-intervention, rated justified r eferrals increased from 63% \nto 68%, while unjustified referrals decreased from 19% to 17%. Those needing \nmore information decreased from 19% to 16%. Poor-qu ality referrals \ndecreased to 4% post intervention. Mixed model anal ysis estimated justified \nreferrals at 61% (95% CI: 55.8–65.5) pre-interventi on and increased to 66% \n(95% CI: 61.5–70.9) post-intervention. Unjustified referrals fell from 20% (95% \nCI: 15.9–24.3) to 17% (95% CI: 13.5–21.4), Rated go od quality referrals \nincreased from 7% (95% CI: 4.6–9.0) to 8% (95% CI: 5.5–10.7), poor-quality \nreferrals decreased from 30% (95% CI: 25.4–35.3) to  27% (95% CI: 22.3–\n31.7). Variations were not statistically significan t. \n\n \n \nSaturday \nAbstract-based Programme \n \n 204  \nConclusion: The intervention is ongoing and needs further evalu ation. \nHowever, providing reasons for declined referrals c an serve as an educational \ntool for clinicians and contribute to the reduction  of low value MRI for LBP in \nradiology departments. \nLimitations: Adherence to referrers’ confidentiality during data  sampling \nprevented confirmation of whether all post-interven tion referrals originated from \nclinicians who received return letters; however, re gion-wide campaigns likely \nmitigated this issue. \nFunding for this study: The Research council of Norway (Project number \n302503) \nEthics committee - additional information: Regional Committees for Medical \nand Health Research Ethics (REK) reference number 3 78396 and Norwegian \nAgency for Shared Services in Education and Researc h (SIKT) reference \n261461. \nAuthor Disclosures:  \nMina Heggelund: Nothing to disclose  \nElin Kjelle: Nothing to disclose \nCatherine Chilute Chilanga: Nothing to disclose \n \n \nWhite matter hyperintensities and silent brain infa rcts in aortic valve \nrepair: the PEARL Study \n*M. Bono*¹, M. Zanardo², V. Bari², B. Cairo², A. Po rta², F. Sardanelli², P. Vitali²; \n¹Varese/IT, ²Milan/IT \n(martinabono93@tiscali.it) \n \nPurpose or Learning Objective: We assessed the burden of white matter \nhyperintensities (WMH), a biomarker of chronic cere brovascular disease, and \nquantified ischemic lesions using diffusion-weighte d magnetic resonance \nimaging (DW-MRI) in patients undergoing surgical ao rtic valve replacement \n(SAVR) or transcatheter aortic valve implantation ( TAVI). Our goal was to \ninvestigate whether a relationship exists between c hronic cerebrovascular \ndisease and acute ischemia in aortic valve repair p atients. \nMethods or Background: This prospective study involved brain MRI scans at \n1.5-T performed within seven days following SAVR or  TAVI. The semi-\nquantitative Fazekas scale was used to classify WMH  severity into low, \nintermediate, and high categories of chronic cerebr ovascular disease. WMHs \nand DWI-positive ischemic lesions were quantified u sing automatic \nsegmentation by Quantib® ND and semi-automatic segm entation adjusted by \na neuroradiologist and a trained radiographer. \nResults or Findings: A total of 55 patients were included in the study, of \nwhom 47 underwent SAVR (62±15 years) and 8 TAVI (83 ±3 years). The mean \nlesion count for SAVR patients was 2.11 ± 7.07, while the mean lesion count \nfor TAVI patients was 7.25 ± 6.88(p = 0.081).An analysis of WMH using the \nFazekas scale indicated that SAVR group had mean Fa zekas score of 1.30 ± \n0.69, while TAVI group 2.25 ± 0.71 (p = 0.053). \nConclusion: While TAVI patients tended to have a higher burden of chronic \ncerebrovascular disease as indicated by the Fazekas  score, the lack of \nsignificant differences in ischemic lesion counts b etween SAVR and TAVI \nsuggests that both procedures pose similar acute is chemic risks. The WMH \nburden may enable risk-stratification of patients w ho undergo TAVI/SAVR and \nidentify those that would benefit most from the ado ption of neuroprotective \ndevices. \nLimitations: Low number of TAVI patients enrolled. \nFunding for this study: Ricerca Finalizzata Code: RF-2016-02361069. \nEthics committee - additional information: Ethics committee Lombardia \nAuthor Disclosures:  \nPaolo Vitali: Nothing to disclose \nVlasta Bari: Nothing to disclose \nFrancesco Sardanelli: Nothing to disclose \nMartina Bono: Nothing to disclose \nAlberto Porta: Nothing to disclose \nMoreno Zanardo: Nothing to disclose \nBeatrice Cairo: Nothing to disclose \n \n \nPreliminary results on white matter hyperintensitie s and lesion counts in \npatients undergoing cardiac surgery with cardiopulm onary bypass: the \nPASCAL study \nA. Nocita¹, M. Zanardo², V. Bari², B. Cairo², P. Si ngh², A. Porta², F. Sardanelli², \nP. Vitali², *M. Bono*¹; ¹Varese/IT, ²Milan/IT \n(martinabono93@tiscali.it) \n \nPurpose or Learning Objective: This preliminary study investigates changes \nin white matter hyperintensities (WMH) and lesion c ounts in patients \nundergoing cardiac surgery with cardiopulmonary byp ass (CPB). The primary \nobjective was to assess the differences in WMH volu mes and lesion counts \npre- and post-surgery, measured through brain magne tic resonance imaging \n(MRI) Fluid-Attenuated Inversion Recovery (FLAIR) 3 D sequence. \n \n \nMethods or Background: The first fifteen patients of the PASCAL study were  \nincluded. Two brain MRI scans were acquired before the surgical intervention \nwith cardiopulmonary bypass (pre), and after the su rgical intervention (post) \nbut within 10 days. Pre-operative and post-operativ e WMH volumes and lesion \ncounts were evaluated using automatic segmentation by Quantib® ND and \nsemi-automatic segmentation adjusted by a neuroradi ologist and a trained \nradiographer. A t-test was conducted for statistica l analysis, while \nreproducibility was evaluated using a Bland-Altman plot. \nResults or Findings: The mean pre-surgical WMH volume was 0.49 cm³, \nwhich increased slightly to 0.57 cm³ post-surgery. A paired t-test showed a \nstatistically significant difference between pre- a nd post-surgical volumes (p = \n0.044). The lesion count analysis showed a mean pre -operative count of 13, \nincreasing to 14 post-surgery. However, this differ ence was not statistically \nsignificant (p = 0.09). The Bland-Altman analysis f or WMH volume \ndemonstrated a mean difference of 0.04 cm³ with lim its of agreement ranging \nfrom -0.15 to 0.23 cm³. For lesion count, the mean difference was 1 with limits \nof agreement from -5 to 7, indicating moderate repr oducibility between the two \nmethods. \nConclusion: The results suggest that while WMH volumes may incr ease after \nCPB surgery, the number of lesions does not signifi cantly change, although the \nsample size limits the generalisability of these fi ndings. Definitive results will be \nobtained once the estimated sample size is reached.  \nLimitations: Preliminary results. \nFunding for this study: Ricerca Finalizzata Code: GR-2021-12372037. \nEthics committee - additional information: Ethics Committee Lombardia 1, \ncode: 06/INT/2023 \nAuthor Disclosures:  \nPaolo Vitali: Nothing to disclose \nAlba Nocita: Nothing to disclose \nVlasta Bari: Nothing to disclose \nPavandeep Singh: Nothing to disclose \nFrancesco Sardanelli: Nothing to disclose \nMartina Bono: Nothing to disclose \nAlberto Porta: Nothing to disclose \nMoreno Zanardo: Nothing to disclose \nBeatrice Cairo: Nothing to disclose \n \n \nHardware as a Predictor of Anxiety in Patients Unde rgoing Magnetic \nResonance Imaging Examinations \n*D. A. A. Costa*, A. Grilo, E. Carolino, M. C. P. R ibeiro; Lisbon/PT \n(diogo2242@gmail.com) \n \nPurpose or Learning Objective: The objective of this study was to assess \nwhich type of Magnetic Resonance (MRI) examination,  the coil used and the \npatient position (to scan the brain or the knee), m ost influence patient’s anxiety \nlevels. \nMethods or Background: Hundred patients underwent MRI scan using a 1,5T \nMagneton Symphony by Siemens Healtineers. Fifty of them performed brain \nand the others knee scan. Before and after the MRI scan, the STAI Inventories \nForm Y-1 (state anxiety) and the Form Y-2 (trait an xiety) was applied. The \nphysiological measurements of Blood Pressure and He art Rate were collected \nat the beginning and at the end of scan. \nResults or Findings: In brain studies, the initial mean anxiety levels ( \u0004̅ =37.28 \n± 12.446) are higher than those collected after examination (\u0004̅ =33.72 ± \n13.389). In the knee examinations the initial mean anxiety values (\u0004̅ =31.96 ± \n10.681) are higher compared to the final values ( \u0004̅ =29.42 ± 8.094). When \ncomparing brain and knee studies, the first shown h igher initial anxiety levels (\u0004̅ \n=37.28 ± 12.446) than patients who undergo knee sca n (\u0004̅ =31.96 ± 10.681). \nThe mean values of Maximum BP ( \u0004̅ =130.06 ± 3.808) and Final HR (\u0004̅ \n=125.10 ± 1.851), collected after brain studies, are higher than the values \ncollected in knee (\u0004̅ =88.24 ± 2.720) and (\u0004̅ =81.12 ± 1.943). \nConclusion: Through Multiple Linear Regression (Stepwise method ), the \nexamination type and the patient's assessment befor e MRI are predictors of \nanxiety, as their probability of occurrence is low.  Higher and significant \ndifferences were found at the beginning of MRI scan . The appliance of different \ntransceiver coils, reduces the surrounding space, c ontributing to increasing the \nlevels of state anxiety. \nLimitations: Lack information about other pathologies. Higher pe riods of \nquestions and evaluation \nFunding for this study: Not Applied \nEthics committee - additional information: Approval by Ethical Commission \nof School of Health Technology with the reference C E-ESTeSL-Nº.107-2022 \nAuthor Disclosures:  \nAna Grilo: Nothing to disclose \nElisabete Carolino: Nothing to disclose \nDiogo André Arrais Costa: Nothing to disclose \nMargarida Carmo Pinto Ribeiro: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 205  \nEducational background in MRI Safety of healthcare professionals \nworking in MRI departments: Insights from the ECSO- MRI Project \nJ. Scheurleer¹, A. V. Diepen¹, *A. De Bock*², H. Bi jwaard¹; ¹Haarlem/NL, \n²Brussels/BE \n \nPurpose or Learning Objective: This study aimed to assess the educational \nneeds and preferences regarding MRI safety training  among healthcare \nprofessionals working in MRI departments. Findings aimed to support the \ndevelopment of a standardised European MRI Safety O fficer (MRSO) \ncurriculum as part of the funded ECSO-MRI project. \nMethods or Background: A survey among radiographers, radiologists, and \nother healthcare professionals across European MRI departments assessed \neducational backgrounds, MRI safety responsibilitie s, and training preferences, \nwith 313 responses analysed. The study received loc al ethics committee \napproval and confirmed European data security requi rements. \nResults or Findings: Of the respondents, 71.6% had clinical experience i n \nMRI, but 49.5% did not regularly update their MRI s afety knowledge. Informal \npeer-to-peer training was the most commonly used le arning method (62.9%). \nHands-on training was the most preferred approach ( 89.5%). Additionally, 70% \nof respondents considered the development of Standa rd Operating Procedures \n(SoP) an advanced MRI safety topic. Furthermore, 40 % highlighted the need \nfor formal basic training for all radiographers wor king with MRI, while 63% \nindicated a need for formal advanced training for k ey users of MRI. Only 31% \nreported following local MRI safety guidelines, and  18.5% adhered to \ninternational standards, underscoring the need for more consistent safety \nprotocols. \nConclusion: Significant gaps in MRI safety education, particula rly in advanced \ntraining, were identified. The ECSO-MRI project aim s to address these by \ndeveloping a standardized curriculum in collaborati on with European \ninstitutions. \nLimitations: The reliance on self-reported data may introduce bi as in \nassessing actual MRI safety practices and training needs. \nFunding for this study: This study was funded by the European Union’s \nErasmus+ program. \nEthics committee - additional information: Ethics Committee of University \nCollege Dublin Research Ethics Research Ethics Refe rence Number is: LS-LR-\n24-220-McNulty \nAuthor Disclosures:  \nAnja Van Diepen: Nothing to disclose \nAnke De Bock: Nothing to disclose \nJelle Scheurleer: Nothing to disclose \nHarmen Bijwaard: Nothing to disclose \n \n \nRadiographers’ Knowledge and Attitudes differences towards Cardiac \nImplant Patients between radiographers in Magnetic Resonance Imaging \nC. Maloney, A. England, N. Moore, R. Young, G. A. C urran, *M. F. Mcentee*; \nCork/IE \n(mark.mcentee@ucc.ie) \n \nPurpose or Learning Objective: Radiographers in magnetic resonance \nimaging (MRI) are at the forefront of patient safet y, and patients with cardiac \nimplantable electronic devices (CIED) have been ref erred for MRI imaging \nmore frequently in recent years. A rise in MR-condi tional devices has resulted \nin CIED patients accessing MRI, and previously pati ents with CIEDs were \nclassed as absolute contraindications. However, the  magnetic fields from the \nMRI scanner can interfere with CIEDs causing damage  to the device itself \nand/or the patient. This study evaluates CIED speci fic knowledge, confidence \nlevels and attitudes amongst radiographers in MRI t owards cardiac implant \npatients. \nMethods or Background: A quantitative, online survey was conducted \namongst qualified radiographers in MRI internationa lly to evaluate participants \nMRI qualification status, what CIED specific knowle dge is held by \nradiographers, confidence levels amongst radiograph ers scanning CIED \npatients, and determine common attitudes amongst ra diographers towards \nimplant patients who present for an MRI scan. \nResults or Findings: 90 responses were recorded and overall, and 58% of \nparticipants held a postgraduate degree in MRI whil e 42% did not. 73% of \nradiographers highlighted they had opportunities to  update their MRI safety in \nthe work setting, while only 36% had opportunities to update their cardiac \nimplant safety knowledge. Overall, radiographers co ncluded that their MRI and \nCIED safety knowledge was sufficient and possessed adequate knowledge of \nthe CIED and magnetic field relationship. However, evident disagreement on \nconfidence statements on explaining MR field impact s on CIED functioning, \nhaving insufficient knowledge on device workings to  confidently scan and \nconflicting safety results indicate that this area requires addressing. \nConclusion: Inclusion of specific CIED safety training for MRI radiographers \nshould be recommended as educational topics to equi p radiographers with the \nknowledge, skills and confidence to competently pro vide safe patient care. \nLimitations: Survey distribution was in English. \nFunding for this study: None \nEthics committee - additional information: Medical School Social Research \nEthics Committee - University College Cork \nAuthor Disclosures:  \nMark F. Mcentee: Nothing to disclose \nChloe Maloney: Nothing to disclose \nNiamh Moore: Nothing to disclose \nRena Young: Nothing to disclose \nGráinne Alison Curran: Nothing to disclose \nAndrew England: Nothing to disclose \n \n \n09:30-11:00 Research Stage 1 \nResearch Presentation Session: \nInterventional Radiology \nRPS 1809 \nWhat's new in genitourinary and female \nhealth interventions? \n \nModerator \nV. Kostadinova; Ljubljana/SI  \n \n \nP-RENAL: a new scoring system for the prediction of  complications and \nrecurrence in kidney malignancies percutaneous abla tion \n*G. Ferrillo*, D. Poretti, N. Buffi, P. Casale, M. Francone, V. Pedicini; Milan/IT \n(giuseppe.ferrillo@humanitas.it) \n \nPurpose or Learning Objective: There is a lack of internationally validated \nscores for the assessment of outcomes in percutaneo us ablation of kidney \ntumours. We designed the P-RENAL score and compared  it to the surgical \nstandard score (RENAL). \nMethods or Background: We performed a retrospective analysis of 146 \npatients treated with RFA (N=90) or MWA (N=56) betw een 01/01/2016 and \n31/03/2021 (minimum follow up of 6 months). RENALan d p-RENAL were \ncalculated in both populations. Primary outcomes we re recurrence (presence \nof vital tumour >1 cm in the treated area) and comp lications (according to the \nCIRSE complications score). P-RENAL is derived from  RENAL, it consists of 6 \nparameters, each assigned a value from 1 to 3: radi us (<2 cm, 2-3cm, >3cm), \nexophytic area and nearness to the collecting syste m (identic to RENAL \nscore), location (1 interpolar mass; 2 lower polar;  3 apical tumours), side (1 \nposterior lateral; 2 medial posterior; 3 medial ant erior), proximity of external \nstructure(>3 cm from an high risk structureas ileal  or colic wall,renal artery, \nspleen or liver); 2 from 3 to 1 cm; 3 points if <1 cm. \nResults or Findings: P-RENAL showed higher sensitivity than RENAL score \nin predicting recurrence after ablation of kidney t umours expecially with MWA \n(AUC 0.86). There was no association of the RENAL s core with complications, \nwhile we observed an independent positive correlati on between p-RENAL and \nthe presence of grade > 2 complications. \nConclusion: P-RENAL has higher sensitivity than RENAL in detect ion of \nrecurrence and complications in percutaneous ablati on. \nLimitations: Retrospective monocentric study. \nFunding for this study: None \nEthics committee - additional information: Retrospective study board \nreview. \nAuthor Disclosures:  \nVittorio Pedicini: Nothing to disclose \nNicolò Buffi: Nothing to disclose \nMarco Francone: Nothing to disclose \nDario Poretti: Nothing to disclose \nPaolo Casale: Nothing to disclose \nGiuseppe Ferrillo: Nothing to disclose \n \n \nThulium laser enucleation of the prostate versus pr ostatic artery \nembolization for big prostates above 80gms: a multi centre study \n*N. M. A. Attia*¹, M. A. El Hamid², M. Abd El Wahab ², A. Salah²; ¹Assiut/EG, \n²Cairo/EG \n(nohamohamedali@yahoo.com) \n \nPurpose or Learning Objective: To compare the postoperative and functional \noutcomes of thulium laser enucleation of the prosta te (ThuLEP) and prostatic \nartery embolization (PAE) for the treatment of larg e-volume benign prostate \nhyperplasia (BPH) > 80 ml. \n\n \n \nSaturday \nAbstract-based Programme \n \n 206  \nMethods or Background: We performed a retrospective, multicentric study of  \n120 patients consecutively treated with THuLEP (60 patients) or PAE (60 \npatients) for symptomatic large BPH >80gms between March 2022 and March \n2023. These two groups were compared by treatment r esponse, postoperative \nrecovery period and complication rate at baseline, 1, 3 and 6 months. \nResults or Findings: Intraoperative complications were significantly hig her in \nthe THULEP group (13.5% vs 0%, p=0.006). There was no statistically \nsignificant difference in the postoperative complic ations between the two \ngroups (10% vs 8.4%, p=0.287). There was significan t delay of catheter \nremoval in the PAE group with subsequent delay in h ospital stay (25% vs 5%, \np=0.002). International Prostate Symptom Score (4.4 ±0.9SD vs 10.7±1.5SD, \np<0.001) and PSA (1.4±0.6SD vs 4.3±5.1SD, p<0.001) were significantly lower \nin the THuLEP versus PAE group after 6 months. Post void residual (PVR) was \nhigher in the PAE group from baseline till 6months (12.5±3.9SD vs \n20.35±4.8SD, p<0.001). Maximum urinary flow rate (Q max) was significantly \nhigher in the THULEP group from baseline (37.93 ±4SD vs 27.3±8.1SD, \np<0.001). Erectile dysfunction was significantly lo wer in the PAE group after 3 \nmonths. \nConclusion: Both ThuLEP and PAE relieve lower urinary tract sym ptoms with \nhigh efficacy and safety, however ThuLEP was superi or to PAE in improving \nthe functional outcomes in the first 6months while PAE had lower \nintraoperative complications and erectile dysfuncti on. \nLimitations: The quality of life was not assessed being a retros pective study. \nThe sample size was relatively small. Longer follow -up time is needed to \ncompare the durability of the symptomatic improveme nt from each procedure. \nFunding for this study: None \nEthics committee - additional information: This study was approved by the \nCommittee of Medical Ethics of the Faculty of Medic ine Assiut University with \nIRB no: 04-2024-300429 \nAuthor Disclosures:  \nMahmoud A. El Hamid: Nothing to disclose \nNoha Mohamed Ali Attia: Nothing to disclose \nAhmed Salah: Nothing to disclose \nMohamed Abd El Wahab: Nothing to disclose \n \n \nA decade of varicocele embolizations: Success and r ecurrence \n*M. Ni Mhiochain De Grae*¹, M. Alkhattab¹, A. Alkad himi², M. Springael²,  \nG. O' Sullivan¹; ¹Galway/IE, ²Dublin/IE \n(meadhbhnimhioch24@rcsi.ie) \n \nPurpose or Learning Objective: Varicoceles are vascular lesions of the \npampiniform plexus, affecting 10-20% of the populat ion and found in 40% of \nmen with primary infertility and 80% with secondary  infertility. (1-4) Pain occurs \nin 2-10% of cases. (1,5) Treatment options include conservative management, \npercutaneous embolization, or surgery. (6) Percutan eous embolization, \nintroduced in the late 1970s, has a technical failu re rate of around 13% and \nrecurrence rates ranging from 0.6% to 45%. (7,8) Th is study evaluates success \nand recurrence rates of percutaneous varicocele emb olizations at Galway \nUniversity Hospital and Galway Clinic between 2009- 2022, with a minimum \nfollow-up of 18 months. \nMethods or Background: Data collected included patient age, procedure \ndate, access site, side of occurrence, previous int erventions, treatment \nmethod, need for re-intervention, and recurrence ra tes. Technical success was \ndefined as successful access to the gonadal vein an d insertion of \ncoil/sclerosant. Clinical success was assessed thro ugh follow-up consultations \nand ultrasound when available. \nResults or Findings: The technical success rate was 94.7%, with 98.1% of  \nsuccessful embolizations achieving clinical success . Of 225 patients, 3.12% \nhad prior failed surgeries, all treated successfull y with IR, and 0.89% required \nsurgical intervention. Telephone follow-ups (42.7% response rate) revealed a \nrecurrence rate of 10.7% with an average follow-up of 8.03 years. The \ncomplication rate was 1.78%. Fertility outcomes sho wed 51.35% of patients \nhad successful pregnancies, while 6.9% experienced ongoing fertility issues. \nConclusion: Our study of 225 patients is the largest to date, s howing higher \nsuccess rates (technical 94.7%, clinical 98.1%) and  a 10.7% recurrence rate \nwith long-term follow-up. \nLimitations: This is a single centre retrospective review. 42.7%  response to \ntelephone consultation follow up. \nFunding for this study: No funding. \nEthics committee - additional information: Local ethics approval in Galway \nUniversity Hospital. \nAuthor Disclosures:  \nMaia Springael: Nothing to disclose \nAmor Alkadhimi: Nothing to disclose \nMeadhbh Ni Mhiochain De Grae: Nothing to disclose \nGerry O' Sullivan: Nothing to disclose \nMaha Alkhattab: Nothing to disclose \n \n \n \n \nThe Clinical Study of Magnetic Resonance-Guided Hig h-Intensity \nFocused Ultrasound (MRgFUS) Treatment for Adenomyos is \n*Q. Zhang*; Shanghai/CN \n(q_zhang18@fudan.edu.cn) \n \nPurpose or Learning Objective: Objective: To investigate the safety and \nefficacy of Magnetic Resonance-Guided High-Intensit y Focused Ultrasound \n(MRgFUS) treatment on adenomyosis. \nMethods or Background: After receiving approval from local Ethics \nCommittee, 62 patients with adenomyosis underwent M RgFUS treatment from \nOctober 2018 to March 2021. The 62 patients with ad enomyosis were \nevaluated for adverse reactions during and after MR gFUS treatment. The \nlesion volume was measured before and after treatme nt using T2-weighted \nMRI images. Non-perfused volume (NPV) was calculate d using T1-weighted \ncontrast-enhanced images, then the ablation rate wa s calculated by dividing \nNPV by the lesion volume. Visual analogue scale (VA S) was used to score \ndysmenorrhea in adenomyosis patients before and aft er treatment. \nResults or Findings: All 62 patients successfully underwent MRgFUS \ntreatment without severe adverse reactions, achievi ng an average ablation rate \nof 69.12%±17.64%. The average lesion volume before ablation was \n83.72±84.02 cm³. At 3 months post-ablation, the ave rage lesion volume was \n62.66±66.31 cm³, at 6 months it was 56.09±66.67 cm³, and at 12 months it was \n68.53±87.04 cm³. There were statistically significa nt differences in lesion \nvolume (p<0.05) before and after ablation treatment . The preoperative VAS \nscore for the 62 patients was 6.68±2.13. At 3 months post-treatment, the \naverage dysmenorrhea score was 3.69±1.95, at 6 mont hs it was 2.72±1.40, \nand at 12 months it was 3.38±1.45. Post-treatment dysmenorrhea scores \nshowed statistically significant decrease compared to pre-treatment (p<0.05). \nConclusion: MRgFUS is a safe, feasible and effective technique in treatment \nfor adenomyosis. \nLimitations: Serological examinations were not performed in pati ents with \nadenomyosis to assess changes in serologic markers after treatment in this \nstudy. \nFunding for this study: Not applicable \nEthics committee - additional information: Ethics Committee of Huashan \nHospital affiliated to Fudan University \nAuthor Disclosures:  \nQi Zhang: Nothing to disclose \n \n \nTreatment of cesarean scar pregnancy with systemic Methotrexate \nfollowed by intra-arterial Methotrexate injection a nd uterine arteries \nembolization: experience from a tertiary center \n*C. Intrieri*, C. Lanza, S. A. Angileri, J. Tintori , C. Ercolino, A. M. Ierardi,  \nG. Carrafiello, V. Chiarpenello; Milan/IT \n(cristinaintrieri95@gmail.com) \n \nPurpose or Learning Objective: The aim of this retrospective study is to \nevaluate the efficacy and the safety of the treatme nt of cesarean scar \npregnancy (CSP) based on a single or cyclic systemi c methotrexate (MTX) \ninjection followed by intra-arterial MTX injection and uterine arteries \nembolization (UAE). \nMethods or Background: A pregnancy’s ectopic implantation on a uterine \nscar tissue following a previous cesarean section i s defined as CSP. Because \nof its possible complication in uterine rupture, ma ssive hemorrhage, placenta \naccreta spectrum, fetal and maternal death, CSP is considered a life-\nthreatening condition. Therefore, treatment is fund amental. \nThe cooperating group of Radiologists and Gynecolog ists at our institution \nevaluated all patients affected by CSP from 2013 to  2023 (n=21) treated with \nsystemic MTX injection followed by intra-arterial M TX injection and UAE with \ngelatin sponge. The inclusion criteria were: labora tory tests (levels of beta-\nhuman chorionic gonadotropin and beta-HCG over the normal threshold), \nultrasound diagnosis of CSP, gestational age ≤8 weeks, stable hemodynamic, \nat least one previous pregnancy and at least one de livery through cesarean \nsection. Exclusion criteria were: contraindications  to MTX administration, the \npresence of abdominal hemorrhage and significant sh ock. \nResults or Findings: Technical success of UAC and clinical success \n(reduction of the the sac size, lack of active vagi nal re-bleeding and declining \nbeta-HCG values) was achieved in all procedures (10 0%). No peri-procedural \ncomplications were recorded. \nConclusion: In our experience, the use of systemic MTX and intr a-arterial \ninjection of MTX plus UAC proved to be effective an d safe to achieve CSP \ntermination. The use of gelatin sponge after intra- arterial MTX administration \nproved to be effective also in preventing massive b leeding in case of \nsuccessive uterine curettage. \nLimitations: Controlled randomized trial with a larger sample is  needed to \nassess the efficacy and the safety of treatment. \nFunding for this study: No funding was received for this study. \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 207  \nEthics committee - additional information: All subjects gave their consent \nfor inclusion in the present study. The study was c onducted in accordance with \nthe World Medical Association (WMA) Declaration of Helsinki [WMA]. \nAuthor Disclosures:  \nVittoria Chiarpenello: Nothing to disclose \nSalvatore Alessio Angileri: Nothing to disclose \nAnna Maria Ierardi: Nothing to disclose \nCarolina Ercolino: Nothing to disclose \nJacopo Tintori: Nothing to disclose \nCristina Intrieri: Nothing to disclose \nCarolina Lanza: Nothing to disclose  \nGianpaolo Carrafiello: Nothing to disclose \n \n \nWhat is the safe observation period following image -guided renal \nbiopsies? \n*F. Taylor*, K. Sehgal, M. Van Wees, K. Li, D. W. D e Boo, L-A. Slater; \nMelbourne/AU \n(fergustay@gmail.com) \n \nPurpose or Learning Objective: To investigate the timing, type, and severity \nof complications following percutaneous, image-guid ed renal biopsy and to \ndetermine if the current observation period of 4 ho urs can be safely reduced. \nMethods or Background: Consecutive image-guided percutaneous targeted \nand non-targeted renal biopsies performed between 2 017-2022 in adult \npatients by radiology medical staff were included. The PACS imaging system \nand electronic medical records (EMR) were accessed to obtain relevant patient \ninformation and procedural reports. Retrospective a nalysis of the type and \ntiming of complications was performed against patie nt demographics and \nbiopsy-related variables. Probabilities were calcul ated for a range of \nobservation periods to assess the proportion of com plications identified by \nshortening the observation period from 4 hours. \nResults or Findings: 332 percutaneous renal biopsies were included and 4 4 \n(13%) complications identified within the 4-hour ob servation. 29 complications \nwere post-operative bleeding, of which 2 developed macroscopic haematuria, \n25 peri-nephric haematoma and 2 had hemodynamic ins tability with either \nhaematuria or peri-nephric hematoma. 64% of all com plications occurred within \nthe first hour, 86% occurring within 2 hours. Of th e 6 complications occurring \nafter 2 hours, 2 were post-operative bleeding in no n-targeted renal biopsies \nrequiring admission, the other 4 did not require ad ditional \nobservation/procedures. \nConclusion: The vast majority of complications after targeted a nd non-\ntargeted renal biopsies tend to occur within the fi rst 2 hours of observation. \nComplications that occurred after 2-hour observatio n were often pain related, \nsomething that can be overcome with a standardised post-operative analgesic \nregime. It may be possible to safely reduce observa tion times following image-\nguided targeted renal biopsies. \nLimitations: Single institute dataset, including targeted and no n-targeted \nbiopsies. Biopsies were performed by Radiologists w ith varied experience and \nsupervision, not ascertained during retrospective d ata collection. The low \nnumber of complications limits our analysis of asso ciated factors. \nFunding for this study: This research did not receive any specific grant fr om \nfunding agencies in the public, commercial, or not- for-profit sectors. This study \nwas not supported by any funding. \nEthics committee - additional information: All procedures performed in \nstudies involving human participants were in accord ance with the ethical \nstandards of the institutional and/or national rese arch committee and with the \n1964 Helsinki declaration and its later amendments or comparable ethical \nstandards. For this type of study formal consent is  not required. \nEthical approval was obtained from the local Human Research Ethics \nCommittee (HREC), reference number: RES-23-0000-015 Q and the need for \ninformed consent was waived. \nAuthor Disclosures:  \nFergus Taylor: Nothing to disclose \nMatthew Van Wees: Nothing to disclose \nKunal Sehgal: Nothing to disclose \nDiederick Willem De Boo: Nothing to disclose \nKenny Li: Nothing to disclose \nLee-Anne Slater: Nothing to disclose \n \n \nProstate artery embolization (PAE): are there predi ctors indicating the \ntechnical success rate? \n*T. Lauenstein*, K. Scherschel, J. Boddenberg, F. V erfürth, N. Ziayee; \nDüsseldorf/DE \n(thomas.lauenstein@evk-duesseldorf.de) \n \nPurpose or Learning Objective: To assess if factors including patients’ age, \nprior history of cardiovascular disease or prostate  size have an impact on the \ntechnical success rate of PAE. \n \nMethods or Background: Data of 163 consecutive patients undergoing PAE \nwas analyzed. Technical success rate was defined as  full when arteries of both \nprostate lobes could be probed, as partial (one lob e) or as absent . These \nfindings were correlated with patients’ age (range 53-93 years; mean 72 \nyears), prior history of cardiovascular disease (pr esent in 88 patients) and size \nof the prostate gland (range 36-265ml; mean 84ml). \nResults or Findings: Bilateral PAE could be achieved in 125 patients. In  29 \npatients only one lobe was embolized and PAE failed  to be successful in 9 \npatients. Patient’s age failed to show a statistica lly significant difference \nbetween these groups (71.4±8.4 years for successful vs. 72.4±9.2 years for \nunsuccessful procedures, P=0.5576, unpaired ttest),  as well as prior history of \ncardiovascular disease (52% for successful vs. 60% for unsuccessful \nprocedures, P=0.4576. Fishers ttest). In contrast, prostate size was \nsignificantly higher in patients with successful bi lobar PAE: 88±40ml for \nsuccessful vs. 68±39ml for unsuccessful procedures (Mann -Whitney-test; \np=0.0021). A consecutive ROC analysis revealed a pr ostate size of 79ml as a \ncut-off value for the success rate of bilobar PAE. \nConclusion: Prostate size was found to be the only predictor fo r a successful \ncompletion of PAE. Patients with volumes of the pro state gland <79ml should \nbe informed to have a higher risk of technical ther apy failure. \nLimitations: Limitations of this study are related to the retros pective nature of \ndata analysis. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nFrank Verfürth: Nothing to disclose \nThomas Lauenstein: Nothing to disclose \nNaim Ziayee: Nothing to disclose \nKatharina Scherschel: Nothing to disclose  \nJörg Boddenberg: Nothing to disclose \n \n \nFirst intermediate term report of safety and outcom es of radiofrequency \nablation of Renal tumours using an automated energy  delivery-controlled \nsystem \n*G. Ferrillo*, E. Baldassarre, T. Sirugo, E. Lanza,  S. Romano, N. Buffi,  \nP. Casale, M. Francone, V. Pedicini; Milan/IT \n(giuseppe.ferrillo@humanitas.it) \n \nPurpose or Learning Objective: To determine the outcomes (Primary efficacy \nand safety) of RF ablation of hystologically proven  renal cell neoplasms using a \nnew single tip internally cooled RF probe. \nMethods or Background: After ethical commettee approval we performed a \nretrospective analyisis of data reguarding the firs t patients (N= 85)treated with \na new RF system whose electrode has an active tip ( 1-5 cm) that erogates \nhigh energies (400-2500mAs) with a pulsing algorith m optimized efficiency. \nThis allows fast ablation times, comparable to MW, with a higher safety due to \nlower peak temperature reached (< 100 °C). Minimum follow up was 6 months, \non average 12.3 months The outcomes were clinical o utcome (success, partial \nsuccess or failure) and safety. Tecnichal success w as defined as complete \nablation evaluated at any follow up CT-scan; partia l success if any area of \nsuspicious recurrence > 1 cm was identified anytime  during follow up. \nSafety was assessed according to the CIRSE complica tions classification \nResults or Findings: The primary success rate was 91%; secondary success  \nrate of 100 % (7 patient retreated). Reguarding saf ety there was one case of a \nCIRSE >2 complication (grade 3, bleeding needing em bolization), and a total of \n16% of CIRSE grade < 2 complications (N = 14). Ther e was no mortality. \nInterestingly, tumours were divided by location and  there was a high \npercentage of central (N= 10) or endophytic lesions  (N = 29), with \nsuperimposable outcomes of the peripheral lesions. \nConclusion: RFA performed with this new modality resulted safe and effective \nand it is associated with short procedural times. T he automatic energy-delivery \ncontrolled system might overcome limitations due to  long procedural times and \nheat-sink effect in RFA. Future studies are needed to confirm these results on \nlarger populations \nLimitations: The study is etrospective and monocentric \nFunding for this study: None \nEthics committee - additional information: Regional ethical committee \nAuthor Disclosures:  \nVittorio Pedicini: Nothing to disclose \nSilvio Romano: Nothing to disclose \nTiziano Sirugo: Nothing to disclose \nNicolò Buffi: Nothing to disclose \nEleonora Baldassarre: Nothing to disclose \nEzio Lanza: Nothing to disclose \nMarco Francone: Nothing to disclose \nPaolo Casale: Nothing to disclose \nGiuseppe Ferrillo: Nothing to disclose \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 208  \nPredicting local tumour control in renal cell carci noma after cryoablation \nby 3D volumetric image features \n*C-Y. Huang*, J-A. Hong, N-W. Chang, C-C. Li, C-A. Liu, S-H. Shen;  \nTaipei City/TW \n(cy.jeffrey.huang@gmail.com) \n \nPurpose or Learning Objective: Cryoablation has been shown to be a safe \nand effective treatment for renal cell carcinoma (R CC); however, there is \ncurrently no consensus on the predictors for local tumour control. The aim of \nthis study was to implement automatic segmentation of renal structures on \npretreatment CT images, extract relevant image feat ures, and identify potential \npredictors for local tumour control in RCC after cr yoablation. \nMethods or Background: A total of 124 patients with RCC managed with \ncryoablation were included. Pretreatment abdominal CT in nephrographic \nphase were obtained and segmented into masks of the  kidney, tumour, cyst, \nand renal sinus using deep learning segmentation mo dels and post-\nprocessing. Relevant image features were extracted,  and logistic regression \nwas utilized to assess the correlation between thes e image features and \nresponse to cryoablation. \nResults or Findings: Among the 124 patients included, failure of local t umour \ncontrol occurred in 15 (12.1%). Tumour volume (P = .005) and the contact area \nbetween the tumour and the renal sinus (P = .001) w ere significantly \nassociated with local tumour control according to u nivariable logistic \nregression. A multivariable logistic regression dem onstrated that higher tumour \nvolume (P=.051, OR=1.038, 95% CI=1.000-1.079) and a  larger contact area \nbetween the tumour and the renal sinus (P=.005, OR= 1.316, 95% CI=1.088-\n1.591) were associated with failure of local tumour  control. The area under the \nROC curve to predict local tumour control after cry oablation was 0.76. \nConclusion: This study demonstrated that volumetric image featu res, \nincluding tumour volume and the contact area betwee n the tumour and the \nrenal sinus, significantly increase the risk of fai lure of local tumour control in \nRCC after cryoablation. \nLimitations: This study was subject to the inherent shortcomings  of its \nretrospective design. Additionally, it was conducte d at a single centre, so \nexternal validation is warranted. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study was approved by the \nInstitutional Review Board of Taipei Veterans Gener al Hospital (2023-09-\n014CC). \nAuthor Disclosures:  \nChien-An Liu: Nothing to disclose \nChih-Ying Huang: Nothing to disclose \nShu-Huei Shen: Nothing to disclose \nChih-Chien Li: Nothing to disclose \nNai-Wen Chang: Nothing to disclose \nJia-An Hong: Nothing to disclose \n \n \nEfficacy of point of care cortical assay testing du ring adrenal venous \nsampling \n*A. Tinney*, K. Lau, J. Dorey; Clayton/AU \n \nPurpose or Learning Objective: The aim of this single centre retrospective \nstudy was to evaluate the QCA false negative preval ence during AVS. \nMethods or Background: Primary aldosteronism is the most common cause \nof secondary hypertension. Adrenal venous sampling (AVS) is the gold \nstandard for subtyping primary aldosteronism and gu iding treatment. Quick \ncortisol assay (QCA) has proven to be a useful tool  to confirm correct catheter \nposition within the adrenal vein, increasing proced ure success rates. All \nconsecutive AVS procedures on adult patients at our  institution from July 2022 \nuntil March 2024 utilising intraoperative QCA were included. AVS without QCA \nwere excluded. QCA analysis was performed in accord ance with manufacturer \nrecommendations. Correct catheter position was conf irmed with dedicated \nangiographic techniques, and with formal biochemica l analysis. \nResults or Findings: 158 successful and 7 unsuccessful procedures were \nincluded. No QCA false positive result was found in  unsuccessful AVS. 37 \n(23.4%) procedures produced a false negative result  on QCA analysis despite \ncorrect catheter. 23/37 (62.2%) of these procedures  had a bilateral false \nnegative QCA, and 14/37 (37.8%) had an asymmetrical  unilateral false \nnegative QCA result. There was no significant diffe rence in patient \ndemographics across cohorts. Of the false negative QCA tests, 10 individual \ntest kits (3.2%) provided a negative result despite  laboratory cortisol results \nreturning a value above the manufacturer reference value of 828nmol/L. \nConclusion: QCA has been shown to be a useful tool intra-operat ively in \nincreasing AVS sampling success rates. Our results demonstrate that false \nnegative QCA results not infrequently occur and pro ceduralists must remain \nvigilant. Detailed patient history, dedicated angio graphic techniques and \ncorrelation with preoperative CT remain vital when interpreting QCA results. \nOngoing research is required to assess patient fact ors which may contribute to \nfalse negative results. \n \n \nLimitations: Retrospective data availability. \nFunding for this study: N/A \nEthics committee - additional information: Local institute low risk ethics \napproval. \nAuthor Disclosures:  \nJames Dorey: Nothing to disclose \nAdrian Tinney: Nothing to disclose \nKen Lau: Nothing to disclose \n \n \n09:30-11:00 Research Stage 2 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 1801 \nInnovative imaging in colorectal cancer \nand pelvic floor disorders \n \nModerator \nS. Rafaelsen; Vejle/DK \n \n \nMRI evaluation of nodal status after neoadjuvant th erapy in rectal cancer \nwith node-by-node pathological comparison \n*Q-Y. Li*, X-Y. Yan, D. Yang, Z. Guan, R-J. Sun, Q.  Lu, L. X. Ting, X. Zhang, \nY-S. Sun; Beijing/CN \n(631213971@qq.com) \n \nPurpose or Learning Objective: To validate the performance of the ESGAR \ncriteria of nodal status after neoadjuvant therapy (NAT) in rectal cancer and to \ninvestigate how morphological features and changes before and after NAT can \nhelp with node-by-node pathological comparison. \nMethods or Background: Rectal cancer patients who received radical surgery  \nafter NAT and had complete pre- and post-NAT MRI we re consecutively and \nprospectively enrolled. For the nodes that achieved  node-by-node matched \nbetween MRI and pathology, their SADs were measured , and morphological \nfeatures (i.e. shape, internal structure, and borde r) were determined on pre- \nand post-NAT axial T2WI. \nResults or Findings: 207 patients were included and 612 nodes achieved \nmatched, including 471 (77.0%) benign and 141 (23.0 %) metastatic nodes. On \nthe post-NAT MRI, the ESGAR criteria, i.e. SAD ≥5mm, yield an AUC, \nsensitivity, and specificity in determining nodal s tatus of 0.67, 52.5%, and \n82.4%, respectively. All morphological features dif fered between benign and \nmetastatic nodes, with AUCs ranging from 0.55 to 0. 65. The prediction model \ncombined of the only morphological independent pred ictor, internal structure, \nand SAD didn’t result in an improved diagnostic per formance compared to \nSAD alone (P=.64). As the changes before and after NAT, there were \ndifferences in the size reduction rate and the chan ge in internal structure, but \nnot in shape and border. The AUC of the only indepe ndent predictor, size \nreduction rate, was only 0.58. \nConclusion: SAD of 5mm is a feasible criterion for determining nodal status \non post-NAT MRI and its diagnostic performance coul d not be improved by \nmorphological features or changes before and after NAT, which suggests the \nlimited efficacy of conventional features and the u rgent need for novel features \nin the future. \nLimitations: The limitation is the lack of histopathological gol d standards for \nevery visible node on MRI. \nFunding for this study: Funding was received from the National Natural \nScience Foundation of China (82271955) and Capital' s Funds for Health \nImprovement and Research (2024-1-1022). \nEthics committee - additional information: This study was approved by the \ninstitutional review board (No. 2019KT76). \nAuthor Disclosures:  \nLi Xiao Ting: Nothing to disclose \nXin-Yue Yan: Nothing to disclose \nRui-Jia Sun: Nothing to disclose \nDing Yang: Nothing to disclose \nYing-Shi Sun: Nothing to disclose \nZhen Guan: Nothing to disclose \nQing-Yang Li: Nothing to disclose \nXiaoyan Zhang: Nothing to disclose \nQiaoyuan Lu: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 209  \nA method of matching nodes between MRI and patholog y in rectal cancer \npatients \n*Q-Y. Li*, X-Y. Yan, D. Yang, Z. Guan, R-J. Sun, Q.  Lu, L. X. Ting, X. Zhang, \nY-S. Sun; Beijing/CN \n(631213971@qq.com) \n \nPurpose or Learning Objective: To develop a method that enables node-by-\nnode matching between preoperative MRI and postoper ative pathology in \nrectal cancer patients, thereby providing reliable node-based ground-truth \nlabels for further radiological studies. \nMethods or Background: This methodological study prospectively enrolled \n535 patients (59 ± 11 years; 326 males) with rectal cancer between 2021 and \n2023. Target nodes were defined as nodal structures  with a short-axis \ndiameter (SAD) of ≥ 3 mm in the mesorectum or around the superior rect al \nartery on MRI. With relative location to the tumor,  rectal wall and mesorectal \nfascia, each target node was localized in three dir ections. Combining the \ncoordinates of each node, a 3D node map centered on  the tumor and including \nall target nodes was constructed for each patient a nd used as a bridge \nenabling node-by-node matching between MRI and path ology. \nResults or Findings: 3,038 target nodes were detected on preoperative MR I, \nof which 2,220 (73.1%) achieved matched between MRI  and pathology. \nOf the 1,707 matched benign nodes, 1,321 (77.4%), 3 78 (22.1%), and 8 (0.5%) \nhad SADs of <5 mm, 5-9 mm and >9 mm, respectively. Whereas of the 513 \nmatched metastatic nodes, 224 (43.7%), 254 (49.5%),  and 35 (6.8%) had \nSADs of <5 mm, 5-9 mm and >9 mm, respectively. Pati ents with lower \nmatching rates tended to have higher T-stages and m ore target nodes on MRI, \nwhereas other factors, e.g. the BMI, therapeutic re gimen, tumor location, and \ntime interval between MRI and pathological examinat ion did not show \nsignificant effect on the matching accuracy. \nConclusion: A matching method between MRI and pathology was dev eloped \nto label numerous nodes with precise statuses in re ctal cancer patients, which \ncontributes to future radiological studies. \nLimitations: The limitation is the small proportion of metastati c nodes \ncompared to benign ones. \nFunding for this study: Funding was received from the National Natural \nScience Foundation of China (82271955) and Capital' s Funds for Health \nImprovement and Research (2024-1-1022). \nEthics committee - additional information: This study was approved by the \ninstitutional review board (No. 2019KT76). \nAuthor Disclosures:  \nLi Xiao Ting: Nothing to disclose \nXin-Yue Yan: Nothing to disclose \nRui-Jia Sun: Nothing to disclose \nDing Yang: Nothing to disclose \nYing-Shi Sun: Nothing to disclose \nZhen Guan: Nothing to disclose \nQing-Yang Li: Nothing to disclose \nXiaoyan Zhang: Nothing to disclose \nQiaoyuan Lu: Nothing to disclose \n \n \nPrognostic impact of MRI-detected risk factors in t otal neoadjuvant \ntherapy for locally advanced rectal cancer \n*G. F. Cicala*, S. Parisi, F. De Cobelli, V. Burgio , M. Ronzoni, R. Rosati,  \nU. Elmore, P. Passoni; Milan/IT \n(cicala.giuseppe@hsr.it) \n \nPurpose or Learning Objective: In Locally Advanced Rectal Cancer (LARC), \nTotal Neoadjuvant Therapy (TNT) has shown to be a v alid therapeutic option \nleading to a significant reduction in recurrence, d istant metastases, and an \nimprovement in disease-free survival. Magnetic reso nance imaging (MRI) plays \na crucial role in the detection of extramural venou s invasion (EMVI), tumor \ndeposits (TDs), mesorectal fascia invasion (MFI) an d helps identify patients \nwho may benefit from a TNT. \nMethods or Background: A retrospective analysis was performed on \nprospectively collected data involving 109 patients  who received TNT between \n2009 and 2022. MRI scans were conducted both before  and after TNT, with a \nfocus on EMVI, TDs, MFI. Following TNT, viable EMVI  and TDs were \nevaluated using a standardized five-point Likert sc ale. \nResults or Findings: Among the 109 patients, 95 patients met the inclusi on \ncriteria, 64.2% were male, with a median age of 60. 3 years. Positive EMVI \nscores were observed in 47.4% of cases, while TDs a nd MFI in 15.7% and \n24.7%, respectively. Positive EMVI, TDs, MFI, and L ikert scores significantly \ncorrelated with reduced time to progression (TTP) a nd poorer overall survival \n(OS). In multivariate analysis, the presence of TDs  (HR 10.28, p=0.002) and a \nLikert score of 4 (HR 16.44, p=0.003) were strong i ndependent predictors of \nshorter TTP. \nConclusion: This study confirms the impact of MRI-detected risk  factors as \nsignificant predictors of prognosis, supporting the  role of a standardized five-\npoint Likert scale in the stratification of patients for personalized treatment. \nLimitations: Treatment changes after TNT can lead to false negat ives, \nnecessitating a multidisciplinary evaluation for pa tients. Larger cohorts are \nneeded to improve the Likert scale's reliability. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Pending confirmation from the \nethics committee \nAuthor Disclosures:  \nGiuseppe Franco Cicala: Nothing to disclose \nValentina Burgio: Nothing to disclose \nSergio Parisi: Nothing to disclose \nMonica Ronzoni: Nothing to disclose \nRiccardo Rosati: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \nPaolo Passoni: Nothing to disclose \nUgo Elmore: Nothing to disclose \n \n \nEvaluation of diffusion-weighted imaging in predict ing response in \nlocally advanced rectal cancer \n*P. N. Franco*, C. Maino, C. R. G. L. O. M. Talei F ranzesi, R. Corso,  \nD. Ippolito; Monza/IT \n(francopaoloniccolo@gmail.com) \n \nPurpose or Learning Objective: To assess the performance of Diffusion-\nWeighted Imaging (DWI) and apparent diffusion coeff icient (ADC) values in \npredicting response to neoadjuvant chemoradiation t herapy (CRT) in patients \nwith locally advanced rectal cancer (LARC). \nMethods or Background: Ninety-four patients with MRI pre- and post-\nneoadjuvant treatment were retrospectively enrolled . Three regions of interest \n(ROIs) were manually drawn on three different tumor  slices for every DWI \nsequence. ROIs were automatically copied to the cor responding ADC maps \nand the system derived three different ADC values ( mean, maximum, and \nminimum), and the standard deviation (SD). Only mea n ADC values were \nconsidered. After surgical intervention, pTNM and M andard tumor-regression-\ngrade (TRG) were obtained. Patients with TRG 1-2 we re classified as \nresponders while patients with TRG 3-5 were classif ied as non-responders. \nResults or Findings: No correlation was found between pre-ADC values and  \nTRG classes, while post-ADC and ΔADC values showed a significant \ncorrelation with TRG classes (r= -0.285, p=0.002 an d r= -0.290, p=0.019, \nrespectively). Post-ADC values were statistically d ifferent between responders \nand non-responders (p=0.019). When considering the relation between overall \nsurvival (OS) and ADC values, pre-ADC showed a nega tive correlation with OS \n(r= -0.381, p=0.001) while a positive correlation w as found between ΔADC \nvalues and OS (r= 0.323, p=0.013). According to ΔADC values, the mean OS \ntime between responders and non-responders showed a  significant difference \n(p=0.030). A statistical difference was found betwe en TRG classes and OS \n(p=0.038) and by dividing patients into responders and non-responders \n(p=0.019). \nConclusion: The pre-ADC and ΔADC values could be used as useful \npredictors for patients' prognosis. Post-ADC values , due to their relationship \nwith TRG classes, could be a useful tool to predict  response. \nLimitations: No correlation between imaging and surgical specime ns; the \nmean interval among CRT, restaging MRI, and surgery  was variable among \npatients. \nFunding for this study: None \nEthics committee - additional information: Upon reviewing the protocol, the \nlocal ethical committee deemed formal approval unne cessary, owing to the \nretrospective, observational, and anonymous nature of this study. \nAuthor Disclosures:  \nCesare Maino: Nothing to disclose \nCammillo Roberto Giovanni Leopoldo Oreste Massimili ano Talei Franzesi: \nNothing to disclose \nRocco Corso: Nothing to disclose \nPaolo Niccolò Franco: Nothing to disclose \nDavide Ippolito: Nothing to disclose \n \n \nEfficacy of interventional transarterial treatment in locally recurrent or \nunresectable colorectal carcinoma: Therapy response  and survival \n*T. J. Vogl*, A-I. Nica, C. Booz, L. S. Alizadeh, I . Yel, T. Biciusca,  \nA. Gökduman, T. Gruber-Rouh, H. Adwan; Frankfurt/DE  \n \nPurpose or Learning Objective: To evaluate the efficacy of transarterial \nchemoperfusion (TACP) and transarterial chemoemboli zation (TACE) as \npalliative and symptomatic treatment options for un resectable colorectal \ncarcinoma (CRC) regarding local tumor response and survival. \nMethods or Background: Between January 2000 and October 2023, 318 \nTACP and 80 TACE procedures were performed in 67 pa tients with locally \nrecurrent or unresectable CRC. Forty-eight patients  were treated with TACP \n(mean 6.2 procedures/patient, range 2-22), 14 with TACE (mean 4.6 \nprocedures/patient, range 2-11) and 5 patients rece ived a combination of both \ntherapies (mean 4.5 procedures/patient, range 2-13) . Local tumor response \n\n \n \nSaturday \nAbstract-based Programme \n \n 210  \nwas retrospectively evaluated using the RECIST crit eria and overall survival \n(OS) and progression-free survival (PFS) were calcu lated using the Kaplan-\nMeier estimator. \nResults or Findings: 49 (73.13%) of the 67 patients had stable disease ( SD), \n15 (22.39%) progressive disease (PD) and 3 patients  (4.48%) partial response \n(PR). Median OS was 16.17 months, median PFS was 11 .25 months. There \nwas no statistically significant difference in OS ( P=0.598) and PFS (P=0.847) \nbetween patients either receiving TACP or TACE or b oth treatments. One year \nafter the first procedure, 27 (40.3%) patients were  still alive. Nine patients \n(13.4%) were still alive after 2 years and six pati ents (8.9%) were still alive after \n3 years. No major complications were reported. \nConclusion: CONCLUSION: TACP and TACE are minimally invasive \nprocedures that offer a treatment option for patien ts with locally recurrent or \nunresectable CRC, potentially preventing tumor prog ression and improving \nquality of life. However, their benefits in the tre atment of CRC warrant further \ninvestigation. \nLimitations: Retrospective, single-center study Varying follow-u p intervals \nShort follow-up period \nFunding for this study: No funding \nEthics committee - additional information: Approval of the ethics committee \nof the Johann Wolfgang Goethe University, Frankfurt  \nAuthor Disclosures:  \nChristian Booz: Nothing to disclose \nIbrahim Yel: Nothing to disclose \nTeodora Biciusca: Nothing to disclose \nAndreea-Ioana Nica: Nothing to disclose \nThomas J. Vogl: Nothing to disclose \nAynur Gökduman: Nothing to disclose \nTatjana Gruber-Rouh: Nothing to disclose \nHamzah Adwan: Nothing to disclose \nLeona Soraja Alizadeh: Nothing to disclose \n \n \nOpportunistic Colorectal Cancer Screening in Comput ed Tomography: \nExploration of the Colon-Liver Axis with Machine Le arning \n*S. Grosu*¹, A. Hinterberger², A. E. Sint¹, J. Rick e¹, M. Ingrisch¹, P. Wesp¹; \n¹Munich/DE, ²Heidelberg/DE \n \nPurpose or Learning Objective: Colorectal cancer (CRC) detection in non-\ndedicated computed tomography (CT) examinations wit hout bowel preparation \nis challenging. Current research suggests that live r diseases are associated \nwith an increased risk of colorectal cancer. The ai m of this study was to identify \npatients with CRC using machine learning (ML)-based  opportunistic analysis of \nthe liver in non-dedicated routine clinical CT scan s. \nMethods or Background: Patients 18 years or older with histologically prov en \nCRC or negative colonoscopy that underwent contrast -enhanced CT of the \nabdomen for various indications within 5 years or l ess to colonoscopy were \nincluded retrospectively. Patients were randomly di vided into a training set \n(75%) and a test set (25%), stratifying for age and  sex. Deep learning-based \nautomated liver segmentation on CT images was perfo rmed. Standardized \nRadiomic image features were extracted from the liv er segmentations. A \nrandom forest ML algorithm was trained on the train ing set to differentiate \nbetween patients with histologically confirmed CRC or negative colonoscopy \n(non-CRC). Algorithm performance was evaluated on t he test set using ROC-\nAUC, sensitivity and specificity. \nResults or Findings: The training set comprised 809 CT scans from 809 \npatients (mean age = 62.2 years; 42% female), 66 (8  %) with CRC. The test \nset comprised 270 CT scans from 270 patients (mean age = 62.2 years; 44% \nfemale), 23 (9 %) with CRC. The ROC-AUC for random forest-based \ndifferentiation between CRC and non-CRC patients wa s 0.62, with a sensitivity \nof 74% at a specificity of 50%. \nConclusion: Our results indicate that opportunistic analysis of  routine clinical \nCT images of the liver might have the potential to detect patients with an \nincreased risk for CRC without additional radiation  exposure or examinations. \nLimitations: Further refinement of the presented model is needed  to further \nincrease its diagnostic performance. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Ethics committe of the LMU \nUniversity Hospital, LMU Munich, Munich, Germany \nAuthor Disclosures:  \nAlena Elisabeth Sint: Nothing to disclose \nSergio Grosu: Nothing to disclose \nMichael Ingrisch: Nothing to disclose \nPhilipp Wesp: Nothing to disclose \nAnna Hinterberger: Nothing to disclose \nJens Ricke: Nothing to disclose \n \n \n \n \n \nPrognostic value of lateral lymph node metastasis i n pretreatment MRI \nfor rectal cancer in patients with neoadjuvant CRT and surgical resection \nwithout LLND: A systemic review and meta-analysis \n*T. Lee*¹, N. Horvat², M. J. Gollub², J. Garcia-Agu ilar², T. H. Kim²; ¹Seoul/KR, \n²New York, NY/US \n(lee8720@gmail.com) \n \nPurpose or Learning Objective: To systematically review and meta-analyze \nthe prognostic significance of lateral lymph node m etastasis (LLNM) on \npretreatment MRI in patients with rectal cancer who  undergo neoadjuvant \nchemoradiation followed by curative surgical resect ion without lateral lymph \nnode dissection (LLND). \nMethods or Background: We searched the MEDLINE and EMBASE \ndatabases until September 27, 2023, utilizing the f ollowing search terms: \n(rectal OR rectum OR colorectal) AND (lateral OR si dewall) AND (lymph OR \nnode). The QUIPS tool was employed to evaluate meth odological quality. We \npooled the association between LLNM on pretreatment  MRI and outcomes \nsuch as local recurrence, distant metastasis, disea se-free survival, and overall \nsurvival using hazard ratio (HR) and odds ratio (OR ) based on random effects \nmodel. \nResults or Findings: We included 9 studies, encompassing 3180 patients. \nLLNM on pretreatment MRI revealed a significant ass ociation with increased \nlocal recurrence rates (HR: 4.11; 95 % CI: [1.87, 9 .02]) and elevated risks for \nboth disease-free (HR: 1.70; 95 % CI: [1.42, 2.03])  and overall survival (HR: \n1.76; 95 % CI: [1.44, 2.15]). As for distant metast asis, our analysis indicated a \npotential trend towards increased rates, though thi s did not reach statistical \nsignificance (HR: 1.67; 95 % CI: [0.85, 3.27]). \nConclusion: Our findings underscore the relationship between LL NM and \nincreased local recurrence and compromised disease- free and overall survival. \nThis emphasizes the potential limitations of relyin g solely on neoadjuvant \nchemoradiation and highlights the potential need to  intensify treatment in select \npatients. \nLimitations: Firstly, we included a relatively small number of s tudies. Second, \na notable heterogeneity was observed in the criteri a used to define LLNM on \nMRI across different studies, which probably led to  heterogenous proportion of \nLLNM-positive patients among included studies. \nFunding for this study: The National Cancer Institute Cancer Center Core \nGrant P30 CA008748 \nEthics committee - additional information: This study was a systematic \nreview and therefore exempt from requiring approval  from our institutional \nreview board. \nAuthor Disclosures:  \nTaehee Lee: Nothing to disclose \nJulio Garcia-Aguilar: Nothing to disclose \nMarc Jeffrey Gollub: Nothing to disclose \nNatally Horvat: Nothing to disclose \nTae Hyung Kim: Nothing to disclose \n \n \nOpen magnetic field and MRI defecography \n*G. Sterlicchio*, I. Carbone, M. Rengo, C. L. Salet ti, D. Bellini; Latina/IT \n(giuseppe.sterlicchio@uniroma1.it) \n \nPurpose or Learning Objective: To investigate the diagnostic value and \nimage quality of Open MRI scanner (0.5T) on the eva luation of pelvic floor \ndisfunctions. \nMethods or Background: Twenty one patients (20 women and 1 man) \nunderwent MRI defecography using both Open MRI scan ner (0.5 T) and high \nfiled MRI scanner (1.5 T). For both scanners, the s ame MRI protocol has been \nadopted, including morphological sequences and dyna mic sequences during \nrest and defecation. Contrast-to-noise ratio (CNR),  and signal to noise ration \n(SNR) were calculated and compared among the differ ent data sets. \nQualitative assessment of image quality was perform ed by 3 readers using 5 \npoints Likert scale. \nResults or Findings: SNR was significantly lower using 0.5T compared to \n1.5T (mean value 33.1 vs 95.2 for T2 sequences and 44.8 vs. 55.6 for dynamic \nsequences; P<0.05). CNR was significantly lower usi ng 0.5T compared to 1.5T \n(mean value 37.1 vs 57.6 for T2 sequences and 27.1 vs. 51.3 for dynamic \nsequences; P<0.05). However, the readers' image qua lity scores showed that \nopen MRI scanner is not inferior to High magnetic f iled. Diagnostic accuracy \nwas the same for both scanners. \nConclusion: Despite the low values of all quantitative quality metrics on 0.5T \ncompared to i.5T, readers perception of image quali ty is the same between \nOpen MRI scanner and High filed MRI scanner. Low ma gnetic field does not \naffect diagnostic accuracy. \nLimitations: Small sample size. \nFunding for this study: None \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 211  \nEthics committee - additional information: Yes \nAuthor Disclosures:  \nIacopo Carbone: Nothing to disclose \nMarco Rengo: Nothing to disclose \nGiuseppe Sterlicchio: Nothing to disclose \nCarlos Leon Saletti: Nothing to disclose \nDavide Bellini: Nothing to disclose \n \n \nAssessment of puborectalis (PRM) and pubococcygeus (PCM) muscles \nthickness by RMI defecography: A promising adjuvant  radiological \nparameter for the identification of spastic pelvic floor syndrome(SPFS) \n*M. Mandolini*, V. Carrozzo, A. Calculli, R. Falett i, A. Ferraris; Torino/IT \n(matilde.mandolini@unito.it) \n \nPurpose or Learning Objective: To investigate the association between \nradiological suspicion of SPFS and the thickness ra tio R(PRM/PCM) in patients \nwith pelvic floor disorders. \nMethods or Background: The retrospective study involved 109 women (age \n17-86 ys, BMI 16-35.5 kg/m2, past pregnancies 76%) who underwent RMI \nDefecography between November 2021 and September 20 24 at our institution. \nThe ESGAR-ESUR recommended protocol was complemente d by \nmeasurements of PRM and PCM branches thickness made  bilaterally at half \nthe length of each bundle and averaged to obtain th e ratio R (PRM/PCM). \nContinuous independent variables were compared with  Mann-Whitney’s Test \nwhen independent; Wilcoxon’s Test and Bland-Altman plot when correlated; \ndichotomic variables were studied with Fisher’sTest . PropensityScoreMatching \n(PSM) was used to reduce the effect of confounding covariates on the \noutcome.The ReceivingOperatingCurve (ROC) was used to estimate the \nperformance of R with the Area Under The Curve (AUC ). \nResults or Findings: The 23 (21%) patients satisfying the radiological c riteria \nfor SPFS recognized by scientific literature formed  the Study Group.PSM \nextracted from the remaining 84 patients, a Control  Group of 23 with baseline \ncomparable (p>0.77, Standardized Mean Difference <0 .20) to that of the Study \npatients. The thickness ratio R was 3.2(3.0-3.5) fo r the former vs 2.5(2.2-2.75) \nfor the latter (p<0.0001). The ROC curve assessed a  good diagnostic ability of \nR for SPSF, (AUC=0.94) with threshold R≥3.0 (sensitivity 0.87, specificity 0.96, \nPPV=0.95, NPV=0.88).The measurements made by two re aders showed good \nintra- and inter-operator agreement. \nConclusion: The performance of the PRM/PCM thickness ratio is p resently \nbeing explored over a larger population; if the pos itive results will be confirmed, \nthis indicator may become an additional tool for ra diological identification of \nSPFS. \nLimitations: The main limitations of the study are the absence o f gold \nstandard method for SPFS diagnosis and the small nu mber of our Study \nGroup. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Being a retrospective preliminary \nstudy based on the use of anonymous data, it has no t yet been subjected to \nscrutiny by the ethics committee. \nAuthor Disclosures:  \nMatilde Mandolini: Nothing to disclose \nAnnarita Calculli: Nothing to disclose \nValentina Carrozzo: Nothing to disclose \nRiccardo Faletti: Nothing to disclose \nAndrea Ferraris: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n09:30-11:00 Research Stage 3 \nResearch Presentation Session: Neuro \nRPS 1811 \nBrain in function \n \nModerator \nN. Bargalló; Barcelona/ES  \n(BARGALLO@clinic.cat) \n \n \nConstructing normative human brain atlases of R1, R 2, proton density, \nand myelin volume fraction using synthetic quantita tive MRI \n*H. M. H. Sbaihat*, A. K. Roenneke, D. Müller, T. L adopoulos, R. Schneider,  \nB. Krieger, B. Bellenberg, C. Lukas; Bochum/DE \n(hasansbaihat@gmail.com) \n \nPurpose or Learning Objective: Quantitative MRI (qMRI) provides valuable \ninsights into tissue-specific MR properties, extend ing the diagnostic capabilities \nof conventional MRI. We aimed to construct normativ e multimodal human brain \natlases to serve as references for tissue alteratio ns in neurological disorders. \nMethods or Background: Fifty-eight healthy controls (HC) underwent qMRI of  \nthe brain at 1.5T using the QRAPMASTER sequence, re sulting in parameter \nmaps for Myelin Volume Fraction (MVF, %), Proton De nsity (PD, %), and \nRelaxation Rates R1 and R2 (s⁻¹). We constructed four high-resolution (1mm³) \natlases in standard space after bias-field correcti on, interpolation, \nnormalization, and smoothing. These atlases allow v isual and quantitative \ncomparisons with individual datasets and will be ma de available to the \nresearch community. Additionally, quantitative data  were extracted from 26 \nwhite matter regions of interest (ROIs). The atlase s and extracted data were \nvalidated using z-score maps of three healthy contr ols and three multiple \nsclerosis (MS) patients. Group differences across t he ROIs between the qMRI \natlases and the testing subjects were assessed usin g t-tests. \nResults or Findings: The resulting atlases demonstrated high anatomical \naccuracy, resolution, and comprehensive brain cover age. The ROI values for \nR1, R2, PD, and MVF were consistent with the publis hed literature. The z-\nscore maps, particularly for R1 and MVF, accurately  reflected individual lesion \npatterns and diffuse tissue changes in each MS pati ent. T-test results for R1, \nR2, PD, and MVF confirmed the alignment between the  HC atlases and the HC \ntesting group, while significant differences were o bserved with the MS testing \nsubjects at a p-value of 0.01. \nConclusion: We successfully generated high-resolution, multi-mo dal qMRI \natlases of the human brain, providing normative bas elines for R1, R2, PD, and \nMVF. The validation process underscores their poten tial for assessing \nmicrostructural brain alterations in individual pat ients with neurodegenerative \ndiseases. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee of the \nMedical Faculty of the Ruhr-University Bochum, Germ any (Approval Np. 20-\n7054-BR). \nArea of Interest: CNS, Neuroradiology brain. \nAuthor Disclosures:  \nRuth Schneider: Nothing to disclose \nTheodoros Ladopoulos: Nothing to disclose \nBarbara Bellenberg: Nothing to disclose \nBritta Krieger: Nothing to disclose \nHasan M H Sbaihat: Nothing to disclose \nAnna Katharina Roenneke: Nothing to disclose  \nCarsten Lukas: Nothing to disclose \nDajana Müller: Nothing to disclose \n \n \nHow does hippocampal volume in mesial temporal scle rosis affect brain \nnetworks during a functional MRI memory task? \n*S. B. Rosa*¹, B. Direito², F. Sales², D. J. Pereir a²; ¹Lisbon/PT, ²Coimbra/PT \n(saradbrosa@gmail.com) \n \nPurpose or Learning Objective: Mesial Temporal Sclerosis (MTS) is the most \nfrequent histopathological abnormality in drug-resi stant temporal lobe epilepsy. \nAnterior temporal lobectomy is a possible treatment , but postoperative memory \ndeficits may follow. MTS patients activate extra-te mporal regions during \nmemory-related tasks, possibly as part of compensat ory networks. Further \nresearch is required to understand these networks a nd how they may impact \nsurgical outcomes. We aimed to understand how hippo campal atrophy affects \nconnectivity during a memory fMRI task in MTS patie nts. \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 212  \nMethods or Background: We retrospectively included drug-resistant MTS \npatients admitted to CLHU (2019-2023) for surgical evaluation. Patients \nunderwent MRI, including 3D T1WI and event-related memory fMRI. \nHippocampal volume ratio was automatically calculat ed using volBrain. CONN \ntoolbox was used to assess the effect of hippocampa l volume on seed-to-voxel \nfunctional connectivity of the right and left hippo campus in patients with right \nand left MTS, respectively, while performing a verb al memory task. \nResults or Findings: Thirteen patients were included (eight with lMTS, f ive \nwith rMTS). Hippocampal volume ratio was 0.1047±0.0 406 for lMTS and \n0.1144±0.0331 for rMTS patients (mean ± standard deviation). Seed-to-voxel \nanalysis showed a positive association between hipp ocampal volume and \nconnectivity with contralateral hippocampus and par ahippocampal gyri in both \ngroups. In lMTS, hippocampal volume was negatively associated with \nconnectivity to right superior parietal lobule and supramarginal gyrus (p＜0.05 \ncluster-size p-FDR corrected). No significant negat ive associations were found \nin rMTS. \nConclusion: As expected, greater hippocampal atrophy correspond ed to \nreduced connectivity with the parahippocampal gyri and contralateral mesial \ntemporal regions. Surprisingly, lMTS patients with greater atrophy showed \nhigher connectivity to right parietal areas, a comp ensatory network not \npreviously described in this condition, to our know ledge. \nLimitations: Small sample size, applying a resting-state fMRI an alysis method \nto task-based fMRI, assessing right hippocampal con nectivity with a left \nhippocampal task. \nFunding for this study: No funding. \nEthics committee - additional information: All patients signed written \nconsent for data usage and all data was anonymized.  \nAuthor Disclosures:  \nFrancisco Sales: Nothing to disclose \nSara Botelho Rosa: Grant Recipient: ESR-EIBIR seed grant \nDaniela Jardim Pereira: Nothing to disclose \nBruno Direito: Nothing to disclose \n \n \nUnveiling the Brain’s Response to Valenced Sounds: Neural Correlates of \nAuditory Emotion \n*F. Aldhafeeri*; Hafar al-Batin/SA \n(aldhafeeri8@hotmail.com) \n \nPurpose or Learning Objective: Complex neural pathways that integrate \nauditory processing with emotional evaluation play a crucial role in the \nperception of emotional sounds. This study aimed to  identify the neural circuits \nthat differentially encode positive and negative va lence during the implicit \nprocessing of emotional stimuli. \nMethods or Background: A block-design fMRI experiment was conducted \nwith thirty healthy participants. The study measure d blood oxygen level \ndependent (BOLD) signal changes in response to plea sant and unpleasant \nsounds from the International Affective Digitized S ounds (IADS), with each \ncondition compared against a neutral baseline \nResults or Findings: Significant activation (pFDRcorrected <0.05) was fo und \nin the medial prefrontal cortex (mPFC), ventral ant erior cingulate cortex \n(vACC), and temporal lobe when contrasting pleasant  sounds with neutral \nconditions. In response to unpleasant sounds, signi ficant activation \n(pFDRcorrected <0.05) was observed in the amygdala,  nucleus accumbens, \nparahippocampal gyri, temporal lobe, visual cortex,  PFC, insula, anterior \ncingulate gyrus, and cerebellum, compared to the ne utral condition \nConclusion: The neural correlates of pleasant and unpleasant st imuli involve \na complex interplay between brain regions that regu late emotional responses. \nThe auditory cortex, amygdala, and nucleus accumben s are key components \nof this process, with distinct activation patterns depending on the emotional \nvalence of the auditory stimuli. Understanding thes e neural mechanisms \nenhances our insight into how sound influences emot ional experiences and \nmay guide the development of therapeutic interventi ons for auditory-related \nemotional disorders \nLimitations: The study was conducted with a relatively small sam ple size of \nthirty participants. While this is typical for many  fMRI studies, the limited \nnumber of participants may affect the generalizabil ity of the findings. A larger \nsample size could provide more robust and reliable results. \nFunding for this study: None \nEthics committee - additional information: Local Research Ethics \nCommittee \nAuthor Disclosures:  \nFaten Aldhafeeri: Nothing to disclose \n \n \n \n \n \n \n \n \nAmygdala multimodal reorganization as an indicator of affective \ndysfunction in tinnitus patients \n*Q. Chen*; Beijing/CN \n(chenqian8319@163.com) \n \nPurpose or Learning Objective: This study aimed to systematically \ninvestigate structural and functional alterations i n amygdala subregions using \nmultimodal MRI in patients with tinnitus with or wi thout affective dysfunction. \nMethods or Background: Sixty patients with persistent tinnitus and 40 heal thy \ncontrols (HCs) were recruited. Based on a questionn aire assessment, 26 and \n34 patients were categorized into the tinnitus pati ents with affective dysfunction \n(TPAD) and tinnitus patients without affective dysf unction (TPWAD) groups, \nrespectively. MRI-based measurements of gray matter  volume, fractional \nanisotropy (FA), fractional amplitude of low-freque ncy fluctuations (fALFF), \nregional homogeneity (ReHo), degree centrality (DC) , and functional \nconnectivity (FC) were conducted within 14 amygdala  subregions for \nintergroup comparisons. Associations between the MR I properties and clinical \ncharacteristics were estimated via partial correlat ion analyses. \nResults or Findings: Compared with HCs, the patients exhibited significa nt \nstructural and functional changes, with more pronou nced WMI changes in the \nTPAD group, predominantly within the left auxiliary  basal or basomedial \nnucleus (AB/BM), right central nucleus, right later al nuclei (dorsal portion), and \nleft lateral nuclei (ventral portion containing bas olateral portions). Moreover, \nthe TPAD group exhibited decreased FC between the l eft AB/BM and left \nmiddle occipital gyrus and right superior frontal g yrus (SFG), left basal nucleus \nand right SFG, and right lateral nuclei (intermedia te portion) and right SFG. In \ncombination, these amygdalar alterations exhibited a sensitivity of 65.4% and \nspecificity of 96.9% in predicting affective dysfun ction in patients with tinnitus. \nConclusion: Although similar structural and functional amygdala  remodeling \nwere observed in the TPAD and TPWAD groups, the cha nges were more \npronounced in the TPAD group. These changes mainly involved alterations in \nfunctionality and white matter microstructure in va rious amygdala subregions; \nin combination, these changes could serve as an ima ging-based predictor of \nemotional disorders in patients with tinnitus. \nLimitations: This is a cross-section study \nFunding for this study: None \nEthics committee - additional information: This study was approved by the \nInstitutional Review Board of Beijing Friendship Ho spital, Capital Medical \nUniversity (No. 2017-P2-134-01). \nAuthor Disclosures:  \nQian Chen: Nothing to disclose \n \n \nThe impact of diet on brain structural and function al networks \n*R. Rajiah*, Q. Aziz, P. Nachev, J. K. K. Ruffle; L ondon/UK \n \nPurpose or Learning Objective: To delineate the association between diet, \nbrain structure and function in patients undergoing  multi-modal MRI. \nMethods or Background: Extensive evidence from lab-based and clinical \nstudies support diet as an essential regulator of b rain function. However \naccompanying neuroimaging evidence is scarce and sm all in scale. We \nstudied 518 healthy participants from the Cambridge  Centre for Aging and \nNeuroscience (CAM-Can) repository (261 male and 257  female, mean age 53 \nyears). Participants were clustered via their self- reported dietary intakes using \nK-means clustering. From MRI, resting-state functio nal, white matter \ntractography, and grey matter volumetry networks co nditional to dietary cluster \nwere investigated using network-based statistics, w ith participant age and sex \nas nuisance covariates. Relevant clinical data comp rising BMI and Hospital \nAnxiety and Depression Score (HADS) were compared u sing Welch’s t-test. \nResults or Findings: Four dietary intake patterns were identified: balan ced, \nmeat-predominant, vegetarian, and raw vegetable-pre dominant. Specific \ndietary types – balanced, meat-predominant and vege tarian – were associated \nwith increased functional, grey matter or white mat ter connectivity between \nregions comprising the bilateral insula, frontomedi al cortex (FMC) and bilateral \nanterior cingulate cortex (ACC) (p=<0.04). A balanc ed diet was associated with \nsignificantly increased functional, structural, and  white matter connectivity \nbetween brain regions, including the orbitofrontal cortex (OFC), bilateral insula \nand FMC compared to all others (p=<0.04). A balance d diet was associated \nwith a significantly reduced depression score (HADS -D), compared to a meat-\npredominant one (p=0.03). The meat-predominant diet  cluster correlated with \nincreased structural connectivity at the bilateral ACC and increased BMI \n(p=0.01), when compared to vegetarian diets (p=0.03 ). \nConclusion: In the largest study of its kind, we reveal the str uctural and \nfunctional brain associations of diet. Our findings  imply physiological correlates \nfor diet-induced brain changes for future research.  \nLimitations: We use self-reported dietary intake, limited by dat a availability. \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 213  \nFunding for this study: The Issac Shapiro grant supported RR. JKR was \nsupported by the Medical Research Council (MR/X0004 6X/1).  PN is supported \nby the Wellcome Trust (213038/Z/18/Z) and the UCLH NIHR Biomedical \nResearch Centre. \nEthics committee - additional information: The Cam-CAN project was \napproved by Cambridgeshire Research Ethics Committe e (reference: \n10/H0308/50). \nAuthor Disclosures:  \nParashkev Nachev: Nothing to disclose \nQasim Aziz: Nothing to disclose \nRebekah Rajiah: Nothing to disclose \nJames Kelsey K Ruffle: Nothing to disclose \n \n \nCurrent European trends in the use of CTA as an anc illary test in the \ndetermination of brain death \n*H. Briody*, I. Alam, R. Bruen, P. Mchugh, P. Rohan , S. Looby; Dublin/IE \n(hayleybriody@rcsi.ie) \n \nPurpose or Learning Objective: To establish the current trends in European \nliterature on the use of computed tomography angiog raphy (CTA) as an \nancillary test in the determination of brain death and to clarify the technical \nparameters and interpretative criteria currently in  use. \nMethods or Background: Brain death is primarily a clinical diagnosis made by \nconfirming the absence of brainstem reflexes and th e presence of apnoea in \nthe setting of irreversible coma where confounding factors have been \nexcluded. However, in specific circumstances where it is not possible to \nperform the required tests to satisfy the clinical criteria, ancillary testing may be \nrequired to support the diagnosis. The radiological  study endorsed by most \nnational guidelines remains four-vessel digital sub traction angiography \nalthough there is a move towards the less invasive,  more accessible CTA. \nResults or Findings: CTA is currently endorsed as an ancillary test for the \ndetermination of brain death in multiple European c ountries including France, \nGermany, Spain, Switzerland, Poland and the United Kingdom. The Polish and \nUK guidelines advise a three-phase study consisting  of a non-contrast phase \nand post-contrast phases at 20 and 40 seconds post intravenous contrast \ninjection. Opacification of the superficial tempora l or facial artery is required to \nconfirm correct contrast administration. Cessation of cerebral circulation is \ndiagnosed when there is bilateral absence of contra st in the middle cerebral \narteries and internal cerebral veins on the second post-contrast phase. \nConclusion: CTA is emerging as a feasible ancillary test for th e diagnosis of \nbrain death and is currently endorsed by a number o f European guidelines. \nThe most widely used technique involves a three-pha se study with evaluation \nof intracranial vessels in four anatomical location s. \nLimitations: This study is limited by a paucity of literature an d lack of \nconsensus guidelines on imaging in brain death. \nFunding for this study: None. \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nPaul Mchugh: Nothing to disclose \nRichard Bruen: Nothing to disclose \nImran Alam: Nothing to disclose \nSeamus Looby: Nothing to disclose \nHayley Briody: Nothing to disclose \nPat Rohan: Nothing to disclose \n \n \nSpeed-reading-induced changes in functional brain n etworks: A \nconnectivity-based analysis \n*T. A. Walpola*, C. Yang, N. Dilhani, R. Iseki, T. Makino, T. N. Hoang,  \nC. D. Kulathilake, I. Ichiro, A. Senoo; Tokyo/JP \n(thishuliwalpola@gmail.com) \n \nPurpose or Learning Objective: Speed-reading is a salient technique used \namong children to improve fast learning skills. The  main aim of the present \nstudy is to assess the alterations of the brain’s f unctional connectivity networks \nidentified in a speed-reading-trained native Sinhal ese cohort. \nMethods or Background: A cohort of 18 healthy native Sinhala-speaking \nadults (>18yrs) who volunteered to undergo fMRI (sc anner: 3.0 T SIGNA \nPremier) were selected ((male, 11: female, 7), Age (mean; Stdv: 31;4.0) and \ndominant hand: Right). Participants read a simple S inhala novel silently inside \nthe scanner. Three scans at one-month intervals eac h were done; the first two \nwere the control scans, and the third was the train ing scan after the \nintervention of speed-reading training. Image acqui sition; fMRI sequence using \nGRE EPI (TR: 1000(ms) TE: 30(ms), characters per ta sk block: 320-350). Data \nanalysis was performed using CONN toolbox v.22.a. \nResults or Findings: The reading speed (mean (Stdv)) increased from 213 \n(87) to 712 (200) wpm in the training group. The se ed-based connectivity \n(SBC) results showed that the language network has increased significant \nconnectivity, with 57 voxels (32%) covering 20% of Right Heschl’s gyrus and \n68 (38%) voxels covering 15% of planum temporale (S ignificant increase \nthreshold: p<0.05 cluster-size p-FDR corrected, vox el threshold: p<0.001 (p-\nuncorrected)). According to the generalized psychop hysiological interactions \n(gPPI) results, the individual ROI analysis showed increased connectivity in the \nleft posterior temporal gyrus with the bilateral vi sual lateral and occipital \nnetworks meanwhile a decreased connectivity with th e visual medial network \n(p-FDR corrected <0.05). \nConclusion: The study concluded that the activation of Heschl’s  gyrus; crucial \nfor sound perception in reading, and planum tempora le; responsible for \nphonological processing highlights the importance o f some auditory processing \nregions related to quick recognition and comprehens ion of text during speed \nreading. \nLimitations: Eye-tracking is not performed \nFunding for this study: None \nEthics committee - additional information: The Ethics have been approved \nby the Ethics Review Committee of Tokyo Metropolita n University, Tokyo, \nJapan Approval No-22022 \nAuthor Disclosures:  \nNiluka Dilhani: Nothing to disclose \nChutian Yang: Nothing to disclose \nTatsuya Makino: Nothing to disclose \nThishuli Anujaya Walpola: Nothing to disclose  \nThanh Ngoc Hoang: Nothing to disclose \nRinako Iseki: Nothing to disclose \nAtsushi Senoo: Nothing to disclose \nChathura Darshana Kulathilake: Nothing to disclose \nIso Ichiro: Nothing to disclose \n \n \nIlluminating Minds: The Transformative Role of Radi ology in Mental \nHealth Diagnosis \nR. Praveenkumar, *F. Abubacker Sulaiman*, J. Lydia;  Chennai/IN \n(fasulaiman@gmail.com) \n \nPurpose or Learning Objective: This abstract examines the contributions of \nradiology to diagnosing and managing mental health disorders, emphasizing \nadvanced imaging modalities' role in understanding neurobiological \nmechanisms. \nMethods or Background: A systematic review of literature was conducted, \nfocusing on the use of magnetic resonance imaging ( MRI), computed \ntomography (CT), and positron emission tomography ( PET) in evaluating \nconditions such as schizophrenia, major depressive disorder, anxiety \ndisorders, and post-traumatic stress disorder (PTSD ). \nResults or Findings: Key findings illustrate the impact of radiology on mental \nhealth care: Structural Imaging: MRI and CT scans i dentify neuroanatomical \nchanges in mental health disorders. For example, st udies show reduced gray \nmatter in the prefrontal cortex of patients with sc hizophrenia and altered \nhippocampal volumes in those with major depressive disorder. Functional \nImaging: PET and functional MRI (fMRI) enhance unde rstanding of neural \ncircuitry. Research indicates hyperactivity in the amygdala of individuals with \nanxiety disorders and altered connectivity in the d efault mode network in \npatients with depression. Biomarker Discovery: Radi ological imaging aids in \nidentifying neurobiological biomarkers. Specific me tabolic patterns in PET \nscans may predict treatment response in major depre ssive disorder, allowing \nfor personalized therapy. Therapeutic Monitoring: I maging techniques monitor \ntreatment efficacy. In patients receiving transcran ial magnetic stimulation \n(TMS) or electroconvulsive therapy (ECT), imaging p rovides real-time \nassessments of changes in brain activity. \nConclusion: Radiology significantly enhances the understanding and \nmanagement of mental health disorders through advan ced imaging \ntechniques, improving diagnostic accuracy and facil itating personalized \ntreatment strategies. \nLimitations: The review highlights the need for standardized ima ging \nprotocols and further research to correlate neuroim aging findings with clinical \noutcomes \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nRathinamoorthy Praveenkumar: Nothing to disclose \nJohnbosco Lydia: Nothing to disclose \nFarook Abubacker Sulaiman: Nothing to disclose \n \n \nAltered Structural-Functional Coupling in Parkinson ’s Disease Patients \nwith Depression: Insights from Multimodal Brain Net work Analysis \n*M. Wang*¹, C. Tan², J. Hakumäki¹; ¹Kuopio/FI, ²Cha ngsha/CN \n \nPurpose or Learning Objective: This study integrates multimodal brain \nnetwork data to explore alterations in structural-f unctional coupling across \nmulti-scale brain networks in Parkinson’s disease p atients with depression \n(PDD). By identifying potential imaging biomarkers,  this research aims to \nenhance sustainable diagnostic processes, improving  healthcare efficiency \nthrough accurate and early detection of depression in Parkinson’s disease. \n\n \n \nSaturday \nAbstract-based Programme \n \n 214  \nMethods or Background: A total of 106 consecutive, drug-naïve Parkinson’s \ndisease patients were prospectively enrolled and di vided into PDD (n = 50, \nHAMD > 17, BDI > 10) and PDND (n = 56, HAMD < 7, BD I < 10) groups. All \nparticipants underwent diffusion tensor imaging (DT I) and resting-state MRI on \na 3.0T scanner. Whole-brain functional and structur al networks were \nconstructed. Structural-functional coupling, intra-  and inter-module connectivity, \nas well as topological parameters (e.g., clustering  coefficient, local efficiency), \nwere compared between the two groups. ROC curve ana lysis was performed \nto evaluate the diagnostic performance of these par ameters in differentiating \nPDD from PDND. \nResults or Findings: PDD patients exhibited significantly lower whole-br ain \nstructural-functional coupling compared to PDND pat ients (P = 0.017, TFCE-\ncorrected). At the module level, PDD patients showe d increased structural-\nfunctional coupling within the default mode network  (P = 0.032, TFCE-\ncorrected). Additionally, PDD patients had signific antly lower clustering \ncoefficients (P = 0.007) and reduced local efficien cy (P = 0.021) across the \nbrain's structural network. ROC analysis combining these metrics \ndemonstrated a sensitivity of 65% and a specificity  of 77.7% for distinguishing \nPDD from PDND. \nConclusion: Altered multi-scale brain network structural-functi onal coupling in \nPDD suggests potential imaging biomarkers for more sustainable and precise \ndiagnosis. These findings may help reduce unnecessa ry interventions and \noptimise healthcare resources by providing a non-in vasive tool for early \ndetection of depression in Parkinson’s disease. \nLimitations: The limitation is the relatively small sample size.  \nFunding for this study: CSC funding No. 202306370155 \nEthics committee - additional information: The study received institutional \nreview board approval, and written informed consent  was obtained from all \nparticipants. \nAuthor Disclosures:  \nChanglian Tan: Nothing to disclose \nJuhana Hakumäki: Nothing to disclose \nMin Wang: Nothing to disclose \n \n \nSystematically altered connectome gradient in patie nt with Type2-\ndiabetes mellitus: Potential effect on cognitive fu nction \n*H. Ran*, K. Huang, T. Zhang; ZunYi/CN \n(haifengran_zmu@163.com) \n \nPurpose or Learning Objective: Type2-diabetes mellitus (T2DM) is known to \naffect brain networks and cognitive function. Conne ctome gradient studies \nhave suggested a primary-to-transmodal gradient in functional brain network. \nHowever, whether this gradient structure is disrupt ed in patients with T2DM are \nstill ambiguous. The aim of this study is to invest igate connectome gradient \nalterations and its potential contribution to cogni tive function in T2DM. \nMethods or Background: We recruited resting-state functional magnetic \nresonance imaging (rs-fMRI) data of 45 participants  (24 T2DM patients and 21 \nhealthy controls) and Montreal Cognitive Assessment  Scale (MoCA) and Mini-\nMental State Examination (MMSE) were performed in p atients with T2DM. We \nstudied the related alterations in the principal an d secondary connectome \ngradient between T2DM and healthy controls at the v oxel and network levels. \nWe further examined the associations between T2DM-r elated changes of \nconnectome gradients and clinical variables. The RV R algorithm was \nemployed to assess the predictive capacity of princ ipal gradients in relation to \ncognitive function scores among patients diagnosed with T2DM \nResults or Findings: Relative to the controls, T2DM patients extended \ngradient at different network-level and voxel-level . In the principal gradient, the \nleft rolandic operculum, precuneus gradient score w as negatively correlated \nwith MMSE score and duration, respectively (r =-0.4 79, -0.481, P=0.028, \n0.027)，and the connectome gradient alterations in ventral attention network \nwas negatively correlated with MMSE and MoCA (r=-0. 484, -0.435, P=0.026, \n0.049). Patients’ principal gradient maps significa ntly predicted their MMSE (r = \n0.465, P = 0.022). \nConclusion: We reported a systematically disrupted functional g radient in \npatients with T2DM and its negative correlation wit h cognitive function. These \nfindings improve our comprehension the neurobiologi cal mechanisms that \nunderlie cognitive function and offer potential ima ging biomarkers for the \nassessment of cognitive function in T2DM. \nLimitations: The sample size of this study was relative small. \nFunding for this study: This study was supported by Intelligent Medical \nImaging Engineering Research Center of Guizhou High er Education \nInstitutions project (Grant No. Qianjiaoji [2023] 0 38). \nEthics committee - additional information: Ethics approval of this research \nwas granted by the Ethic Committee of the Affiliate d Hospital of Zunyi Medical \nUniversity[KLL-2024-512] \nAuthor Disclosures:  \nTijiang Zhang: Nothing to disclose \nKexin Huang: Nothing to disclose \nHaifeng Ran: Nothing to disclose \n \n \nBrain Structural Connectivity alteration and its ro le in Language \nprocessing in Post-stroke Aphasia \n*N-T. Hoang*, T. Hada, C. D. Kulathilake, T. A. Wal pola, N. Dilhani, A. Senoo; \nTokyo/JP \n(hnthanh@huemed-univ.edu.vn) \n \nPurpose or Learning Objective: Our main objectives are to evaluate the brain \nconnectivity in post-stroke aphasia (PSA) and its c orrelation with language \nability. \nMethods or Background: Tractography of 20 PSA was reconstructed by \nusing deterministic algorithms. Brodmann atlas was used to brain parcellation. \nConnectivity analysis with following network measur es: density (den), \nclustering coefficient (Cc), transitivity (trans), characteristic path length (CPL), \nsmall worldness (Sw), global efficiency (G_eff), ra dius (r), diameter (d), \nassortativity coefficient (A_coeff), rich-club coef ficient (club_coeff) were \nextracted. All statistical analysis were performed using SPSS version 27. \nResults or Findings: There are significantly lower value of CPL (p = 0.0 46), \nA_coeff (p = 0.026), diameter (p = 0.031), and high er value of G_eff (p = 0.046) \nin the high-level repetition group (n = 6) compared  to the low-level repetition \ngroup (n = 14). This findings indicate the brain ne tworks of the high-level group \nare more efficient in transferring information. Add itionally, the network \nconnections in the high-level group are more random  or diverse regarding \nnode characteristics, whereas networks in the low-l evel group may display \nmore organized connections between similar nodes. T here are strong positive \ncorrelations between the speaking sentence repetiti on and d (r = 0.593, p = \n0.006), Sw (r = 0.580, p = 0.007), G_eff (r = 0.566 , p = 0.009), and strong \nnegative correlation between the speaking sentence repetition and CPL (r = \n0.571, p = 0.009). \nConclusion: The high-level repetition group transfers informati on across \nregions more efficiently than the low-level group. It indicates the potential for \nrehabilitation strategies and emphasizing the need for personalized \napproaches in aphasia treatment. \nLimitations: This research have some limitations. Our small samp le cannot \nrepresent for PSA population. Additionally, the lac k of longitudinal research is \ninsufficient to determine the impact of network con nectivity on language \nrecovery. \nFunding for this study: Research grant was provided by the Tokyo Global \nPartner Scholarship Program \nEthics committee - additional information: This study is being approved by \nthe ethics committee \nAuthor Disclosures:  \nNiluka Dilhani: Nothing to disclose \nThishuli Anujaya Walpola: Nothing to disclose \nNgoc-Thanh Hoang: Nothing to disclose \nAtsushi Senoo: Nothing to disclose \nChathura Darshana Kulathilake: Nothing to disclose \nTakuya Hada: Nothing to disclose \n \n \nExploring brain region changes following acute caff eine intake based on \nASL and OEF \n*Z. Shu*; Zhejiang, Hangzhou/CN \n(cooljuty@hotmail.com) \n \nPurpose or Learning Objective: This study aims to employ Arterial Spin \nLabeling (ASL) and Oxygen Extraction Fraction (OEF)  to assess the impact of \ncaffeine consumption on local cerebral blood flow p erfusion and brain oxygen \nmetabolism. \nMethods or Background: A prospective study was conducted involving 18 \nhealthy young volunteers with no habitual caffeine consumption. ASL and OEF \nimages were acquired both prior to and 90 minutes f ollowing the administration \nof 200 mg of caffeine. Voxel-based analysis was emp loyed to quantify \nalterations in brain oxygen metabolism across vario us cerebral regions before \nand after caffeine consumption. Regional cerebral b lood flow perfusion was \nassessed in areas exhibiting changes in brain oxyge n metabolism, with a focus \non evaluating the extent of perfusion variation sub sequent to caffeine intake. \nAdditionally, the correlation between OEF indices i n regions with modified \ncerebral blood flow and shifts in Karolinska Sleepi ness Scale (KSS) scores \nwas analyzed. \nResults or Findings: In regions exhibiting alterations in brain oxygen \nmetabolism following caffeine consumption, signific ant reductions in cerebral \nblood flow perfusion were identified within the med ial occipital-temporal areas \nof both hemispheres, the right anterior cingulate g yrus, the left postcentral \ngyrus, and the left cerebellum (P<0.05). Notably, o nly the left cerebellum \ndemonstrated an increase in brain oxygen metabolism . A negative relationship \nbetween brain oxygen metabolism and cerebral blood flow perfusion in the left \ncerebellum after caffeine intake (r=-0.5, p<0.05). Furthermore, a negative \ncorrelation was observed between changes in brain o xygen metabolism and \nshifts in KSS scores before and after caffeine cons umption (r=-0.738, p<0.05). \nConclusion: The increase in brain oxygen metabolism in the left  cerebellum \nmay compensate for the reduced cerebral blood flow perfusion following acute \ncaffeine intake, which could be a key factor in red ucing drowsiness. \n\n \n \nSaturday \nAbstract-based Programme \n \n 215  \nLimitations: The research sample size only focuses on young peop le \nFunding for this study: The work was supported by the National Natural \nScience Foundation of China (Grant No.82101983) \nEthics committee - additional information: Ethics Committee of Zhejiang \nProvincial People's Hospital \nAuthor Disclosures:  \nZhenyu Shu: Nothing to disclose \n \n \n09:30-11:00 Research Stage 4 \nResearch Presentation Session: Chest \nRPS 1804 \nTechnological advancements in chest \nimaging: MRI, photon counting CT and \nmore \n \nModerator \nP. Ciet; Rotterdam/NL  \n(p.ciet@erasmusmc.nl) \nAuthor Disclosures:  \nPierluigi Ciet: Advisory Board: European Medicine A gency (EMA); Board \nMember: European Society of Pediatric Radiology (ES PR) Research \nCommittee, AI and Cardiothoracic Taskforces; Consul tant: Siemens \nHealthineers, Vertex Pharmaceutical; Employee: Eras mus MC - Sophia \nChildren's Hospital; Grant Recipient: Research Dutc h Council, Horizon \nPathfinder; Research Grant/Support: General Electri cs; Speaker: ECR, ESPR, \nERS, InSpIrEd. \n \n \nCharacterization of Interstitial Lung Abnormalities  and Prediction of \nDisease Progression with MRI \n*D. Kütting*, J. A. Luetkens, D. Thomas, A. M. C. B oehner, T. Dell, A. Faron; \nBonn/DE \n(daniel@kuetting.de) \n \nPurpose or Learning Objective: Interstitial lung abnormalities (ILA) impact \nsurvival and quality of life, yet predictive imagin g markers for progression are \nlacking. MRI holds promise in enhancing ILA phenoty ping, potentially enabling \ntailored follow-up strategies while reducing repeti tive CT imaging and \nassociated radiation exposure. \nMethods or Background: Assessment of detectability and estimation of \ndisease progression of ILA in a single center, lung  cancer screening cohort \n(224 participants) receiving same day CT/MRI. Radio logists independently \nevaluated chest images for presence of ILA using st andardized criteria. Follow-\nup exams were reviewed for disease progression. MRI  sequences included T2-\nTSE MVXD, T2-STIR, and diffusion-weighted imaging ( DWI). Statistical \nanalyses evaluated the agreement between CT and MRI  findings and MRI's \ndiagnostic performance for ILA detection and progre ssion prediction. \nResults or Findings: Among the 224 participants (mean age 58.5 ± 5.7 yea rs; \n45% female), 26 exhibited ILA on baseline CT, with 65% categorized as \nsubpleural fibrotic. Baseline CT findings served as  the reference standard. MRI \ndetected ILA in 30 cases, 20 of which were congruen t with CT findings, \nyielding a sensitivity of 76.9% and a specificity o f 94.9% (McNemar's test, \np=0.3173). MRI detected ILA in 19/26 cases using T2 -TSE MVXD, 20/26 using \nT2-STIR (7/20 with hyperintense signal), and 6/26 u sing DWI (3/6 with \nhyperintense signal). Seven subjects showed progres sive disease on follow-\nup, with 6 of the subjects initially having a subpl eural fibrotic pattern. \nHyperintense signals in STIR and DWI sequences pred icted progression, with \nhazard ratios of 6.79 and 5.43, respectively. The c ombination of hyperintense \nsignals in STIR and DWI had a positive predictive v alue of 100%. \nConclusion: MRI reliably detected ILA and predicted disease pro gression, \nparticularly in the fibrotic subtype. MRI offered v aluable insights for ILA \nmonitoring and phenotyping, potentially improving p atient management and \nreducing radiation exposure. \nLimitations: Limited amount of patients \nFunding for this study: No funding \nEthics committee - additional information: University Hospital Bonn \nAuthor Disclosures:  \nJulian Alexander Luetkens: Nothing to disclose \nTatjana Dell: Nothing to disclose \nAnton Faron: Nothing to disclose \nDaniel Thomas: Nothing to disclose \nAlexander Marc Christian Boehner: Nothing to disclo se \nDaniel Kütting: Nothing to disclose \nVisualized quantitative evaluation of regional vent ilation and perfusion in \npatients with COPD using MRI phase-resolved functio nal lung imaging \n(PREFUL) \nZ. M. Xie, X. Gao, J. Gu, Z. Zhang, H. Yu, *L. Zhu* ; Shanghai/CN \n \nPurpose or Learning Objective: To evaluate the clinical value of phase-\nresolved functional lung imaging(PREFUL)in diagnosi ng and regional \nspecificity assessment of ventilation and perfusion  status of chronic obstructive \npulmonary disease (COPD) patients of different seve rity. \nMethods or Background: 100 healthy volunteers, 40 patients with COPD(18 \nas GOLD1, 10 as GOLD2, 7 as GOLD3, and 5 as GOLD4 ) underwent MRI \nusing 3D PREFUL under free breathing at 3.0 T(Free- breathing 1H MRI \nacquisition, no contrast agent administration). The  PREFUL postprocessing \nmethod was used for the extraction of dynamic perfu sion and ventilation \nparameters. Mean ventilation and perfusion maps, ve ntilation flow-volume \nloops(FVL) correlation, ventilation defect percenta ge(VDP), perfusion defect \npercentage (QDP), map of ventilation/perfusion defe cts (V/Q defects), and \nmatched defect percentage on both perfusion and ven tilation maps (VQM) \nwere calculated. \nResults or Findings: Compare to the homogenous ventilation and perfusion  \nmaps of healthy volunteers, COPD patients showed si gnificant heterogeneity. \nThe mean ventilation and perfusion percentage in CO PD patients were \nsignificantly lower than the healthy volunteers (P< 0.01), while the FVL is \nstatistically higher in COPD patients (P<0.01). The  ventilation map showed \nregional differences in visual agreement with emphy sema on CT and all 3D \nPREFUL-derived ventilation parameters correlated wi th FEV1 and FEV1/FVC \nin the patients with COPD(all P<0.05). Besides, our  data showed a trend of \ncorrelation between different GOLD grades and VDP ( P<0.05) in the COPD \npatients, but no significant difference in QDP amon g groups. Compared to \npatients with low-grade COPD (GOLD 1-2), severe COP D (GOLD 3-4) had \nhigher VQM which indicated a better consistency of regional defect in \nventilation and perfusion maps(P<0.01). \nConclusion: MRI PREFUL plays a promising role in evaluating the  severity of \nCOPD and visually predicting regional ventilation a nd perfusion defect in 3D \nlung imaging. \nLimitations: The study included a relatively small sample size o f COPD \npatients \nFunding for this study: This work was supported in part by the National \nNatural Science Foundation of China (8207070786), Y oung Scientists Fund of \nthe National Natural Science Foundation of China (8 2302188), Shanghai \nPujiang Program (22PJD069). \nEthics committee - additional information: Shanghai Chest Hospital ethics \ncommittee \nAuthor Disclosures:  \nZi Ming Xie: Nothing to disclose \nJunfeng Gu: Nothing to disclose \nZhengqi Zhang: Nothing to disclose \nXiaokun Gao: Nothing to disclose \nHong Yu: Nothing to disclose \nLin Zhu: Nothing to disclose \n \n \nAI-Enhanced 3D Gradient Echo MRI: An Alternative fo r Lung Nodule \nDetection and Assessment \nA. W. Marka¹, M. Steinhardt¹, M. Graf¹, L. Rahn¹, K . Weiss², M. R. Makowski¹, \nD. C. Karampinos¹, J. Gawlitza¹, *S. Ziegelmayer*¹;  ¹Munich/DE, ²Hamburg/DE \n \nPurpose or Learning Objective: Recent years have seen significant progress \nin pulmonary MR imaging for lung nodule detection t hrough optimization and \nnew sequences. This study evaluates the capabilitie s of a 3D gradient echo \nMRI sequence for detection and classification of pu lmonary nodules, \nspecifically in relation to the Lung CT Screening R eporting and Data System \n(Lung-RADS). \nMethods or Background: In this prospective trial, patients with benign and  \nmalignant lung nodules admitted between December 20 21 and July 2024 \nunderwent low-dose chest CT and pulmonary MRI using  a 3D gradient echo \nsequence, accelerated by parallel imaging, compress ed sensing, and deep \nlearning (CSAI). Three radiologists (4, 9, and 10 y ears of experience), blinded \nto clinical information, independently evaluated th e MR images. Nodule \ndetection, characterization (size, morphology), and  Lung-RADS assessment \nwere performed for all patients. To quantify interr eader agreement, intraclass \ncorrelation coefficient (ICC) for nodule measuremen ts and Cohen’s kappa for \nLung-RADS classifications were calculated. \nResults or Findings: A total of 75 patients (mean [SD] age, 65±12 years;  33 \nwomen [44%]) with 135 pulmonary nodules were includ ed and analyzed. \nNominal scan time was 3:53 min. The CSAI sequence a chieved a detection \nrate of 96,3%, with 5 missed nodules all being ≤4mm. The mean nodule \ndiameter for MRI deviated from CT by 0.1 mm (1.96 S D5.87mm; -1.96 SD-\n5.67mm). Nodule size for CT and MRI showed excellen t inter-rater agreement \n(ICC-CT: 0.995, CI95: 0.993, 0.996; ICC-CSAI: 0.993 , CI95: 0.991, 0.995). \nLung-RADS category agreement between CT and MRI was  almost perfect for \n\n \n \nSaturday \nAbstract-based Programme \n \n 216  \nReader 2 (k=0.86) and Reader 3 (k=0.90), while Read er 1 showed substantial \nagreement (k=0.69). \nConclusion: Pulmonary MRI with an accelerated 3D gradient echo sequence \nshowed high detection rates for pulmonary nodules w ith comparable Lung-\nRADs scores and morphological assessments to CT. \nLimitations: -Heterogeneous cohort \n-No follow-up scans \nFunding for this study: None \nEthics committee - additional information: It was approved by the local \nethical review board (protocol number 692/21S). \nAuthor Disclosures:  \nMarkus Graf: Nothing to disclose \nMarcus R. Makowski: Nothing to disclose \nAlexander Wolfgang Marka: Nothing to disclose \nSebastian Ziegelmayer: Nothing to disclose \nKilian Weiss: Nothing to disclose \nLeonie Rahn: Nothing to disclose \nDimitrios C. Karampinos: Nothing to disclose \nJoshua Gawlitza: Nothing to disclose \nManuel Steinhardt: Nothing to disclose \n \n \nConjugate Gradient and Deep Learning Reconstruction s: Utility for Lung \nMRI with Ultra-Short TE to Reduce Acquisition Time with Keeping Image \nQuality and Nodule Detection Capability \n*Y. Ohno*, H. Nagata, T. Ueda, M. Nomura, T. Yoshik awa, D. Takenaka,  \nY. Ozawa; Toyoake/JP \n(yohno@fujita-hu.ac.jp) \n \nPurpose or Learning Objective: To determine capability of Conjugate \ngradient reconstruction (CG-recon) and deep learnin g reconstruction (DLR) for \nreducing acquisition time with keeping image qualit y and nodule detection \nperformance on UTE-MRI. \nMethods or Background: 35 patients with lung nodule underwent UTE-MRI \nobtained with CG-recon and grid-reconstruction (Gri d-recon) by original \n(UTEoriginal), 1/2 (UTE1/2) and 1/4 (UTE1/4) sampli ng spoke numbers at 1.5T \nand 3T systems. Then, each UTE-MRI was reconstructe d with and without \nDLR. Standard protocol in this study was UTEorigina l obtained by Grid-recon \nand reconstructed without DLR. In each patient, sta ndard reference for nodule \nwas determined by thin-section CT. To determine the  influence of sampling \nspoke number reduction and reconstruction method di fferences, signal-to-\nnoise ratios (SNRs) of lung and nodule, overall ima ge quality and nodule \npresence probability were assessed by ROI measureme nts or 5-point scales. \nSNRs and overall image quality were compared betwee n each UTE-MRI and \nstandard protocol by Student’s t-test or Wilcoxon’s  signed rank test. Then, \nROC analysis was performed to compare nodule detect ion capability between \neach UTE-MRI and standard protocol. \nResults or Findings: DLR was significantly improved SNRs of all \nUTEoriginals and UTE1/2 obtained by CG-recon as com pared with standard \nprotocol (p<0.05). Overall image qualities of each UTE1/4 and all UTE1/2s \nexcept that obtained by CG-recon and reconstructed with DLR were \nsignificantly lower than that of standard protocol (p<0.05). Area under the curve \n(Az) of standard protocol (Az=0.97) was significant ly larger than that of all \nUTE1/4s (0.82<Az<0.92, p<0.0001) and UTE1/2 obtaine d by Grid-recon and \nreconstructed without DLR (Az=0.94, p=0.03), althou gh it was significantly \nsmaller than that of UTEoriginals obtained by CG-re con (Az=0.98, p<0.05). \nConclusion: CG-recon and DLR can reduce acquisition time withou t \ndegradation of image quality and nodule detection o n UTE-MRI. \nLimitations: Limited study population and nodule numbers \nFunding for this study: Canon Medical Systems Corporation \nEthics committee - additional information: Fujita Health University Hospital \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Smoking Re search Foundation \nResearch/Grant Support: Grant-in-Aid for Scientific  Research from the \nJapanese Ministry of Education, Culture, Sports, Sc ience and Technology \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grant-in-Aid  for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Canon Medi cal Systems \nCorporation Research/Grant Support: Grants-in-Aid f or Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Canon Medic al Systems \nCorporation Research/Grant Support: Smoking Researc h Foundation \n \n \n \n \nVisual analysis of dynamic oxygen-enhanced MRI (OE- MRI): Comparison \nwith V/Q SPECT and MR perfusion in chronic thromboe mbolic pulmonary \nhypertension \n*G. Agarwal*¹, D. Gopalan¹, M. Naik¹, N. Soneji¹, B . Statton¹, B. Ariff¹,  \nM. Tibiletti², G. Parker², S. Copley¹; ¹London/UK, ²Manchester/UK \n(girija.agarwal@nhs.net) \n \nPurpose or Learning Objective: Chronic thromboembolic pulmonary \nhypertension (CTEPH) is an under-recognized conditi on associated with \nsignificant morbidity yet is surgically treatable. Oxygen-enhanced MRI (OE-\nMRI) is an emerging tool for quantifying and mappin g regional gas delivery and \nuptake without the need for hyperpolarized gas. Thi s study evaluates the \naccuracy of visual analysis of OE-MRI for diagnosin g CTEPH compared to the \ncurrent standard of care: V/Q SPECT and MR perfusio n (MR-P). \nMethods or Background: Prospective study conducted from 2018 to 2023. \nParticipants with a clinical suspicion of CTEPH und erwent T1-weighted \ndynamic OE-MRI on a 1.5T scanner, V/Q SPECT and MR- P. Four consultant \nradiologists (2 MRI, 2 NM specialists) scored relev ant scans independently \n(positive, negative, or indeterminate) for CTEPH bl inded to clinical data with \ndifferences resolved by consensus. The reference st andard was the \nmultidisciplinary team diagnosis of CTEPH. \nResults or Findings: A total of 58 patients were included, with 49 under going \nboth OE-MRI and V/Q SPECT, and 48 undergoing all th ree modalities. Studies \nconsidered indeterminate for CTEPH were excluded fr om sensitivity and \nspecificity analyses (2/49 V/Q, 1/48 MR-P, and 1/49  OE-MRI). The sensitivity \nof OE-MRI, V/Q SPECT and MR-P were 0.932 (95% CI 0. 78-0.981), 0.964 \n(95% CI 0.823-0.994) and 0.931 (95% CI 0.78-0.931) respectively. Specificities \nwere 0.789 (95% CI 0.567-0.915), 0.947 (95% CI 0.75 4-0.991) and 0.833 (95% \nCI 0.608-0.942) respectively. There was no statisti cally significant difference \nbetween OE-MRI and V/Q or OE-MRI and MR-P using McN emar test (P > .05). \nConclusion: Visual analysis of OE-MRI maps is a valuable adjunc t in \ndiagnosing CTEPH with similar sensitivity but lower  specificity than VQ SPECT \nand MR-P. Further analysis of quantitative data is required to fully assess the \nrole of this technique. \nLimitations: Limitations are a relatively small sample size (n=5 8) and only \nvisual (not quantitative) analysis. \nFunding for this study: Study funded by NIHR Imperial Biomedical Research \nCentre (BRC) grant \nEthics committee - additional information: Informed consent obtained from \nall patients for the extra oxygen-enhanced MRI sequ ence, the other \ninvestigations were standard of care. \nAuthor Disclosures:  \nDeepa Gopalan: Nothing to disclose \nSusan Copley: Nothing to disclose \nMitesh Naik: Nothing to disclose \nGeoff Parker: Employee: Bioxydyn Ltd Board Member: Quantitative Imaging \nLtd and Bioxydyn Ltd Shareholder: Quantitative Imag ing Ltd and Bioxydyn Ltd \nBen Statton: Nothing to disclose \nGirija Agarwal: Nothing to disclose \nMarta Tibiletti: Board Member: Bioxydyn Limited Sha reholder: Quantitative \nImaging Ltd Employee: Bioxydyn Limited \nNeil Soneji: Nothing to disclose \nBen Ariff: Nothing to disclose \n \n \nEarly clinical experiences for chest imaging with a  new photon counting \nCT system combining cadmium zinc telluride detector s and super-high \nresolution deep-learning image reconstruction \nS. S. Schalekamp, L. J. Oostveen, M. Simmelink, W-J . Van Der Woude,  \nP. P. P. Van Der Tol, M. Prokop, *E. J. Smit*; Nijm egen/NL \n(ewoudsmit@gmail.com) \n \nPurpose or Learning Objective: To assess the image quality of chest scans \nacquired using a photon-counting CT (PCCT) scanner with cadmium-zinc-\ntelluride (CZT) detectors and super-high resolution  deep-learning image \nreconstruction (SHR-DLR). \nMethods or Background: We analyzed the chest images from two \nconsecutive cohorts of 18 and 25 patients who under went imaging on a \nprototype PCCT scanner for various indications. Ima ges were reconstructed \nusing both normal resolution (NR: 0.62mm sections, 512-matrix, hybrid-\niterative-reconstruction) and super-high resolution  (SHR: 0.21mm sections, \n1024-matrix, deep-learning-reconstruction) protocol s. An experienced chest \nradiologist assessed image quality using a 5-point scale (poor to excellent) for \noverall quality, sharpness, detail visibility, nois e, and artifacts. A homogeneous \nregion in the left ventricle was used to measure im age noise. The number of \nvisible bronchial branching generations was quantif ied in three lung regions: \nthe upper-right-lobe(1R), the upper-left-lobe(5L), and the right-lower-lobe(10R). \nBronchus volumes were automatically calculated (fir st cohort only). Statistical \nsignificance was determined using a signed rank tes t (p<0.05). \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 217  \nResults or Findings: The SHR-DLR images were rated to have higher overal l \nimage quality (4.7 vs 3.6), image sharpness (4.8 vs  3.3), detail visibility (4.7 vs \n3.6) compared to the NR images, while having lower perceived image noise \n(4.3 vs 3.3; all p<0.01). No image artifacts were o bserved with either protocol \n(both 4.0). Although measured noise levels were sim ilar between SHR-DLR \n(33.2 HU) and NR (34.2 HU), SHR-DLR images demonstr ated a finer noise \ntexture. SHR-DLR enabled visualization of finer bro nchial details, with 1.2 more \nperipheral branches visible on average (p<0.01). Au tomatic segmentation \nshowed larger bronchus volumes in the SHR-DLR image s (54.1 mL) compared \nto the NR images (47.2 mL;p<0.01). \nConclusion: PCCT with CZT detectors and SHR-DLR reconstruction provides \nexcellent spatial resolution and superior visualiza tion of the bronchial system \nwhile maintaining low image noise. \nLimitations: No comparison to conventional CT. \nFunding for this study: None \nEthics committee - additional information: Waived \nAuthor Disclosures:  \nPieternel P P Van Der Tol: Nothing to disclose \nSteven S Schalekamp: Nothing to disclose \nLuuk J. Oostveen: Nothing to disclose \nMathias Prokop: Research/Grant Support: Canon Medic al Systems Speaker: \nCanon Medical Systems \nWillem-Jan Van Der Woude: Nothing to disclose \nMirte Simmelink: Nothing to disclose \nEwoud J. Smit: Speaker: Canon Medical Systems \n \n \nFeasibility study on Photon-Counting CT-derived vir tual non-contrast \nimages substitute for true non-contrast images \nL. Lei, *Y. Zhou*; Zhengzhou City/CN \n(zyh921209zyh@163.com) \n \nPurpose or Learning Objective: To explore the feasibility of the virtual non-\ncontrast images derived from Photon-Counting CT (PH CT) substitute for true \nnon-contrast images. \nMethods or Background: 40 patients underwent pre-and arterial-venous dual-\nphase post-contrast chest imagining on a PHCT and h ad previously undergone \na chest CT with a standard energy-integrating detec tor system (EID-CT) \nscanner were retrospectively included in this study . The images were \nretrospectively analyzed. The arterial VNC images ( VNC-A) and venous VNC \nimages (VNC-V) were derived from raw datasets using  dedicated software \nrespectively. Two radiologists assessed image quali ty using a five-point Likert \nscale and performed measurements of vessels and lun g parenchyma for \nsignal-to-noise ratio (SNR), contrast-to-noise rati o (CNR), and in the case of \nsolid lung masses-to-lung parenchyma contrast ratio . \nResults or Findings: The image noise of all tissues among the four kinds  of \nimages had significant differences, images noise of  VNC-A and VNC-V images \nwere lower than TNC images and EID-CT images (P ＜0.05), and VNC-V \nimages had the highest SNR and CNR. Good equivalenc e between VNC and \nTNC images was observed in all relevant tissues wit h Bland-Altman analysis. \nImage quality subjective scoring of EID-CT, TNC, VN C-A, and VNC-V were \n5.00(1.00), 5.00(1.00), 5.00(0.75), 5.00(1.00), res pectively which had no \nsignificant differences (P=0.20). \nConclusion: The VNC image derived from PHCT enhanced image migh t be \nused as a substitute for the TNC image, and the ima ge quality is higher than \nconventional EID-CT images. \nLimitations: Not applicable \nFunding for this study: Not applicable \nEthics committee - additional information: This study was approved by the \nHuman Research Ethics Committee. \nAuthor Disclosures:  \nLimin Lei: Nothing to disclose \nYuhan Zhou: Nothing to disclose \n \n \nQuantitative Lung Imaging using Ultra High-Resoluti on Spectral \nCapabilities of CZT-based Photon-Counting Detector CT: A Feasibility \nStudy \n*S. Sharma*¹, S. Ross¹, T. Labno¹, R. Zhang¹, X. Zh an¹, R. Thompson², Z. Yu¹, \nA. Pourmorteza³; ¹Vernon Hills, IL/US, ²Cleveland, OH/US, ³Atlanta, GA/US \n(shsharma@mru.medical.canon) \n \nPurpose or Learning Objective: To evaluate ultra-high-resolution spectral \ncapabilities of CZT-based photon-counting detector CT (PCD-CT) for \nquantitative lung imaging. \nMethods or Background: A COPDGene2 phantom, with three reference \nfoams (20-lb, 12-lb, and 4-lb with HU-120kVp of -70 3, -824, and -937) and \nairways (inner-diameter (ID): 2.5-6 mm, wall-thickn ess (WT): 0.4-1.5 mm), was \nscanned on a CZT-based PCD-CT (120 kVp, 0.4x0.5 mm focal-spot, and  \n \n \nCTDIvol =12.8 mGy). Scans were reconstructed in nor mal-resolution (NR) and \nultra-high-resolution (UHR) spectral modes (pixel s ize (PS): 0.125 mm, slice \nthickness (ST): 0.2 and 0.6 mm, respectively), with  a lung kernel (FC52) and \niterative denoising, followed by generation of 40-1 50 keV VMIs. For evaluation, \nthe following were quantified: (1) HU bias between UHR and NR VMIs (ST=3.0 \nmm), (2) contrast (C) and contrast-to-noise ratios (CNRs) for ground-glass \nnodules (GGNs) and emphysema (ES) (ST=3.0 mm) (20-l b, 12-lb, and 4-lb \nfoams were surrogates for GGN, normal lung, and ES) , and (3) IDs and WTs in \nVMI with maximum CNR(GGN) and CNR(ES) (ST=0.6 mm). \nResults or Findings: Bias between UHR-VMI and NR-VMI was found to be <5 \nHU for all materials. Noise was greater in UHR-VMI (8.2-22.6%) than NR-VMI. \nC(GGN) and C(ES) improved in UHR-VMI compared to no n-spectral UHR \nimages at <60 keV (max: 4.0 HU) and >60 keV (max: 3 .5 HU), respectively. \nThe 70 keV VMI was optimal for both CNR(GGN) and CN R(ES) due to minimal \nnoise amplification. For airways, errors (mean±σ) (mm) in 70 keV UHR-VMI \nwere lower: (UHR/NR) 0.06±0.09/0.15±0.19 (WT) and -0.18±0.17/-0.29±0.31 \n(ID), with highest improvements for smaller airways . \nConclusion: Spectral-UHR imaging with CZT-based PCD-CT offers \ndiagnostically-relevant contrast improvements betwe en diseased and normal \nlung over non-spectral images, with reduced measure ment errors for airways. \nLimitations: Study used a phantom approximating compositional si milarity to \nnormal and diseased lung instead of real patients. \nFunding for this study: N/A \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nZhou Yu: Employee: Canon Medical Research USA, Inc.  \nSteven Ross: Employee: Canon Medical Research USA, Inc. \nRichard Thompson: Employee: Canon Healthcare USA, I nc. \nAmir Pourmorteza: Nothing to disclose \nTom Labno: Employee: Canon Medical Research USA, In c. \nRuoqiao Zhang: Employee: Canon Medical Research USA , Inc. \nXiaohui Zhan: Employee: Canon Medical Research USA,  Inc. \nShobhit Sharma: Employee: Canon Medical Research US A, Inc. \n \n \nUltra low dose Photon Counting CT versus low dose p hoton counting CT \nin patients with cystic fibrosis \n*L. G. Murkes*, M. Lidegran, M. Sund, P. Hillergren , S. Diaz; Stockholm/SE \n(lena.gordon-murkes@regionstockholm.se) \n \nPurpose or Learning Objective: To introduce ultra low dose photon counting \nCT (ULDPCCT) as the main diagnostic follow up metho d in patients with cystic \nfibrosis (CF) and thereby reduce radiation dose wit h preserved or improved \ndiagnostic value. \nMethods or Background: Patients with CF undergo lifelong yearly follow up \nalternating CT and chest radiography from an early age. CT is the method of \nchoice and provides important information about the  course of the disease, \nmaking low and ultralow dose methods imperative. Th is prospective study \nincluded 71 CF patients between 7 and 66 years of a ge. A specific study \nprotocol was set up on a photon counting detector C T, Siemens Naeotom \nAlpha. All patients included underwent an inhaled a nd exhaled ULDPCCT and \nlow dose PCCT (LDPCCT) examinations at their yearly  follow up. Radiation \ndoses were collected for each scan and patient. The  median was also \ncalculated. Images from all scans were assessed sep arately by two paediatric \nradiologists with different years of experience usi ng a modified Bhalla scoring \nsystem. Interobserver agreement was calculated with  Cohens’ kappa \ncoefficient. P-values of <0.05 were considered stat istically significant. \nResults or Findings: The effective dose median (IQR) was 0,11 mSv (0.1-\n0.13) for the ULDPCCT and 0,77 mSv (0.66-0.87) for the LDPCCT \nrespectively. There was no statistically significan t difference between the \nBhalla scoring when comparing ULDPCCT versus LDPCCT  scans with p-value \n0,71 (0,37;1,04). Interobserver agreement was subst antial (Kappa value 0.65 \nfor ULDPCCT and 0.71 for LDPCCT) \nConclusion: A tailored ULDPCCT scan protocol might be used as y early \nfollow up diagnostic tool in CF patients with detai led diagnostic value and \nreduction of the radiation dose at approximately 1/ 7th of a regular LDCT, \nthereby reducing the accumulative dose contribution  to the patient. \nLimitations: Relatively limited amount of patients. \nFunding for this study: Funding was provided by \" Riksförbundet Cystisk \nFibros \" for the statistical analysis \nEthics committee - additional information: The study was approved by the \nswedish ethics commitee. According to the declarati on of Helsinkii. Dnr 2023-\n01227-01 \nAuthor Disclosures:  \nMarika Lidegran: Nothing to disclose \nPierre Hillergren: Nothing to disclose \nLena Gordon Murkes: Nothing to disclose \nSandra Diaz: Nothing to disclose \nMarie Sund: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 218  \nAI-based body composition analysis of COPD patients ’ CT scans –  \na multicentric study \n*B. K. Budai*¹, S. Hettinger¹, V. M. Wagner¹, V. Pa lm¹, R. Hosch², F. Nensa², \nO. Von Stackelberg¹, H-U. Kauczor¹, J. Biederer¹; ¹ Heidelberg/DE, ²Essen/DE \n \nPurpose or Learning Objective: This study focused on AI-based CT body \ncomposition analysis (BCA) as an alternative to bio electrical impedance \nanalysis (BIA) for identifying COPD patients at hig h risk of sarcopenia. We \naimed to construct CT-based linear regression model s for predicting the \npatients’ fat mass (FM), fat-free mass (FFM), skele tal muscle mass (SMM), \nand total muscle mass (MM). \nMethods or Background: A total of 571 COPD patients (349 males (61.1%), \naged 65.5 ± 8.6y) from a prospective multicentric study (COSYCONET) \nunderwent baseline chest CT scans and BIA. The AI-b ased BCA of inspiratory \nchest CTs was performed by the “Body and Organ Anal ysis” (BOA) algorithm. \nVolumes of muscles, bones, and fatty tissues were c onverted to mass in kg \nusing standard human tissue densities. Linear regre ssion with estimated thorax \nweight fitted to patient weight was used to extract  residuals which combined \nwith the respective CT-based measures, age, sex, we ight, and height were \nused for predicting BIA-based results. The reliabil ity of CT-predicted body \ncomposition measures was evaluated with intraclass correlation coefficients \n(ICC). The performance of the CT-based FFMI in iden tifying high-risk \nsarcopenia patients was assessed with ROC curve ana lysis. \nResults or Findings: The CT-based estimated body composition measures \ncorrelated well with BIA with ICCs of 0.90 (FM), 0. 94 (FFM), 0.92 (SMM), and \n0.92 (MM). The CT-based FFMI achieved an ICC=0.88 a nd predicted high-risk \nsarcopenia patients with an AUC, accuracy, sensitiv ity, and specificity of 0.903, \n88.3%, 88.7%, and 88.2%, respectively. Gwet’s AC1 o f 0.82 suggested \nexcellent agreement between the two approaches. \nConclusion: AI-based body composition analysis of chest CT scan s could be \nused to assess BIA-based body composition measures of COPD patients and \nto identify patients at high risk of sarcopenia. \nLimitations: The limitation of the study is that no external val idation was \nperformed. \nFunding for this study: Funding was provided by the German Federal \nMinistry of Education and Research (BMBF) (Projektt räger: DLR e.V. Bonn, \nFunding ref. 01GI0884) \nEthics committee - additional information: This study was approved by the \nlocal ethics committee and the central ethics commi ttee of the multicenter \nstudy. \nAuthor Disclosures:  \nJürgen Biederer: Nothing to disclose \nSophia Hettinger: Nothing to disclose \nHans-Ulrich Kauczor: Nothing to disclose \nOyunbileg Von Stackelberg: Nothing to disclose \nVerena Maria Wagner: Nothing to disclose \nViktoria Palm: Nothing to disclose \nRené Hosch: Nothing to disclose \nBettina Katalin Budai: Nothing to disclose \nFelix Nensa: Nothing to disclose \n \n \nArtificial intelligent based automated detection of  chest x-ray \nabnormalities as a support for young radiologists \n*L. Giuliani*, G. M. Masci, N. Landini, P. Giuliani , V. Panebianco, C. Catalano; \nRome/IT \n(lucagiuliani92@virgilio.it) \n \nPurpose or Learning Objective: To investigate whether AI represents an \nadded value for chest X-ray (CXR) interpretation. \nMethods or Background: A dataset of CXR performed between March 2023 \nand January 2024 were retrospectively selected from  the institutional PACS by \na senior thoracic radiologist. All CXR were evaluat ed by two young radiologists \nwith 1 year of experience in chest imaging, who ass essed the presence of \nseveral findings (consolidation, nodule, atelectasi s, fibrosis, calcification, \npneumothorax, cardiomegaly, pleural effusion, media stinal enlargement, \npneumoperitoneum). All examinations were then analy zed with an AI tool \n(Lunit INSIGHT CXR, Version 3.110) which assessed t he same features. \nFinally, an additional AI-assisted evaluation was p erformed by the two \nradiologists. The ground truth was established by t he radiologist in charge of \nimage selection who classified the abnormalities as  either visible or not visible \non the radiograph. \nResults or Findings: A total of 548 CXR examinations were selected. From  \nthe 500 CXR analyzed, a total of 876 findings were reported either from the \nradiologists or from the AI tool. The two radiologi sts showed a \nsensitivity/specificity of 80.6%/93.2% and 87.2%/95 .3%, respectively. AI \nshowed a sensitivity/specificity of 96.9%/64.3%. Wi th AI assistance, the \nsensitivity of the two radiologists increased to 82 .9% (+2.3%) and 89.3% \n(+2.1%), while specificity decreased to 87.2% (-6%)  and 89.5% (-5.8%). The  \n \n \nabnormalities for which the radiologists showed hig her disagreement with AI \nwere fibrosis and calcification (p<0.0001), whilst the abnormalities for which the \nradiologists more often changed interpretation afte r AI evaluation were \nnodules, calcification, and pneumothorax (p<0.0001) . \nConclusion: AI does not show significant increase of the diagno stic \nperformance compared to standard radiological evalu ation of CXR. \nParticularly, AI slightly increases sensitivity but  at cost of a significant decrease \nin specificity. Therefore, standard radiological in terpretation still remains the \ngold standard for CXR. \nLimitations: It was conducted retrospectively \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Sapienza University, Rome, \nRif.7226, Prot.0473/2024 \nAuthor Disclosures:  \nValeria Panebianco: Nothing to disclose \nNicholas Landini: Nothing to disclose \nPaolo Giuliani: Nothing to disclose \nGiorgio Maria Masci: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLuca Giuliani: Nothing to disclose \n \n \nDark-field Chest Radiography for Pneumothorax Asses sment \n*F. T. Gassert*, H. Bast, T. Urban, M. Lochschmidt,  L. Kaster, T. Koehler,  \nM. R. Makowski, F. Pfeiffer, D. Pfeiffer; Munich/DE  \n \nPurpose or Learning Objective: Conventional imaging techniques have \nlimitations in early detection of pneumothorax, par ticularly for small \npneumothoraces. Therefore, the purpose of this stud y was to evaluate the \npotential of dark-field chest radiography in improv ing the detection and \nassessment of pneumothorax. \nMethods or Background: This study included 100 participants, comprising 36  \npatients with clinically diagnosed pneumothorax and  64 healthy controls. All \nparticipants underwent dark-field X-ray chest radio graphy using a prototype \nsystem that simultaneously acquires attenuation-bas ed and dark-field images. \nSensitivity, specificity, accuracy, reading time, a nd diagnostic confidence were \ncompared between attenuation-based radiographs and the combination of \nattenuation-based radiographs with dark-field image s (dark-field overlays). \nResults or Findings: Dark-field radiography increased sensitivity for \npneumothorax detection from 84.2% (attenuation-base d radiographs) to 87.4% \n(dark-field overlays; p = .26), while specificity r emained constant (97.3% vs. \n97.4%). The median reading time was significantly r educed from 30.8 seconds \nto 10.3 seconds (p < .001), and diagnostic confiden ce improved significantly \nacross all readers (p < .001). \nConclusion: Dark-field chest radiography enhances the detection  of \npneumothorax, significantly reducing reading time a nd increasing diagnostic \nconfidence without compromising specificity. \nLimitations: Inclusion criteria and the imaging modality they we re assessed on \nwere different for pneumothorax patients and contro ls. While for pneumothorax \npatients, a conventional radiograph showing a pneum othorax was sufficient, \ncontrols had to show a normal CT scan to be include d. \nFunding for this study: We acknowledge financial support through the \nEuropean Research Council (AdG 695045), the Center for Advanced Laser \nApplications (CALA), the Federal Ministry of Educat ion and Research (BMBF) \nand the Free State of Bavaria under the Excellence Strategy of the Federal \nGovernment and the Länder, the German Research Foun dation (GRK2274), \nas well as by the Technical University of Munich–In stitute for Advanced Study. \nThis work was carried out with the support of the K arlsruhe Nano Micro Facility \n(KNMF, KNMF, www.kit.edu/knmf), a Helmholtz Researc h Infrastructure at \nKarlsruhe Institute of Technology (KIT). \nEthics committee - additional information: Ethics committee of the \nTechnical University of Munich \nAuthor Disclosures:  \nHenriette Bast: Nothing to disclose \nFlorian Tilman Gassert: Nothing to disclose \nThomas Koehler: Employee: Philips \nMarcus R. Makowski: Nothing to disclose \nTheresa Urban: Nothing to disclose \nLennard Kaster: Nothing to disclose \nFranz Pfeiffer: Nothing to disclose \nMaximilian Lochschmidt: Nothing to disclose \nDaniela Pfeiffer: Nothing to disclose \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 219  \n12:30-13:30 Research Stage 1 \nResearch Presentation Session: Breast \nRPS 1902 \nIntroduction of artificial intelligence in \nbreast screening \n \nModerator \nS. Schiaffino; San Donato Milanese/IT  \n(schiaffino.simone@gmail.com) \nAuthor Disclosures:  \nSimone Schiaffino: Board Member: European Journal o f Radiology, Eurorad, \nEuropean Radiology Experimental; Speaker: GE Health care \n \n \nImpact on quality performance indicators after impl ementing AI in a \nbreast cancer screening program in Germany \nK. Hamm¹, *A. Rodriguez Ruiz*², T. Jordan¹, B. Vett er¹, C. Engel³, C. Entrup⁴, \nM. Engelke⁵; ¹Chemnitz/DE, ²Nijmegen/NL, ³Leipzig/DE, ⁴Koblenz/DE, \n⁵Hamburg/DE \n \nPurpose or Learning Objective: To evaluate breast cancer screening quality \nindicators after implementation of an AI system for  support reading \nmammograms. \nMethods or Background: Two prospective and consecutive collected cohorts \nof women attending breast cancer screening with mam mography in a region of \nGermany where identified, just before and after imp lementation of an AI \ndecision support system to aid radiologists reading  mammograms (Transpara \nversion 1.7, ScreenPoint Medical). Before AI implem entation, all mammograms \nwere double read without AI. Afterwards, mammograms  were double read \nusing AI as concurrent decision support. All mammog rams were acquired with \nsame devices (Siemens Mammomat Inspiration). A tota l of X radiologists \nassessed the exams in this screening program. Scree ning quality indicators \n(cancer detection rate, recall rate, false positive  rate, PPV2) were compared in \nthe cohorts of women before and after implementatio n of AI using multivariate \nlogistic models adjusted for age, breast density, a nd interval from previous \nexamination. \nResults or Findings: 59.676 women attending screening before AI \nimplementation (2020-2021) and 58.546 women after A I implementation (2022-\n2023) were included in the analysis. Average age wa s 60 years old in both \ncohorts. Average number of months between rounds wa s 838 days in the no-\nAI cohort and 815 days in the AI cohort. After impl ementing AI, cancer \ndetection rate increased (349 screen-detected cance rs, 6.0/1000 vs 286 \nscreen-detected cancers, 4.8/1000, p=0.01), recall rate remained stable (2.5% \nvs 2.6%, p=0.29), false positive rate was reduced ( 1.9% vs 2.1%, p=0.002), \nand PPV2 increased (69%, 349/509 vs 60%, 286/477, p =0.009). \nConclusion: Implementing AI to support radiologists reading mam mograms in \na breast cancer screening program in Germany is saf e and effective, improving \ncancer detection rates and reducing false positives . \nLimitations: This prospective study has a non-paired non-randomi zed design. \nFunding for this study: None. \nEthics committee - additional information: Approved by local ethics \ncommittee. \nAuthor Disclosures:  \nAlejandro Rodriguez Ruiz: Employee: ScreenPoint Med ical \nMartin Engelke: Nothing to disclose \nKlaus Hamm: Nothing to disclose \nTorsten Jordan: Nothing to disclose \nBert Vetter: Nothing to disclose \nChristoph Engel: Nothing to disclose \nChristian Entrup: Nothing to disclose \n \n \nBreast Cancer Characteristics after the introductio n of Artificial \nIntelligence-supported double-reading in a Mammogra phy Screening \nProgram: comparison of baseline and subsequent roun ds \n*C. M. Weiss*, E. Di Gaetano, E. Cattarin, R. Cerni ato, G. Soppelsa, I. Vinci; \nTreviso/IT \n(claudiamaria.weiss@aulss2.veneto.it) \n \nPurpose or Learning Objective: To analyse the prognostic factors of breast \ncancers (BCs) detected with artificial intelligence -supported double-reading \n(AI-DR) and attempt to determine the long-term impa ct of these changes, \nparticularly on possible overdiagnosis. \n \n \nMethods or Background: AI-DR was applied to all digital screening \nmammograms (DSM) from November 2021 to June 2024: 9 9320 in the AI-\nbaseline-screen (AIBS) and 21237 in the AI-subseque nt-screen (AISS). The \ncollected data were compared by retrospective analy sis to determine whether \nAIBS screen-detected BCs differed from AISS. We use d the Z-test to compare \nthe proportions of the data between AIBS and AISS. \nResults or Findings: With a total of 1093 screen-detected BCs (AIBS: \n944/99320; AISS: 149/21367), the study revealed a d ecrease (-26.6%) in the \ncancer detection rate (CDR) per 1000 in AISS compar ed to AIBS (6.97vs9.5). \nThe recall rate (RR) was lower (-42.3%) in AISS tha n in AIBS (1.8%vs3.1%). \nNo significant differences were found in the percen tage of invasive BCs \n(AIBS82%vsAISS81.9%) and in situ BCs (AIBS18%vsAISS 18.1%). Higher \npercentages of luminal BCs were observed in the AIB S than in AISS \n(91.1%vs82.6%), while in AISS, there were higher pe rcentages of high-grade \n(AIBS25.3%vs AISS33.6%), HER2positive (AIBS 8.3%vsA ISS12.4%) and \ntriple-negative BCs (AIBS2.6%vsAISS 5%). \nConclusion: The reduction of RR and CDR in AISS aligns with the  expectation \nof later screening focusing on disease onset or pro gression cases. In AISS, \ncompared to AIBS, more aggressive BCs were detected , while less aggressive \nBCs were reduced. This might suggest an improved pe rformance in the \nsecond round of screening, with a positive impact o n the reduction of \noverdiagnosis. \nLimitations: Data on interval and advanced cancers are lacking, which would \nallow for an analysis of long-term clinical outcome s. \nFunding for this study: No funding \nEthics committee - additional information: Not requested \nAuthor Disclosures:  \nGiorgia Soppelsa: Nothing to disclose \nEleonora Di Gaetano: Nothing to disclose \nElisa Cattarin: Nothing to disclose \nIvana Vinci: Nothing to disclose \nClaudia Maria Weiss: Nothing to disclose \nRoberta Cerniato: Nothing to disclose \n \n \nMulticenter Analysis of AI Assessed Mammography Tec hnologist \nPositioning Variability Between Breast Screening Pr ograms \n*G. Spear*¹, L. R. Margolies², J. Payne³, S. E. E. Iles³, J. Seely⁴, N. Sharma⁵, \nS. H. Heywang-Köbrunner⁶, T. W. W. Vomweg⁷, M. Abdolell³; ¹Chicago, IL/US, \n²New York, NY/US, ³Halifax, NS/CA, ⁴Ottawa, ON/CA, ⁵Leeds/UK, ⁶Munich/DE, \n⁷Koblenz/DE \n(ggspear21@gmail.com) \n \nPurpose or Learning Objective: Variability in mammographic positioning \nquality, both between breast screening programs (BS Ps) and across different \npositioning errors, presents a challenge to establi shing standardized \nmammography quality service delivery. Although high -quality mammography \nhelps ensure diagnostic accuracy, and training can enhance image quality, \nthere is a lack of supporting population-level empi rical data. This study aims to \nquantitatively assess mammography technologists’ po sitioning error rates \nacross BSPs. \nMethods or Background: The MAMMO.IQ study encompassed a total of \n249,817 screening mammograms acquired between Decem ber 1, 2019, and \nFebruary 28, 2021, from seven BSPs across North Ame rica and Europe. The \npositioning errors assessed included: exaggeration,  portion cut off, posterior \ntissues missing, nipple not in profile, too high on  IR, pectoralis shape/position, \nsagging, IMF missing/obscured, PNL difference, and compression. The \nCoefficient of Variation (CV) assessed variability in error rates, (1) between \nBSPs, and (2) between positioning errors. The withi n-BSS CV for each unmet \npositioning criterion was computed using rates for all technologists within a \nBSS. \nResults or Findings: Images acquired by 310 technologists were analyzed.  \nOver/under compression had the lowest variability ( CV=16.44%) indicating \nconsistent practices. Too High on IR exhibited the highest variability \n(CV=71.52%) reflecting a high level of inconsistenc ies. The MLO Inadequate \nPectoralis Length had a CV of 50.45%, representing the median level of \nvariability. \nConclusion: This study highlights variability in mammography te chnologists’ \npositioning errors between and within BSPs. While s ome positioning criteria \nshow consistent practices, others may benefit from improved standardization. \nUnderstanding inconsistencies in mammography servic e delivery helps identify \nopportunities to standardize positioning practices and reduce variability, \nleading to more equitable, high-quality care and fe wer positioning errors. \nLimitations: Missing data on technologist experience, staffing, and COVID-19 \nresponse measures limits the understanding of facto rs driving disparities. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Ethics approvals were obtained \nfrom participating BSPs (NSHA-REB#1026590). \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 220  \nAuthor Disclosures:  \nNisha Sharma: Advisory Board: Densitas \nLaurie R. Margolies: Nothing to disclose \nSylvia H. Heywang-Köbrunner: Nothing to disclose \nSian E. Elizabeth Iles: Nothing to disclose \nJean Seely: Nothing to disclose \nJennifer Payne: Other: Densitas \nToni Werner W Vomweg: Nothing to disclose \nGeorgia Spear: Nothing to disclose \nMohamed Abdolell: CEO: Densitas \n \n \nCharacteristics of Breast Cancers before and after the introduction of \nArtificial Intelligence-supported double reading in  a Mammography \nScreening Program \n*C. M. Weiss*, E. Di Gaetano, E. Cattarin, R. Cerni ato, G. Soppelsa, I. Vinci; \nTreviso/IT \n(claudiamaria.weiss@aulss2.veneto.it) \n \nPurpose or Learning Objective: Screening supported by artificial intelligence \n(AI) increased cancer detection compared to screeni ng without AI. However, it \nis still unclear whether the additional cancer dete ction improves outcomes or \nleads to overdiagnosis of breast cancers (BCs). \nMethods or Background: From January 2019 to October 2021, 134259 \nwomen underwent digital screening mammography (DSM)  with human-double-\nreading (HDR) and from November 2021 to June 2024, 131406 DMS with AI-\nsupported HDR (AI-HDR) DM. The collected data (canc er detection rate [CDR], \nrecall rate [RR] and tumour characteristics) were c ompared by retrospective \nanalysis to determine whether the BCs detected by s creening differed between \nHDR and AI-HDR. We used the Z-test to compare the p roportions of the data \nbetween HDR and AI-HDR. \nResults or Findings: With a total of 2044 screen-detected BCs \n(HDR:938/134259; AI-HDR:1106/131406), the study rev ealed a significant \nincrease (+20.5%) in CDR per 1000 with AI-HDR compa red to HDR \n(8.42vs6.99 per 1000, respectively). The RR was low er (-14.9%) with AI-HDR \nthan with HDR (2.6%vs3.1%). The AI-HDR showed the f ollowing differences in \nBCs rates compared to HDR: lower for invasive BCs ( 82.1%vs84.1%), and \nhigher for in situ BCs (17.9%vs15.9%); higher moder ate-grade BCs \n(65.5%vs61.5%), and lower high-grade BCs (26.4%vs 2 7.9%); higher luminal \nBCs (88.2%vs83.2%), lower HER2positive (8.7%vs13.2% ), and lower triple \nnegatives (2.9%vs3.6%). \nConclusion: It can be concluded that the use of AI-HDR produced  statistically \nsignificant differences in detecting various tumour  subtypes compared to HDR. \nIn particular, it seems to have increased the detec tion of less aggressive BCs \nand reduced unnecessary recalls. \nLimitations: Data on interval and advanced cancers are lacking, which would \nallow for an analysis of long-term clinical outcome s. \nFunding for this study: No fundings \nEthics committee - additional information: Not requested \nAuthor Disclosures:  \nGiorgia Soppelsa: Nothing to disclose \nEleonora Di Gaetano: Nothing to disclose \nElisa Cattarin: Nothing to disclose \nIvana Vinci: Nothing to disclose \nClaudia Maria Weiss: Nothing to disclose \nRoberta Cerniato: Nothing to disclose \n \n \nArtificial intelligence as an initial reader for do uble reading in breast \ncancer screening: A prospective initial study of 32 ,822 mammograms of \nthe Egyptian population \nS. A. Mansour, R. M. Kamal, M. M. Gomaa, E. Sweed, S. Hussien, E. Abdalla, \n*Y. M. Nada*, G. Mohamed, A. F. I. Moustafa; Cairo/ EG \n(dr.yasmin.nada.nl@gmail.com) \n \nPurpose or Learning Objective: Although artificial intelligence (AI) has \npotential in the field of screening of breast cance r, there are still issues. It is \nvital to make sure AI doesn't overlook cancer or ca use needless recalls. The \naim of this work was to investigate the effectivene ss of indulging AI in \ncombination with one radiologist in the routine dou ble reading of \nmammography for breast cancer screening. \nMethods or Background: The study prospectively analyzed 32822 screening \nmammograms. Reading was performed in a blind-paired  style by i) two \nradiologists and ii) one radiologist paired with AI . A heatmap and abnormality \nscoring percentage were provided by AI for abnormal ities detected on \nmammograms. Negative mammograms and benign-looking lesions that were \nnot biopsied were confirmed by a 2-year follow-up. \n \n \n \n \nResults or Findings: Double reading by the radiologist and AI detected 1 324 \ncancers (6.4%); on the other side, reading by two r adiologists revealed 1293 \ncancers (6.2%) and presented a relative proportion of 1·02 (p<0·0001). At the \nrecall stage, suspicion and biopsy recommendation w ere more presented by \nthe AI plus one radiologist combination than by the  two radiologists. The \ninterpretation of the mammogram by AI plus only one  radiologist showed a \nsensitivity of 94.03%, a specificity of 99.75%, a p ositive predictive value of \n96.571%, a negative predictive value of 99.567%, an d an accuracy of 99.369% \n(from 99.252% to 99.472%). The positive likelihood ratio was 387.260, \nnegative likelihood ratio was 0.060, and AUC “area under the curve” was 0.969 \n(0.967 to 0.971). \nConclusion: AI could be used as an initial reader for the evalu ation of \nscreening mammography in routine workflow. Implemen tation of AI enhanced \nthe opportunity to reduce false negative cases and supported the decision to \nrecall or biopsy. \nLimitations: The study is a single institute work. \nFunding for this study: No source of funding. \nEthics committee - additional information: The study has been approved by \nthe Baheya Charity Hospital research center. \nAuthor Disclosures:  \nSahar Abdelkhalek Mansour: Nothing to disclose \nSamar Hussien: Nothing to disclose \nMohammed Mohamed Gomaa: Nothing to disclose \nEngy Abdalla: Nothing to disclose \nEnas Sweed: Nothing to disclose \nYasmin Mohamed Nada: Nothing to disclose \nGhada Mohamed: Nothing to disclose \nRasha Mohamed Kamal: Nothing to disclose \nAmr Farouk Ibrahim Moustafa: Nothing to disclose \n \n \nSimulating single reading for high-risk examination s in the randomized \ncontrolled Mammography Screening with Artificial In telligence trial \n(MASAI) \n*V. Josefsson*, D. Schmidt, H. Sartor, O. Hagberg, K. Lang; Malmö/SE \n(viktoria.josefsson@med.lu.se) \n \nPurpose or Learning Objective: To assess the value of double reading of \nhigh-risk examinations in the MASAI trial. \nMethods or Background: In the randomised controlled MASAI trial AI \nsupported screening was compared to standard double  reading. AI was used \nto triage exams to single or double reading dependi ng on malignancy risk and \nas detection support. Of the 53 048 participants in  the intervention arm, 3800 \nexams were high risk and underwent double reading w hile the remaining \nexams underwent single reading. In this retrospecti ve study, we assessed the \nrelative performance in the intervention arm, compa ring simulated single and \nfactual double reading of high-risk exams and its e ffect on cancer detection, \nrecalls, and false positives. Cancers solely detect ed by the second reader were \ndescribed. \nResults or Findings: The simulated single reading scenario resulted in 8 .9% \n(308 vs. 338) fewer detected cancers and 5.9% fewer  recalls (1045 vs. 1110) \ncompared to the factual outcome in the intervention  arm. Corresponding \nsimulated vs. factual rates were 5.8/1000 vs. 6.4/1 000 for cancer detection, \n2.0% vs. 2.1% for recalls and 1.4% vs. 1.5% for fal se positive. Of the 30 \ncancers solely detected by the second reader, 24 (8 0.0%) were invasive and \n21 (70.0%) were classified as T1. Of the invasive c ancer, 23 (95.8%) were \nlymph-node negative and 8 (33.3%) non-luminal A, of  which four were triple \nnegative. \nConclusion: Double reading of high-risk exams improved cancer d etection \nwithout unduly increasing false positives. The addi tional cancers detected were \nmostly small, lymph-node negative invasive cancers,  including those of \nsignificant prognostic subtypes. These findings sup port the continued use of \ndouble reading for high-risk exams in AI-supported screening. \nLimitations: Single-institution trial. Retrospective simulation.  \nFunding for this study: The Swedish Cancer Society \nRegional Cancer Centers in Collaboration \nLund University ALF-funds \nEthics committee - additional information: The Swedish Ethical Review \nAuthority  \n2020-04936  \n2023-026848-02 \nAuthor Disclosures:  \nKristina Lang: Nothing to disclose \nViktoria Josefsson: Nothing to disclose \nHanna Sartor: Nothing to disclose \nOskar Hagberg: Nothing to disclose \nDavid Schmidt: Nothing to disclose \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 221  \nUsing prior mammograms to improve specificity of an  AI system for \nbreast cancer detection: a large-scale retrospectiv e multi-site validation \n*A. Rodriguez Ruiz*¹, S. Pires¹, R. Peeters¹, G. Ro driguez-Esteban¹,  \nD. Sperber¹, C. De Wolf², J. L. Raya Povedano³, S. Romero Martin³, R. Mann¹; \n¹Nijmegen/NL, ²Geneva/CH, ³Cordoba/ES \n \nPurpose or Learning Objective: To investigate how the use of prior \nmammograms impacts breast cancer detection performa nce of an AI system. \nMethods or Background: Mammograms from women attending three \nEuropean screening programs were collected based on  availability of prior \nimages and at least 2 years follow-up, including or iginal radiologists \nassessments. Each case was analyzed by a breast can cer detection AI \nproduct (Transpara, ScreenPoint Medical, v2.1), res ulting in two cancer risk \nscores: using as input the current mammogram alone and using prior \nmammograms. AI specificity was compared between usi ng priors or not, \nmatching the single radiologist sensitivity . Subse quently, the combination of a \nsingle radiologist and AI was modelled and compared  to double human \nreading. P-values using McNemar and binomial confid ence intervals were \ncomputed. \nResults or Findings: 37,148 cases were included (20,300 from Switzerland , \n916 from Spain 15,932 from The Netherlands), with 1 ,034 recalled cases \n(2.8%), 247 screen-detected cancers (6.6/1000), and  59 interval cancers \n(1.6/1000). 56% of cases had 1 prior mammogram, 44%  had 2 or 3. Images \nwere acquired with Hologic, Siemens, GE, Planmed an d Philips machines. At \nthe average sensitivity of a single radiologist (71 .2%), AI specificity increased \nwhen using prior images, from 98.1% (98.0-98.2%) to  98.8% (98.7-98.9%), \nrepresenting a 37% reduction in false positives (fr om 1.9% to 1.2%, P<0.001). \nCombining AI using priors with a radiologist achiev ed comparable sensitivity \n(83.0% vs 82.0%, P=0.66) and higher specificity (96 .3%, 96.0-96.5%) than \ndouble human reading before consensus (95.4%, 95.2- 95.6%), representing \n20% fewer false positives (P<0.001). The improved s pecificity was higher for \ncases with breast density C/D (+0.8%, P<0.001) than  for A/B (+0.3%, \nP=0.005). \nConclusion: A higher specificity was achieved by a breast cance r detection AI \nsystem using prior mammograms, potentially offering  better aid to radiologists \nin breast cancer screening. \nLimitations: Retrospective design. \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nSantiago Pires: Employee: ScreenPoint Medical \nDaan Sperber: Employee: ScreenPoint Medical \nAlejandro Rodriguez Ruiz: Employee: ScreenPoint Med ical \nRuud Peeters: Employee: ScreenPoint Medical \nSara Romero Martin: Nothing to disclose \nJose Luis Raya Povedano: Nothing to disclose \nGonzalo Rodriguez-Esteban: Employee: ScreenPoint Me dical \nChris De Wolf: Nothing to disclose \nRitse Mann: Nothing to disclose \n \n \n12:30-13:30 Research Stage 2 \nResearch Presentation Session: Vascular \nRPS 1915 \nCarotid and intracranial artery imaging \n \nModerator \nV. Silvestri; Seclin/FR  \n(valentina.silvestri@ghsc.fr) \n \n \nSub-1-minute Relaxation-Enhanced Angiography withou t Contrast and \nTriggering of the Extracranial Arteries \n*J. P. Janssen*¹, K. Kaya¹, R. A. Terzis¹, J. Trist ram¹, R. J. Gertz¹, L. Goertz¹, \nL. Pennig¹, C. H. Gietzen¹, K. Weiss²; ¹Cologne/DE,  ²Hamburg/DE \n(jan.janssen@uk-koeln.de) \n \nPurpose or Learning Objective: To evaluate the acceleration of a 3D \nisotropic flow-independent non-contrast MRA (REACT)  of the neck using \nCompressed SENSE (CS) combined with deep learning-b ased reconstruction \n(CS-AI). \nMethods or Background: Thirty-four volunteers received cervical REACT at \n3T ((acquired threefold: (1) CS factor 7 (scan time : 1:20 min), (2) CS factor 10 \n(0:55 min), and (3) CS-AI factor 10 (0:55 min)). Tw o radiologists rated the \nimage quality of seven arterial segments and overal l image noise. Additionally, \na pairwise forced-choice comparison was conducted. Apparent signal- (aSNR) \nand contrast-to-noise ratios (aCNR) were measured, and image sharpness \nwas assessed by calculating the edge rise distance (ERD). Multiple t-tests and \nnon-parametric tests with Bonferroni correction wer e performed for comparison \nto CS7, which was considered as the current clinica l standard. \nResults or Findings: Compared to CS7, CS10 showed lower image quality \nscores (p<0.001) while CS10-AI obtained higher resu lts (p=0.010). Image \nnoise was similar between CS7 and CS10 (p=0.138) wh ile CS10-AI yielded a \nlower noise (p=0.008). Forced choice revealed prefe rences for CS7 over CS10 \n(p<0.001), but no preference between CS7 and CS10-A I (p>0.999). Compared \nto CS7, aSNR and aCNR were lower in CS10 (p<0.001) and the ERD was \nlonger (p=0.004), while CS10-AI provided better aSN R and aCNR (p=0.001) \nand showed no difference in ERD (p=0.776). \nConclusion: CS-AI enables the acquisition of cervical REACT in less than one \nminute without compromising image quality. Further studies are required to \nconfirm these results in patients and to evaluate t he diagnostic performance \nregarding vascular findings such as stenosis or dis section. \nLimitations: No pathologies were assessed. No comparison was mad e with \nestablished reference standards. \nFunding for this study: Not applicable. \nEthics committee - additional information: Approved by our institutional \nreview board (reference number: 20-1296_1) and regi stered in the national \nregistry for clinical trials (DRKS00030210). \nAuthor Disclosures:  \nLenhard Pennig: Speaker: Guerbet GmbH Speaker: Phil ips Healthcare \nCarsten H. Gietzen: Nothing to disclose \nJuliana Tristram: Nothing to disclose \nRoman Johannes Gertz: Speaker: Guerbet GmbH Speaker : Philips Healthcare \nKenan Kaya: Nothing to disclose \nKilian Weiss: Employee: Philips GmbH \nRobert Angelo Terzis: Nothing to disclose \nJan Paul Janssen: Nothing to disclose \nLukas Goertz: Nothing to disclose \n \n \nCarotid artery assessment in dual-source photon-cou nting CT: impact of \nlow-energy virtual monoenergetic imaging on image q uality, vascular \ncontrast and diagnostic assessability \n*A-I. Nica*¹, C. Booz¹, G. M. Bucolo¹, L. S. Alizad eh¹, T. Vogl¹, T. D'Angelo²,  \nH-L. Kaatsch³, D. Overhoff³, S. Waldeck³; ¹Frankfur t/DE, ²Messina/IT, \n³Koblenz/DE \n \nPurpose or Learning Objective: The purpose of this study is to evaluate the \nimpact of low-energy VMI reconstructions on quantit ative and qualitative image \nquality, vascular contrast, and diagnostic assessab ility of the carotid arteries in \nphoton-counting CTA. \nMethods or Background: A total of 122 patients (67 male) who had \nundergone dual-source photon-counting CTA scans of the carotid artery were \nretrospectively analyzed in this study. Standard 12 0 kV CT images and low-\nkeV VMI series from 40 to 100 keV with an interval of 15 keV were \nreconstructed. Quantitative analyses included the e valuation of vascular CT \nnumbers, signal-to-noise ratio (SNR), and contrast- to-noise ratio (CNR). CT \nnumber measurements were performed in the common, e xternal, and internal \ncarotid arteries. Qualitative analyses were perform ed by three board-certified \nradiologists independently using five-point scales to evaluate image quality, \nvascular contrast, and diagnostic assessability of the carotid arteries. \nResults or Findings: Mean attenuation, CNR and SNR values were highest i n \n40 keV VMI reconstructions (HU, 1362.32 ± 457.81; CNR, 33.19 ± 12.86; SNR, \n34.37 ± 12.89) followed by 55-keV VMI reconstructions; all three mean values \nat these keV levels were significantly higher compa red with standard 120 kV \nCT series (HU, 154.43 ± 23.69; CNR, 16.34 ± 5.47; SNR, 24.44 ± 7.14) \n(p < 0.0001). The qualitative analysis showed highest ra ting scores for 55 keV \nVMI reconstructions followed by 40 keV and 70 keV V MI series with a \nsignificant difference compared to standard 120 kV CT images regarding \nimage quality, vascular contrast, and diagnostic as sessability of the carotid \narteries (all comparisons, p < 0.01). \nConclusion: Low-keV VMI reconstructions at a level of 40–55 keV  significantly \nimprove image quality, vascular contrast, and the d iagnostic assessability of \nthe carotid arteries compared with standard CT seri es in photon-counting CTA. \nLimitations: Single-center retrospective study \nFunding for this study: No funding was received \nEthics committee - additional information: The local IRB approved this \nstudy. \nAuthor Disclosures:  \nChristian Booz: Speaker: Siemens Healthineers \nThomas Vogl: Nothing to disclose \nDaniel Overhoff: Nothing to disclose \nAndreea-Ioana Nica: Nothing to disclose \nHanns-Leonhard Kaatsch: Nothing to disclose \nStephan Waldeck: Nothing to disclose \nTommaso D'Angelo: Speaker: Philips Speaker: Bracco \nLeona Soraja Alizadeh: Nothing to disclose \nGiuseppe Mauro Bucolo: Nothing to disclose \n\n \n \nSaturday \nAbstract-based Programme \n \n 222  \nA Novel Approach in Vascular Imaging: AI-Driven 3D Reconstruction of \nCarotid Arteries for Enhanced Stroke Risk Assessmen t \nK. Gasbarrino¹, *A. Benjamin*¹, T. Beiko¹, J. Ramir ez-Garcia Luna¹, R. Khan¹, \nL. H. Gonzalez Torres¹, S. Levasseur¹, S. Taj², K. Khan¹; ¹Montreal/CA, \n²Columbia, MD/US \n \nPurpose or Learning Objective: The standard approach to assessing stroke \nrisk via 2D carotid ultrasound is limited by operat or variability, a lack of 3D \nvessel visualization, and subjective interpretation , resulting in a nearly 30% \nmisclassification rate. To address these challenges , we developed AI-powered \nsoftware that transforms 2D ultrasound images into precise 3D models of \ncarotid arteries and automates vessel measurements.  \nMethods or Background: We applied a multi-class U-Net AI model, trained on  \n~4000 2D ultrasound images from 113 North American patients with \ncardiovascular risk factors. Two independent sonogr aphers annotated these \nimages, identifying key vascular structures, includ ing medial-adventitial \nboundary, intimal-luminal boundary, and plaque. 3D reconstructions were \nachieved by integrating 2D image segmentations with  positional data captured \nfrom an electromagnetic sensor (Northern Digital In c, Canada) during a single \nB-mode sweep of the carotid artery. Algorithms were  developed for automated \nmeasurement of vessel diameter, artery stenosis, an d classification of disease \nseverity. Validation was conducted using a carotid artery phantom with a \npredefined 70% stenosis (R.G. Shelley Ltd, Canada),  along with clinical \nevaluation in 8 patients to compare performance aga inst the current standard \nof care. \nResults or Findings: The AI model demonstrated strong performance, \nachieving a DICE coefficient of 0.86 in detecting v essel structures. The \nsoftware successfully generated 3D models, with vas cular metrics showing a \n99% agreement with the known stenosis in the phanto m model. Intra-operator \nvariability was minimal, with stenosis measurements  showing only minor \ndeviations (71.42±3.42%). In the clinical study, a 90% reduction in ultrasound \nscan was achieved, while maintaining diagnostic acc uracy equivalent to that of \na vascular radiologist with >10 years of experience . \nConclusion: Our software represents a significant advancement i n carotid \nartery imaging, delivering a ten-fold improvement i n scan efficiency while \nachieving expert-level diagnostic accuracy with min imal variability. \nLimitations: N/A \nFunding for this study: Ontario Brain Institute; Québec's Ministère de \nl'Économie, de l'Innovation et de l'Énergie \nEthics committee - additional information: The study was approved by \nAdvarra IRB (Pro00068778) \nAuthor Disclosures:  \nLuis H. Gonzalez Torres: Nothing to disclose \nRafia Khan: Consultant: Sonaro Inc \nJose Ramirez-Garcia Luna: Shareholder: Sonaro Inc \nSophie Levasseur: Nothing to disclose \nKarina Gasbarrino: Founder: Sonaro Inc \nThierry Beiko: Employee: Sonaro Inc. \nKashif Khan: Founder: Sonaro Inc \nSabir Taj: Advisory Board: Sonaro Inc \nAlex Benjamin: Research/Grant Support: Sonaro \n \n \nAssociation Between Pericarotid Fat Density and Hem orrhagic \nTransformation After Endovascular Therapy for Acute  Extracranial \nInternal Carotid Artery Occlusion \n*Z. Cui*, J. Zhang; Shanghai/CN \n(510349687@qq.com) \n \nPurpose or Learning Objective: This study aimed to investigate the \nassociation between pericarotid fat density (PFD) a round the occlusion and \nhemorrhagic transformation (HT) risk and functional  outcome in acute ischemic \nstroke (AIS) patients with extracranial internal ca rotid artery (e-ICA) occlusion \nwho underwent endovascular thrombectomy (EVT). \nMethods or Background: This multicenter retrospective study included a \ncohort of patients with e-ICA occlusion after EVT b etween June 2019 and \nMarch 2024. PFD was assessed using semi-automated q uantitative software \nat pre-operation neck CT angiography (CTA). The ass ociations between PFD \nand HT, and functional outcome (mRS score 0-3 vs 4- 6) were analyzed using \nmultivariable logistic regression. A mediation anal ysis was conducted to \nexplore whether HT mediates the relationship betwee n PFD and functional \noutcome. Additionally, We sought the association be tween PFD and admission \nC-reactive protein (CRP) levels. \nResults or Findings: 101 patients were included and divided into an HT g roup \n(n=36) and a non-HT group (n=65). PFD was independe ntly associated with \nHT (adjusted odds ratio [aOR]: 1.84, 95% CI: 1.28 t o 2.66, P<0.001) and \nunfavorable functional outcome (1.41, 1.04 to 1.91,  p=0.030). The AUC values \nwere 0.79 (95% CI, 0.70 to 0.89) and 0.68 (0.57 to 0.78), indicating a favorable \npredictive performance for the HT risk and unfavora ble prognosis prediction. \nMediation analysis revealed that HT explained more than 60% of the \nrelationship between PFD and worse functional outco me. In addition, higher \nPFD was positively correlated with elevated CRP lev els. \nConclusion: PFD is an independent predictor of HT and a worse f unctional \noutcome at 90 days in patients with AIS and e-ICA o cclusion who underwent \nEVT. Evaluation of PFD provides opportunities for H T risk stratification and \noutcome prediction. \nLimitations: Given the sample size, we were unable to apply the more \nstringent inclusion criteria of restricted to the e xtracranial ICA. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Informed consent was waived for \nall participants due to the retrospective nature of  the study. \nAuthor Disclosures:  \nJun Zhang: Nothing to disclose \nZhimeng Cui: Nothing to disclose \n \n \nThe influence of visual signals on blood flow in th e central retinal and \ninternal carotid arteries \nM. Beraia, D. Gachechiladze, *A. Siradze*, N. Eliav a, M. Lazarashvili,  \nN. Nikabadze, S. Siradze, J. Giorgelashvili, L. Udu mashvili; Tbilisi/GE \n(anetasiradze@yahoo.com) \n \nPurpose or Learning Objective: The retina offers a unique window into brain \nstructure and functional disorders due to its anato mical, physiological, and \nembryological similarities with the brain. This res earch explores the influence of \nvisual-verbal/nonverbal stimuli on blood flow in bo th the retina and brain. \nMethods or Background: A duplex Ultrasound study was conducted with 25 \nvolunteers (11 males, 14 females, aged 21–35), exam ining blood flow in the \ncentral retinal artery (CRA) and internal carotid a rtery (ICA) under two types of \nvisual stimuli: verbal irritation (Shakespeare’s so nnets) and nonverbal (pictures \n– find the hidden figures). Blood flow parameters ( Vsys, Vdia, PI, and RI) were \nmeasured in intervals 1-15 and 25-40 seconds after stimuli initiation, to assess \nthe nature (neuro/humoral) of blood flow regulation  (RBC circulation time: 20 \nseconds). Initial 5sec for the baseline images. CE- MRI angiography (TOF) was \nused to exclude vascular anomalies. \nResults or Findings: In the CRA, Vsys increased from 7.5–11.5 cm/sec to \n10.7–14.3 cm/sec, with Vdia at 4.1–4.9 cm/sec. RI r ose from 0.53 to 0.64, and \nPI from 0.71 to 0.94. In the ICA, Vsys rose from 80 –130 cm/sec to 95–170 \ncm/sec, with Vdia at 24–45 cm/sec. RI increased fro m 0.67 to 0.74, and PI \nfrom 1.01 to 1.17. Changes in the CRA and ICA were unidirectional (r = 0.7). In \nnonverbal cases, Vsys and RI were higher in the CRA  (P < 0.05) and ICA (P < \n0.01). The time of the stimuli initiation did not c hange the results. These details \nindicate the mostly sympathetic regulation of cereb ral blood flow. \nConclusion: Quantitative blood circulation studies in these art eries may be \nused as the functional ultrasound diagnostics of th e retina and brain. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number UID 2438. \nAuthor Disclosures:  \nNino Eliava: Nothing to disclose  \nLuarsabi Udumashvili: Nothing to disclose \nAneta Siradze: Nothing to disclose \nDudana Gachechiladze: Nothing to disclose \nMariam Lazarashvili: Nothing to disclose \nSalome Siradze: Nothing to disclose \nMerab Beraia: Nothing to disclose \nJano Giorgelashvili: Nothing to disclose \nNini Nikabadze: Nothing to disclose \n \n \nIntracranial arterial calcification detection; a co mparison between ultra-\nhigh-resolution photon-counting CT, conventional en ergy-integrating CT \nand micro-CT \n*J. Van Der Bie*¹, B. P. Berghout¹, R. P. J. Budde¹ , J. Gutierrez²,  \nM. Van Straten¹, D. Bos¹; ¹Rotterdam/NL, ²New york,  NY/US \n(j.vanderbie@erasmusmc.nl) \n \nPurpose or Learning Objective: To assess the performance of photon-\ncounting detector CT (PCD-CT) in detecting and quan tifying intracranial arterial \ncalcifications and comparing the performance to con ventional energy-\nintegrating CT (EID-CT), using micro-CT (µCT) as th e reference standard. \nMethods or Background: Thirty histopathological cross-sections of \nintracranial arteries were scanned with PCD-CT, EID -CT, and µCT. µCT was \noptimized for image quality (reference standard), w hile clinical protocols were \nused for PCD-CT and EID-CT. Various reconstruction kernels (EID-CT: \nHv40/Hv49/Hv59; PCD-CT: Hv40/Hv48/Hv56/Hv64/Hv72/Hv 89) were used to \nenhance spatial resolution. Two experienced observe rs independently \nevaluated the presence of calcifications in all acq uisitions and were compared \nby Cohen’s Kappa and concordance percentages. For o bjective analysis, mass \nscores were used to assess both the detection and m ass. The objective \nmeasure was analyzed using Bland-Altman plots. \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 223  \nResults or Findings: Observer 1 detected calcifications in 24 samples an d \nObserver 2 in 23 samples using µCT (90% concordance , κ=0.706). EID-CT \nwith Hv59 showed the highest interobserver agreemen t (97% concordance, \nκ=0.911), but low detection rates (observer 1: 27%, observer 2: 25%) \ncompared to µCT. PCD-CT yielded better detection wi th Hv48 (observer 1: \n70% detection rate, 90% concordance, κ=0.706) and Hv56 (observer 2: 80% \ndetection rate, 77% concordance, κ=0.314). Mass scores indicated the highest \ndetection with PCD-CT Hv64, though with increased n oise. The Hv48/Hv56 \nkernels were deemed optimal, yielding sensitivity, and specificity (for observers \n1/2) of 83%/92%, 50%/83%, respectively, and accurac y of 77%/90%. \nConclusion: PCD-CT outperformed EID-CT in detecting intracrania l \ncalcifications. Nevertheless, small calcifications sometimes go undetected \ncompared to µCT. Hv48 and Hv56 kernels are recommen ded for optimal \nresults, balancing detection rates and noise. \nLimitations: In the observer study, detection was assessed on a sample basis, \nleading to potential oversight of smaller calcifica tions in segments where larger \nones were present, which is reflected in the mass s cores. \nFunding for this study: This study has received funding by Smart*Light is \npartially funded by the Interreg V Flanders-Netherl ands program with financial \nsupport from the European Regional Development Fund  (ERDF). \nEthics committee - additional information: All brain sources were approved \nby the IRB at their respective institutions. The In stitutional Review Board \nwaived written informed consent: \nAuthor Disclosures:  \nRicardo P. J. Budde: Speaker: Siemens Healthineers Advisory Board: Bayer \nMarcel Van Straten: Nothing to disclose \nDaniel Bos: Nothing to disclose \nBernhard P. Berghout: Nothing to disclose \nJose Gutierrez: Nothing to disclose \nJudith Van Der Bie: Nothing to disclose \n \n \nCerebral blood flow alterations and host genetic as sociation in \nindividuals with long COVID: A transcriptomic-neuro imaging study \n*Y. Wang*¹, F. Zhou²; ¹Nanchang/CN, ²NanChang City,  Jiangxi Province/CN \n(1694865434@qq.com) \n \nPurpose or Learning Objective: Neuroimaging studies have indicated that \naltered cerebral blood flow (CBF) was associated wi th the long-term symptoms \nof long COVID. long COVID were found to be strongly  associated with host \ngene expression. Nevertheless, the relationships be tween altered CBF, clinical \nsymptoms, and gene expression in the central nervou s system (CNS) remain \nunclear in individuals with long COVID \nMethods or Background: First, CBF pattern was computed from arterial spin \nlabeling sequence in long COVID. Next, using CNS ge ne expression data from \nthe AHBA transcriptomic dataset, we conducted the s patial correlation between \nCBF and gene expression to defined the CBF-related genes. Functional \nenrichment analyses were applied to understand the biological functions of \nCBF-related genes. The cell type-specific expressio n analyses is utilized to \nidentify the CNS cell types most closely associated  with long COVID-19 \nResults or Findings: Lower CBF in left frontal-temporal gyrus was associ ated \nwith higher fatigue and worse cognition in long COV ID. This CBF pattern was \nspatially associated with the expression of 2,178 g enes, which significantly \noverlap with the genes reported to interact with SA RS CoV-2 proteins (odds \nratio= 1.60, P= 0.0036). Functional enrichment anal yses indicated these 2,178 \ngenes were enriched in the molecular functions and biological pathways of \nCOVID-19. Additionally, these genes were strongly a ssociated with the \noligodendrocyte progenitor cells, astrocytes, and m yelinating oligodendrocytes \nof the cortex. ALL above results were corrected for  multiple comparisons \nConclusion: Lower CBF is associated with persistent clinical sy mptoms in \nlong COVID individuals, possibly as a consequence o f the complex interactions \namong multiple COVID-19-related genes, which contri butes to our \nunderstanding of the impact of adverse CNS outcomes  and the trajectory of \ndevelopment to long COVID. \nLimitations: The heterogeneity of clinical symptoms at the time of scanning \nmay affect gene expression in individuals with long  COVID. \nFunding for this study: The author(s) disclosed receipt of the following \nfinancial sup port for the research, authorship, an d/or publication of this article: \nThis work was supported by the COVID-19 Research Pr oject of the leading \nmedical discipline in Jiangxi Province, Jiangxi Pro vince Double Thousand \nTalent Plan (jxsq2023201039), Clinical Research Cen ter for Medical Imaging \nIn Jiangxi Province (20223BCG74001) and Jiangxi Pro vince Key Laboratory \nfor Precision Pathology and Intelligent Diagnosis ( 2024SSY06281). \nEthics committee - additional information: The present study was approved \nby the Ethics Committee of the First Affiliated Hos pital of Nanchang University \n(IIT2023018) \nAuthor Disclosures:  \nYao Wang: Nothing to disclose \nFuqing Zhou: Nothing to disclose \n \n \n \n12:30-13:30 Research Stage 3 \nResearch Presentation Session: Paediatric \nRPS 1912 \nImaging the growing brain \n \nModerator \nC. Carducci; Rome/IT  \n(chi.carducci@gmail.com) \n \n \nChildren's brain development is linked to the traje ctory of epigenetic-\nbased inflammatory scores \nJ. Chuah, A. M. A. Manahan, S. Y. Chan, H. Pei, M. Fortier, M. Meaney,  \n*A. P. Tan*; Singapore/SG \n(dnrtanap@nus.edu.sg) \n \nPurpose or Learning Objective: Dysregulation of immune activation has \nbeen consistently shown in patients with mental hea lth disorders. In this study, \nwe mapped the trajectories of DNA-methylation based  inflammation scores \nand examined how these trajectories relate to expos ure to maternal \ndepression, subsequent cognitive outcomes, and brai n development at multiple \nlevels. \nMethods or Background: Inflammation scores were calculated for 293 \nchildren from DNA methylation data based on epigeno me-wide association \nstudies of serum C-reactive proteins at ages 9 and 48 months. We stratify \nthese children into quartiles based on their inflam mation scores at baseline and \nmap the trajectories of inflammation scores for eac h quartile. Next, we \nexamined if children with different inflammation sc ore trajectories have different \nexposure to maternal depression, executive function  performance, and brain \nchanges evaluated using multimodal MRI at ages 4.5,  6.0, and 7.0 years. \nResults or Findings: We observed a decreasing trend of our DNA-methylati on \nbased inflammation scores, primarily driven by chil dren with higher levels of \ninflammation scores at baseline. Children with lowe r levels of inflammation \nscores at 9M and a slower decrease in inflammation scores between 9M and \n48M were exposed to higher levels of maternal depre ssive symptoms and \nshowed poorer executive function performance at age s 4.5 and 7. Children \nwith different inflammation score trajectories exhi bit significantly different brain \nstructure and function, involving predominantly bra in regions involved in \nexecutive function performance, emotion, and reward  processing. \nConclusion: Children with lower baseline inflammation scores an d slower rate \nof decrease across childhood are exposed to higher levels of maternal \ndepression, possibly related to a blunted immune re sponse from chronic stress \nexposure. This has a significant downstream impact on cognitive and brain \ndevelopment at multiple levels. \nLimitations: Evaluation of executive function performance with q uestionnaires \nFunding for this study: This research was supported by grants \nNMRC/TCR/004-NUS/2008 and NMRC/TCR/012-NUHS/2014 fr om the \nSingapore National Research Foundation (NRF) under the Translational and \nClinical Research Flagship and grant OFLCG/MOH-0005 04 from the Open \nFund Large Collaborative Grant Programmes and admin istered by the \nSingapore Ministry of Health’s National Medical Res earch Council (NMRC), \nSingapore. In RIE2025, GUSTO is supported by fundin g from the NRF’s \nHuman Health and Potential (HHP) Domain, under the Human Potential \nProgramme. Additional funding was provided by the S ingapore Institute for \nClinical Sciences, Agency for Science Technology an d Research (A*STAR), \nSingapore. MJM is supported by funding from the Hop e for Depression \nResearch Foundation, USA, the Toxic Stress Network of the JPB Foundation, \nUSA, the Jacobs Foundation, Switzerland, and the NR F and A*STAR’s Human \nPotential Programme (H22P0M0001), Singapore. SYC is  supported by funding \nfrom the NMRC Open Fund – Young Individual Research  Grant (MOH-001149-\n00). EHT is supported by the NMRC Clinician-Scienti st Award (CSA) (MOH-\n001415). APT is supported by funding from the NMRC Transition Award (MOH-\n001273-00) and A*STAR (Brain-Body Initiative, iGran ts call ID #21718). \nEthics committee - additional information: The study was approved by the \nNational Healthcare Group Domain Specific Review Bo ard (D/2009/021 and \nB/2014/00411) and the SingHealth Centralized Instit utional Review Board \n(D/2018/2767 and A/2019/2406). All investigations w ere conducted according \nto the principles expressed in the Declaration of H elsinki. Written consent was \nobtained from all guardians on behalf of the enroll ed children. \nAuthor Disclosures:  \nAi Peng Tan: Nothing to disclose \nMarielle Fortier: Nothing to disclose \nMichael Meaney: Nothing to disclose \nShi Yu Chan: Nothing to disclose \nJasmine Chuah: Nothing to disclose  \nHuang Pei: Nothing to disclose \nAisleen M. A. Manahan: Nothing to disclose \n\n \n \nSaturday \nAbstract-based Programme \n \n 224  \nGeometric microstructural changes of white matter i n infants with \nperiventricular white matter injury and spastic cer ebral palsy \n*M. Wang*¹, H. Zhu², T. Huang³, J. Cheng², H. Jiang ¹; ¹Wuxi/CN, ²Beijing/CN, \n³Zhengzhou/CN \n(wyflipped@gmail.com) \n \nPurpose or Learning Objective: To investigate the geometric microstructural \nchanges in white matter in infants with periventric ular white matter injury and \nspastic cerebral palsy (PWMI-SCP) and facilitate ea rly prediction. \nMethods or Background: PWMI is a high-risk factor for SCP. However, early \nidentification of PWMI-SCP infants remains challeng ing. A novel mathematical \nframework, “Director Field Analysis” (DFA), reflect s changes in the \nmicrostructural geometry of white matter and offers  new insights into the \npathological mechanisms of PWMI-SCP. DFA provides s pecific quantitative \nmetrics, including splay, twist, bend, and total di stortion index. Analysis of \nvariance, correlation analysis, and receiver operat ing characteristics analysis \nwere performed. Corrected p-values < 0.05 were cons idered significant. \nResults or Findings: The PWMI-SCP group exhibited significantly elevated  \nDFA metrics, primarily in the corpus callosum, post erior thalamic radiata, and \ncorona radiata, comparing to the PWMI without SCP g roup, which were \nassociated with enlarged lateral ventricles, reduce d deep nuclear volumes and \nmotor dysfunction. Mediation analysis indicated tha t increased geometric \nmicrostructure in the corpus callosum partially med iates the relationship \nbetween the lateral ventricles and motor function. A multi-parameter model \nbased on DFA metrics can effectively predict PWMI-S CP with an AUC of 0.95. \nConclusion: Abnormal increases in white matter geometric micros tructure in \nthe sensorimotor circuit may be one of the neural s ubstrates underlying the \nmanifestation of SCP in PWMI infants. Monitoring fi ber-orientational alterations \nmay provide new insights into early prediction of P WMI-SCP. \nLimitations: First, this study focused on infants aged 6−36 mo n ths ; the re fo re , \nfuture research should investigate changes in white  matter geometric \nmicrostructure in infants aged 0-6 months to improv e the early diagnosis rate \nof PWMI-SCP. Additionally, biomolecular research is  required to understand \nthe interactions between brain morphology, tissue m echanics, and white matter \ngeometric microstructure to elucidate the pathophys iological mechanisms of \nPWMI-SCP. \nFunding for this study: This study was financially supported by grants from  \nthe STI 2030 - Major Project (Grant No. 2022ZD02090 00); the Hong Kong \nglobal STEM scholar scheme; the internal fund of th e Hong Kong Polytechnic \nUniversity; the Wuxi Municipal Health Commission's 'Double Hundred' Medical \nHealth Young Elite Talent Project (Grant No. BJ2023 088) and the Wuxi \nMunicipal Science and Technology Bureau's Medical a nd Health Tackling \nProject (Grant No. Y20232012); the Jiangsu Province  Graduate Research \nInnovation Project (KYCX24_2647); National Natural Science Foundation of \nChina (Grant No. 61971017, No. 82204933); Open Rese arch Fund of the State \nKey Laboratory of Cognitive Neuroscience and Learni ng (Grant No. \nCNLZD2101). \nEthics committee - additional information: The study received institutional \nreview board approval, and written informed consent  was obtained from all \nparticipants. \nAuthor Disclosures:  \nJian Cheng: Nothing to disclose \nHua Zhu: Nothing to disclose \nTingting Huang: Nothing to disclose \nMiaoyan Wang: Nothing to disclose \nHaoxiang Jiang: Nothing to disclose \n \n \nAltered coupling of cerebral blood perfusion and ne uronal activity in \nchildren with MRI-negative drug refractory Epilepsy  \n*H. Ran*, T. Zhang; Zunyi/CN \n(haifengran_zmu@163.com) \n \nPurpose or Learning Objective: Drug refractory left temporal lobe epilepsy \n(DRLTLE) often give rise to neuronal activity and c erebral vascular \nhemodynamics changes, which may result in neurovasc ular decoupling. \nHowever, neuroimaging evidence on neurovascular dec oupling remains \nscarce, this study aimed to assess the manifestatio n of neurovascular coupling \n(NVC) in childhood DRLTLE using resting-state fMRI (rs-fMRI) and arterial spin \nlabeling imaging (ASL). \nMethods or Background: Based on the collected rs-fMRI and ASL imaging \ndata, degree centrality (DC) and cerebral blood flo w (CBF) were calculated \nrespectively. Across voxel CBF-DC correlations used  to evaluate the NVC \nwithin whole brain, and NVC of brain region was ass essed by the CBF/DC \nratio. We performed correlation analysis to evaluat e the relationship between \nthe variables. Finally, we explored classification problems between DRLTLE \nand healthy control (HC). \nResults or Findings: Compared HC group, the DRLTLE children with higher \nacross voxel CBF-DC correlations. The brain regions  of abnormal CBF, DC, \nand CBF/DC ratio in predominantly in the default mo de and the executive \ncontrol network, the abnormally CBF, DC values in s ome brain regions were \nsignificantly correlated cognitive function. The cl assification model using \nCBF/DC ratio as features achieved the 72.8% accurac y, 0.764 area under the \ncurve, 68.5% sensitivity, 87.5% specificity, the cl assification accuracy were \nhigher than the model using CBF or DC feature. \nConclusion: The study reveals the cerebral blood perfusion, neu ronal activity, \nglobal and regional NVC alteration in children with  MRI-negative DRLTLE non-\ninvasively, associated with lower cognitive perform ance. These findings \nindicating that NVC-based study can better integrat e information of neuronal \nactivity and cerebral hemodynamics, offering a new insight into the \nneuropathological mechanisms of DRLTLE, and NVC may  help clinical \nclassification for childhood DRLTLE. \nLimitations: The sample size was relatively small and the potent ial impact of \nantiepileptic drugs could not totally eliminated. \nFunding for this study: This study was supported by National Natural \nScience Foundation of China (Grant Nos .82171919) a nd Intelligent Medical \nImaging Engineering Research Center of Guizhou High er Education \nInstitutions project (Grant No. Qianjiaoji [2023] 0 38) \nEthics committee - additional information: The ethic committee of Zunyi \nMedical University reviewed and granted ethical app roval of this research \n(Ethical Batch Number: lunshen [2021] 1-080) \nAuthor Disclosures:  \nTijiang Zhang: Nothing to disclose \nHaifeng Ran: Nothing to disclose \n \n \nMagnetic resonance imaging of children with the use  of an incubator \nP. D. Mika, J. Cydejko, *B. Rowinski*, D. Świętoń, E. Szurowska; Gdańsk/PL \n(browinski@uck.gda.pl) \n \nPurpose or Learning Objective: Important limitation of pediatric MRI imaging \nis requirement of general anesthesia (GA) in the yo ungest group of patients. \nImaging in GA is not only stressful for children bu t also extends the time of the \nprocedure. The solution is to use an incubator dedi cated to work in a magnetic \nfield. The MRI dedicated incubator allows a full sp ectrum of MRI imaging. \nAnalysis of examinations with the use of an MRI inc ubator performed at the \nDepartment of Radiology of the University Clinical Center in Gdansk. \nMethods or Background: A retrospective evaluation of MRI examinations \nperformed using an incubator carried out in the you ngest group of patients. \nThe study includes an analysis of the tests perform ed in terms of the child's \nage and the examined area. The paper presents the b enefits of using an MRI \nincubator. Sequences prepared specifically for test  protocols in which the \nincubator is used are discussed Patients were teste d in the \"feed and sleep\" \nprotocol. \nResults or Findings: Between 2020 and 2023, a total of 146 MRI \nexaminations were performed. Most of the studies we re brain scans without \nthe administration of a contrast agent. Only in 1 c ase supplementary general \nsedation was necessary. \nConclusion: 1. The study showed a sharp increase in the number of studies \nusing an incubator. 2. The incubator dedicated for MRI examinations allows to \nlimit the amount of anesthesia in the youngest pati ents to 4 kg of body weight. \n3. Increase in the cost of effectiveness of MRI exa minations in the youngest \ngroup of patients due to the lack of anesthesia cos ts. \nLimitations: Movement artifacts \nWeight over 4kg \nImplants not allowed for MRI \nFunding for this study: No \nEthics committee - additional information: No \nAuthor Disclosures:  \nPaulina Danuta Mika: Nothing to disclose \nEdyta Szurowska: Nothing to disclose  \nBartosz Rowinski: Nothing to disclose \nDominik Świętoń: Nothing to disclose \nJoanna Cydejko: Nothing to disclose \n \n \nBrain Perfusion Imaging by Arterial Spin Labelling Predicts Postsurgical \nSeizure Freedom in Pediatric Focal Lesional Epileps y \n*A. G. Gennari*, L. Gaito, D. Cserpan, R. Kottke, R . Tuura O’Gorman,  \nG. Ramantani; Zurich/CH \n(gennari_antonio@libero.it) \n \nPurpose or Learning Objective: In children with pharmacoresistant focal \nlesional epilepsy, lesion-associated brain perfusio n changes captured by \narterial spin labelling (ASL) are an emerging imagi ng tool improving lesion \ndetection. However, their correlation with postsurg ical seizure outcomes is still \nunexplored. This study aims to determine whether in cluding ASL-derived \nperfusion changes in surgical planning is associate d with favorable \npostsurgical seizure outcomes in children with foca l cortical dysplasia (FCD) or \nlow-grade epilepsy-associated tumors (LEAT). \nMethods or Background: We retrospectively analyzed MRI scans from 18 \nchildren (median age at MRI: 4.8 years, IQR: 1.9–11 .5) who underwent \nsurgical resection for pharmacoresistant epilepsy a nd had at least 1 year of \npost-surgical follow-up. All patients received pres urgical ASL imaging along \n\n \n \nSaturday \nAbstract-based Programme \n \n 225  \nwith pre- and postsurgical structural MRI. Image po stprocessing, including \nsegmentation and coregistration, was used to qualit atively and quantitatively \nevaluate the completeness of resection of both the anatomical lesion and the \nASL-detected perfusion changes. The DICE similarity  index was adopted in \nquantitative analysis to grade the segmentations’ a lignment. These findings \nwere then correlated with seizure outcomes. \nResults or Findings: Fourteen (78%) patients achieved complete seizure \nfreedom. Qualitative analysis showed that complete resection of the ASL-\ndetected perfusion changes significantly correlated  with seizure freedom \n(p=0.009). Quantitative analysis indicated that hig her degrees of alignment \nbetween perfusion and resection cavity segmentation s, as measured by DICE \nscore, were associated with seizure freedom (p=0.04 3), while lesion volume \ninclusion was not (p=0.44). \nConclusion: Including ASL perfusion imaging in the presurgical evaluation can \nhelp better define the epileptogenic zone, improvin g postsurgical seizure \noutcomes, and supporting it as a complementary tool  in surgical planning for \npharmacoresistant pediatric focal lesional epilepsy . \nLimitations: Limitations: - small sample size; - focus on MRI-vi sible lesions \nonly, limiting the generalizability to MRI-negative  patients; - manual \nsegmentation, which may limit the reproducibility o f our results. \nFunding for this study: We thank the National Science Foundation (SNSF: \n208184) (to G.R.), the Anna Mueller Grocholski Foun dation, and the Theodor \nund Ida Herzog-Egli-Stiftung (to A.G.G.) for fundin g. The funders had no role in \nthe design or analysis of the study. \nEthics committee - additional information: The collection and analysis of \npatient data were approved by and performed accordi ng to the guidelines and \nregulations of the local ethics committee (KEK-ZH P B-2024-00298). All parents \ngave written, informed general consent to reuse cli nical data for research. \nAuthor Disclosures:  \nRuth Tuura O’Gorman: Nothing to disclose \nAntonio Giulio Gennari: Research/Grant Support: The  Anna Mueller Grocholski \nFoundation and the Theodor und Ida Herzog-Egli-Stif tung funded Dr. Gennari \npost-doc \nDorottya Cserpan: Nothing to disclose \nGeorgia Ramantani: Research/Grant Support: Prof Ram antani received a grant \nfrom the National Science Foundation (SNSF: 208184)  \nRaimund Kottke: Nothing to disclose \nLuca Gaito: Nothing to disclose \n \n \nRefining Diagnostic Accuracy in Pediatric Metabolic  Brain Disorders: \nIntegrating MRI, Proton Spectroscopy, and Diffusion -Weighted Imaging \n*R. Agarwal*¹, U. Gupta², N. Jha²; ¹Bengaluru/IN, ² Ghaziabad/IN \n(aritika1994@gmail.com) \n \nPurpose or Learning Objective: This study aims to assess MRI signal \nabnormalities in pediatric metabolic brain disorder s, focusing on diffusion-\nweighted imaging (DWI) and proton magnetic resonanc e spectroscopy (MRS). \nThe goals are to identify specific imaging patterns , correlate them with clinical, \nbiochemical, and genetic data, and enhance diagnost ic accuracy for early \nintervention. \nMethods or Background: Metabolic brain disorders in children are inherited  \nconditions leading to progressive neurodegeneration , where early diagnosis is \ncritical. This study evaluated 30 pediatric patient s (aged 0-12 years) with \nsuspected metabolic brain disorders using MRI, DWI,  and proton MRS, along \nwith biochemical and genetic testing. MRI findings were categorized based on \nthe involvement of white matter, grey matter, or bo th. \nResults or Findings: Of the 30 patients, 24 (80%) were diagnosed with \nmetabolic brain disorders, predominantly in males ( 79.17%) and in the 0-3 \nyears age group (62.5%). Common symptoms included r egression of \ndevelopmental milestones (70.83%) and seizures (58. 33%). MRI showed white \nmatter involvement in 10 cases, grey matter in 5, a nd both in 9. Diagnoses \nincluded X-linked adrenoleukodystrophy, metachromat ic leukodystrophy, Leigh \ndisease, Wilson disease, glutaric aciduria type I, and neuronal ceroid \nlipofuscinosis. MRI findings included symmetrical T 2 hyperintensities in \n66.67%, diffusion restriction in 41.67%, and distin ctive MRS peaks such as \nelevated N-acetylaspartate in Canavan disease and l actate in Leigh disease. \nConclusion: DWI and proton MRS are crucial for early diagnosis of pediatric \nmetabolic brain disorders. Identifying characterist ic imaging patterns enhances \ndiagnostic precision and facilitates timely interve ntion, improving patient \noutcomes. \nLimitations: The study's small sample size may limit the general izability of the \nfindings, and the cross-sectional design does not a ddress disease progression \nover time. Future research should involve larger co horts and longitudinal \nstudies to validate these results and assess the lo ng-term efficacy of imaging-\nbased diagnostic methods. \nFunding for this study: None \nEthics committee - additional information: Not required \nAuthor Disclosures:  \nRitika Agarwal: Nothing to disclose \nNarendran Jha: Nothing to disclose \nUjjwal Gupta: Nothing to disclose \nDynamic network dysfunction in children with idiopa thic generalized \nepilepsy and its association with cognitive impairm ent and gene \nexpression profiles \n*H. Ran*, K. Huang, T. Zhang; Zunyi/CN \n(haifengran_zmu@163.com) \n \nPurpose or Learning Objective: Idiopathic generalized epilepsy(IGE) has \nbeen considered as a network disease, recurrent sei zures may result in \nnetwork reconfiguration and cognitive impairments. The dynamic changes in \nfunctional network in IGE children and the relation ship with cognitive \nimpairment and gene expression profiles needs to be  explored. \nMethods or Background: 26 IGE children and 35 healthy controls(HC) were \nrecruited, the modular variability(MV) of construct ed time-varying multi-layer \nnetwork was calculated and compared between groups based on rs-fMRI. The \ncorrelation analysis was performed between MV and c ognitive function scores \nand clinical variables. Allen Human Brain Atlas wer e used to identify gene sets \nassociated with dynamic network remodeling in IGE. The associated biological \nprocesses, pathways were identified by gene enrichm ent tools. \nResults or Findings: Compared to HC, IGE children demonstrated changed \nMV mainly located in the sensorimotor areas, salien ce/ventral attention, and \ndefault mode network, and at the sub-network level,  children with IGE exhibited \nincreased MV in the default mode network(p<0.05, FD R). MV changes in the \nleft prefrontal, precuneus cortex were negatively c orrelated with the verbal IQ, \nfull scale IQ and performance IQ scores, respective ly(r=-0.400, -0.419, -0.408; \np=0.042, 0.032, 0.038), while MV in the right orbit ofrontal cortex was positively \ncorrelated with the verbal IQ and full scale IQ sco res, respectively(r=0.488, \n0.442; p=0.011, 0.023). Gene expression profiles wa s associated with dynamic \nnetwork dysfunction in IGE(r=0.499, pperm  < 0.05). Enrichment analysis \nindicated that the genes related to dynamic network  reorganization were \nprincipally enriched in dendrite, axon, and nervous  system development. \nConclusion: In IGE children, altered dynamic functional network s has been \nidentified and correlated with cognitive function a nd gene expression, revealing \nthe complex relationship between the dynamic change s of macroscopic \nmodules and genetic pathological mechanisms in IGE patients. \nLimitations: The sample size of this study was relative small. \nFunding for this study: This study was supported by National Natural \nScience Foundation of China (Grant Nos .82171919) a nd Intelligent Medical \nImaging Engineering Research Center of Guizhou High er Education \nInstitutions project (Grant No. Qianjiaoji [2023] 0 38) \nEthics committee - additional information: Ethics approval of this research \nwas granted by the Ethic Committee of the Affiliate d Hospital of Zunyi Medical \nUniversity[KLL-2021-347] \nAuthor Disclosures:  \nTijiang Zhang: Nothing to disclose \nKexin Huang: Nothing to disclose \nHaifeng Ran: Nothing to disclose \n \n \n12:30-13:30 Research Stage 4 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 1905 \nArtificial intelligence in abdominal and \noncological imaging \n \nModerator \nG. P. Raval; Rajkot/IN  \n(drgaurang.raval@gmail.com) \n \n \nTotal Segmentator: Integration and validation into PACS workstation for \nabdominal CT scans. A feasibility study \n*G. Lappas*¹, N. Patlakas¹, P. Giannikopoulos¹, M. Triantafyllou²,  \nG. I. Kalaitzakis³, M. Klontzas², K. Petropoulos¹; ¹Athens/GR, ²Crete/GR, \n³Heraklion/GR \n \nPurpose or Learning Objective: This study aims to validate the performance \nof the open-source Total Segmentator for the segmen tation of abdominal \norgans in CT scans and integration into the existin g PACS workstation. \nMethods or Background: The model's segmentation capability was quantified \nusing Dice Similarity Coefficient (DSC) and Normali zed Surface Distance \n(NSD) across five datasets (N=1243). Data variabili ty including statistics \nradiomics analysis was performed. Grad-CAM and Mont e-Carlo were deployed \nfocusing on the understanding of model decision-mak ing and robustness, \nrespectively. Οne assistant professor of radiology and one senior radiology \nresident rated the predicted segmentations’ quality  from Greek hospital CT \nscans (N=100) and integration into the clinical rou tine. \n\n \n \nSaturday \nAbstract-based Programme \n \n 226  \nResults or Findings: The model demonstrated high accuracy in segmenting \nmost organs, e.g., DSC - CI: 0.85-0.97, while showe d lower performance on \ngallbladder, pancreas and prostate, e.g., DSC – CI:  0.71-0.85. Those results \ndepict the robustness of the model across the five datasets which present high \nvariability depicted by normalized volume ( μ±σ: 0.24±0.18) and normalized \nintensity (μ±σ: 0.39±0.21) coming in agreement with radiomics fin dings. Grad-\nCAM and Monte-Carlo noise results provided insights  into the model's \ndecision-making process and highlighting areas for potential improvement. The \nclinical feasibility study indicates a promising to ol generating high quality \nsegmentations with 97% of those requiring minimal m anual adaptations. \nConclusion: Total Segmentator showed robust performance in segm entation \nof the major abdominal organs with poorer performan ce for gallbladder, \npancreas and prostate. The integration into PACS wo rkstation proves the \nmodel's potential for routine use in Greek hospital s marking a significant step \ntowards more efficient radiological flow. \nLimitations: The model had lower performance for gallbladder, pa ncreas and \nprostate while 7% of the validated cases had disagr eements larger than 50% \nvolume-wise. The available number of CT scans used for rating by radiologists \nwas limited. \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMichail Klontzas: Nothing to disclose \nKonstantinos Petropoulos: Board Member: IKnowHealth  S.A. \nNektarios Patlakas: Nothing to disclose \nGeorgios Ioannis Kalaitzakis: Nothing to disclose  \nPetros Giannikopoulos: Nothing to disclose \nGeorgios Lappas: Nothing to disclose \nMatthaios Triantafyllou: Nothing to disclose \n \n \nResults of the ULS23 Challenge on automatic 3D univ ersal lesion \nsegmentation in computed tomography \nM. J. J. De Grauw¹, E. Scholten¹, E. J. Smit¹, M. J . Rutten², B. Van Ginneken¹, \nM. Prokop¹, *A. Hering*¹; ¹Nijmegen/NL, ²'S-Hertoge nbosch/NL \n(alessa.hering@radboudumc.nl) \n \nPurpose or Learning Objective: Generalizable automatic segmentation \nmethods are well-suited for application in the dive rse clinical contexts \nencountered during tumour follow-up in CT. The ULS2 3 challenge establishes \nthe state-of-the-art in automatic 3D lesion segment ation quality, measurement \naccuracy, and prediction robustness. \nMethods or Background: Current benchmarks often focus on organ-specific \nlesion segmentation, yet diverse clinical cases req uire fast, generalist models. \nThe ULS23 challenge focuses on universal 3D lesion segmentation across \nchest-abdomen-pelvis CT, with 38,693 diverse lesion s in the training dataset. \nThe evaluation dataset contains 775 clinically rele vant lesions from 284 \npatients across two Dutch tertiary care centers. We  developed a strong \nbaseline method based on the nnUnet and invited the  research community to \nsubmit their solutions to the challenge. Post-chall enge, we conducted \nexperiments to explore lesion type influence, uncer tainty, and robustness. \nResults or Findings: The ULS23 challenge encouraged over 50 internationa l \nresearchers to develop solutions for automatic 3D l esion segmentation. During \nthe official challenge period, 153 submissions were  recorded during the \ndevelopment phase with seven teams submitting to th e final leaderboard. The \nU-mamba framework achieved the highest challenge sc ore, excelling in \nsegmentation quality (70.8% ± 23.5% Dice), axial measurement accuracy \n(Long-axis 10.3%, Short-axis 11.8% Symmetric Mean A bsolute Percentage \nError), and robustness when evaluated using repeate d segmentation (79.7% ± \n24.2% Dice). Bone, pancreas and colon lesions remai n challenging, and while \nsegmentation consistency improves with performance,  models still showed \nsignificant variability, affecting measurement accu racy. \nConclusion: The results of the ULS23 challenge demonstrate the potential of \n3D universal lesion segmentation using large, aggre gated datasets as a viable \nalternative to organ-specific models. However, sign ificant variability in \nsegmentation performance, particularly for bone, pa ncreas and colon lesions, \nindicates the need for further improvements. Addres sing these performance \ninconsistencies could help reduce variance and enha nce overall clinical \napplicability. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nMathias Prokop: Patent Holder: Mevis Medical Soluti on Consultant: Canon \nMedical Systems, Siemens Healthineer Shareholder: T hirona \nErnst Scholten: Nothing to disclose \nMax Jacobus Johannes De Grauw: Nothing to disclose \nAlessa Hering: Nothing to disclose \nBram Van Ginneken: Founder: Plain, Thirona \nEwoud J. Smit: Nothing to disclose \nMatthieu J.C.M. Rutten: Nothing to disclose \n \nAutomated Matching of Lesions in Cancer Follow-Up U sing 3D Siamese \nNeural Networks and CT \n*A. Vergara*, A. Jimenez-Pastor, A. Alberich-Bayarr i; Valencia/ES \n(alejandrovergara@quibim.com) \n \nPurpose or Learning Objective: There is a lack of automated and reliable \ntools to automatically track lesion changes over ti me, as well as to streamline \ntreatment response reporting, such as RECIST-1.1 an d others. The primary \ngoal of this work was to address the automatic matc hing of lesions in the \nthoracoabdominal region across timepoints through 3 D-Siamese Neural \nNetworks (SNN). \nMethods or Background: A retrospective dataset of 253 longitudinal CT \nexams from metastatic NSCLC patients was used, with  a high variability in \nlesion location, including lymph nodes/lungs/liver/ adrenal glands \n(43.87%/34.02%/11.82%/6.17%, respectively) and volu me (0.1-1000cm3). A \n3D-SNN architecture was used to compare the similar ity between tumor pairs \nand matching the same lesion across consecutive sca ns. The final model \nresulted from a two-step training process, initiall y evaluating 144 \nhyperparameter settings and subsequently retraining  using the top-performing \nones. Different configurations were created combini ng input information (CT \nimage, CT image and lesion segmentation mask, and C T image with several \nHU windows), learning rate (LR), LR schedulers, con volutional block \ncomplexity (1/2 convolutional layers), and loss fun ctions (contrastive and BCE). \nAn 80/20 training/test split ensured comparable les ion size and location \nheterogeneity in both sets. \nResults or Findings: The best model employed the CT image as input, a LR  \nof 10e-4, a step LR scheduler, BCE loss, and 1-conv olutional-layer block, \ndelivering 92.80%/91.10%/92.4% in accuracy/precisio n/recall in the test set. \nThe accuracy decomposed by location was 93.79%/88.3 2%/92.50%/94.35% \nfor lymph nodes, lung, liver and adrenal gland, res pectively. Considering lesion \nvolumes, the accuracy was 90.57%/93.35%/94.48% for small (<10cm3), \nmedium (10-100cm3), and large (>100cm3) ones. \nConclusion: 3D SNNs are a promising and reliable technique to a ccurately \nmatch tumor lesions over consecutive timepoints, re gardless of lesion size and \nregion. \nLimitations: Current work focuses on matching lesions individual ly, not the \nentire patient, which is future work. \nFunding for this study: None \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nAlejandro Vergara: Nothing to disclose \nAna Jimenez-Pastor: Nothing to disclose \nAngel Alberich-Bayarri: Nothing to disclose \n \n \nComparative Analysis of Language Models for Automat ed Interpretation \nof Longitudinal TACE Reports in Hepatocellular Carc inoma \n*E. Can*¹, E. Kotter¹, K. Vogt¹, M. Brönnimann², A.  Elkilany³, W. Uller¹,  \nK. Bressem⁴, L. C. Adams⁴; ¹Freiburg/DE, ²Bern/CH, ³Leipzig/DE, ⁴Munich/DE \n(elif.can@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: The increasing complexity of radiology data \nin hepatocellular carcinoma (HCC) requires innovati ve solutions to ensure \nconsistent and efficient interpretation. This study  aims to evaluate the \nperformance of four leading language models (GPT, G emini, Llama, and \nLlama405b) in extracting and interpreting key clini cal data from longitudinal \nTACE reports. By automating this process, we seek t o reduce the burden on \nradiologists, enhance decision-making accuracy, and  improve workflow \nefficiency in interventional radiology. \nMethods or Background: We analyzed the performance of each model on 50 \nanonymized TACE reports. The models were assessed f or accuracy in \nextracting clinical data (diagnosis, BCLC staging, mRECIST assessment), as \nwell as vascular involvement and metastases identif ication. A detailed error \nanalysis was conducted to evaluate consistency and precision across tasks. \nPerformance metrics included diagnosis accuracy, li ver segment identification, \nand error severity. \nResults or Findings: All models demonstrated high accuracy (90-100%) in \nbasic tasks such as diagnosis identification and pr ocedure date extraction. \nHowever, in complex tasks like mRECIST assessment a nd lymph node status \nevaluation, performance varied significantly. Gemin i outperformed in segment \nidentification (4.6/5) and vascular involvement (2. 9/5), while all models \nstruggled with mRECIST accuracy (0-10%). Llama and Llama405b exhibited \nslightly higher error rates (7.7, 8.0) compared to GPT (6.5) and Gemini (6.6). \nConclusion: While language models show promise in automating TA CE report \ninterpretation, challenges remain in specialized me dical tasks. Gemini showed \nthe best overall performance, but significant impro vements are needed in \nmRECIST assessment and anatomical reporting. Furthe r research is required \nto refine these models for clinical application, po tentially reducing radiologists’ \nworkload and enhancing decision-making processes. \nLimitations: The models' generalizability across other forms of interventional \nradiology remains to be validated. \nFunding for this study: No funding provided. \n\n \n \nSaturday \nAbstract-based Programme \n \n 227  \nEthics committee - additional information: Approved by the ethics \ncommittee of the University Medical Center Freiburg . \nAuthor Disclosures:  \nElif Can: Nothing to disclose \nLisa C. Adams: Nothing to disclose \nAboelyazid Elkilany: Nothing to disclose \nKatharina Vogt: Nothing to disclose \nMichael Brönnimann: Nothing to disclose \nKeno Bressem: Nothing to disclose \nWibke Uller: Nothing to disclose \nElmar Kotter: Nothing to disclose \n \n \nCT-based Foundation Model Enhanced Preoperative Pre diction of \nMicrovascular Invasion in Hepatocellular Carcinoma:  a multicenter study \n*L. Deng*, W. Xia, J. Xia, W. Dai, F. Yan, R. Li; S hanghai/CN \n(dl13075@rjh.com.cn) \n \nPurpose or Learning Objective: To develop CT-based foundation model and \nmulti-instance learning (MIL) framework to preopera tively predict microvascular \ninvasion (MVI) in hepatocellular carcinoma (HCC). \nMethods or Background: CT-based foundation models were developed \nthrough self-supervised learning using public CT im age datasets. Patients with \npathologically proven HCC were included from two ce nters, and the CT image \nsequences of non-contrast, arterial and portal veno us phase were acquired. \nThe features from slices of HCC tumor region were e xtracted by foundation \nmodels, and the features were aggregated by MIL to predict MVI status. The \npredictions of all sequences were combined to obtai n a final prediction by \nlogistics regression. The performance of proposed m ethod was evaluated by \narea under the receiver operating characteristic cu rve (AUC). \nResults or Findings: The CT image patches of 36,811 lesions from \nDeepLesion, LiTS and 3D-IRCADb were used for founda tion model \ndevelopment. A total of 617 HCC patients (median ag e, 69 years; IQR, 52-67 \nyears; men 510) were included and divided into trai ning set (center 1, n=493) \nand independent test set (center 2, n=124). By usin g a few slices of HCC \ntumor region (median number, 3; IQR, 2-4), the prop osed method achieved \nAUCs of 0.80, 0.78, 0.74, and 0.83 for non-contrast , arterial phase, portal \nvenous phase, and all sequences combined, respectiv ely. The proposed \nmethod significantly outperformed the method withou t foundation model \n(AUC=0.72, 0.67, 0.71 and 0.64 for each sequence, P <.05) and the radiomics \nmodel (AUC=0.72, P<.05). \nConclusion: By leveraging a few representative CT slices and av oiding the \nneed for full tumor delineation, the foundation mod el-based method achieves \nimproved performance compared to previous methods, demonstrating the \ncrucial role of the foundation model for accurate M VI prediction and facilitating \nmore efficient clinical decision-making. \nLimitations: Not applicable. \nFunding for this study: Not applicable. \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nWei Xia: Nothing to disclose \nRuokun Li: Nothing to disclose \nFuhua Yan: Nothing to disclose \nJi Xia: Nothing to disclose  \nWenwen Dai: Nothing to disclose \nLin Deng: Nothing to disclose \n \n \nArtificial intelligence for evaluation of magnetic resonance imaging-\ndetected extramural vascular invasion in rectal can cer \n*H. Huang*, K. Zhao, Z. Liu, C. Liang; Guangzhou/CN  \n(huanghaitao@gdph.org.cn) \n \nPurpose or Learning Objective: Extramural vascular invasion (EMVI) is a \ndetectable magnetic resonance imaging (MRI) marker that reflects both tumor \ninvasive and metastatic potential. Since EMVI+ is s een as an independent \nindicator of worse prognosis, clinicians may pursue  more intensive treatment \nstrategies for affected patients. However, EMVI ass essment was influenced by \nobserver experience and subjective factors, limitin g its practical effectiveness. \nThis study aims to develop and validate an interpre table deep learning-based \napproach for automated mrEMVI identification throug h voxel-level \nsegmentation, providing objective and consistent de tection method. \nMethods or Background: This is a multicenter study that included a total o f \n2,501 rectal cancer patients, with 1,830 in the tra ining cohort and 671 in the \nvalidation cohorts. Dice similarity score was used to measure segmentation \nperformance, while the inter-reader agreement of mr EMVI was calculated \nusing Cohen’s Kappa (κ). The prognostic value of mrEMVI statuses identifi ed \nby the artificial intelligence (AI) model was evalu ated by Kaplan–Meier curves \nand the Cox model. \n \n \nResults or Findings: Our model demonstrated excellent performance in \nidentifying mrEMVI, achieving accuracy of 81.54% an d 84.72% in the two \nvalidation cohorts. The model demonstrated a high l evel of inter-reader \nconsistency with senior radiologists in identifying  mrEMVI status (κ: 0.713–\n0.736). AI-mrEMVI+ patients have significantly shor ter 3-year disease-free \nsurvival (DFS) and 5-year overall survival (OS) com pared to AI-mrEMVI− \npatients (DFS: 62.23% vs 84.91%, HR=2.67 95% CI: 1. 95–3.66, P<0.001; OS: \n68.71% vs 87.14%, HR=2.64 95% CI: 1.75–3.97, P<0.00 1). \nConclusion: We provide a more objective and consistent approach  for the \ndetection of mrEMVI, demonstrating potential in pro gnostic prediction, and \noffering promising contributions to optimizing the treatment of rectal cancer \npatients. \nLimitations: the model exhibited some false positives, primarily  because the \nmodel mistakenly identified larger blood vessels in  the mesorectal area as \nmrEMVI. \nFunding for this study: National Science Foundation for Young Scientists of  \nChina (82202267). \nEthics committee - additional information: This study has obtained approval \nfrom the ethics review committees of all participat ing hospitals. Considering the \nretrospective design of the study, the requirement for written informed consent \nfrom patients was waived. \nAuthor Disclosures:  \nChanghong Liang: Author: Administrative support, ma nuscript review \nZaiyi Liu: Author: Administrative support, manuscri pt review \nHaitao Huang: Author: Study design, analyzed the da ta, manuscript Writing \nKe Zhao: Author: Study design, analyzed the data \n \n \nPROVIZ Proof-of-Technology: Performance of a machin e learning \nsoftware for detection of clinically significant pr ostate cancer on \nbiparametric MRI in a prospective clinical study \n*R. Segre*, M. Sunoqrot, G. Nketiah, P. Davik, S. L angorgen, M. Elschot,  \nT. Frost Bathen; Trondheim/NO \n(rebecca.segre@ntnu.no) \n \nPurpose or Learning Objective: PROVIZ is a machine learning software \ndesigned to detect clinically significant prostate cancer (csPCa, defined as \nGGG > 1) on MRI as a reference for targeted biopsy.  The aim of this study is to \nevaluate feasibility (technical issues in < 10% of processed cases), safety \n(absence of Serious Adverse Device Effects, SADEs),  and performance of \nPROVIZ. \nMethods or Background: Prospective, proof-of-technology study on 80 \nconsenting men. Inclusion criteria: biopsy-naive me n undergoing MRI for \nsuspicion of prostate cancer. MR images were first delineated according to PI-\nRADS v. 2.1 by an experienced radiologist. Subseque ntly, automated detection \n(up to three suspicious areas) was performed using PROVIZ. Delineations of \nall lesions were used as a reference for targeted b iopsies, providing the ground \ntruth. \nResults or Findings: To date, 73/80 patients have completed the study. \nRegarding feasibility, one technical issue was expe rienced among the \nparticipants. In terms of safety, no SADEs were obs erved. As to performance, \nPROVIZ scored a high AUROC (92.3%) and can be retro spectively tuned to \nreach the same patient-level sensitivity of the rad iologist at PI-RADS 3 \n(94.6%), while gaining an improvement in specificit y (72.2% vs 52.8%). At this \noperating point, PROVIZ would have referred 7 less patients to biopsy than the \nradiologist (45/73 vs 52/73). On a lesion-level, PR OVIZ showed a slightly lower \nsensitivity than the radiologist (42.9% vs 45.5%) b ut fewer false positives per \ncase (30.1% vs 41.1%). \nConclusion: Preliminary results of this prospective study indic ate that PROVIZ \nis feasible and safe to use, with a performance for  detection of csPCa \ncomparable to an experienced radiologist. As a supp ort tool for the radiologist, \nPROVIZ shows potential for reducing false positive predictions, therefore \nminimizing unnecessary biopsies. \nLimitations: Incomplete data collection. Single-center study. \nFunding for this study: Norwegian University of Science and Technology \n(NTNU), Research Council of Norway (Grant Number 29 5013), The Liaison \nCommittee between the Central Norway Regional Healt h Authority and the \nNorwegian University of Science and Technology (Gra nt Numbers 983005100, \n982992100 and 90368401), St. Olavs Hospital - Trond heim University Hospital, \nCentral Norway Regional Health Authority. \nEthics committee - additional information: REK (Regionale komiteer for \nmedisinsk og helsefaglig forskningsetikk) approval no. 479272. \nAuthor Disclosures:  \nSverre Langorgen: Nothing to disclose \nMatthijs Elschot: Nothing to disclose \nRebecca Segre: Nothing to disclose \nTone Frost Bathen: Nothing to disclose \nMohammed Sunoqrot: Nothing to disclose \nPetter Davik: Nothing to disclose \nGabriel Nketiah: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 228  \nDiagnostic performance of a fully automated AI algo rithm for lesion \ndetection and PI-RADS classification in patients wi th suspected prostate \ncancer \n*H. Engel*¹, A. Nedelcu¹, R. Grimm², H. Von Busch²,  A. Sigle³, T. Krauß¹,  \nJ. Weiß¹, M. Benndorf⁴, B. Oerther¹; ¹Freiburg im Breisgau/DE, ²Forchheim /DE, \n³Freiburg/DE, ⁴Detmold/DE \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of a \nfully automated AI algorithm with lesion detection and PI-RADS classification in \na cohort of consecutive patients verified by target ed and extensive systematic \nbiopsies. \nMethods or Background: This retrospective, single-centre study included \nconsecutive patients who underwent 3T multiparametr ic prostate magnetic \nresonance imaging (MRI) performed between 05/2017 a nd 05/2020, followed \nby targeted transperineal ultrasound-fusion guided and systematic biopsy. The \nAI algorithm (syngo.via Prostate MR, VB60S HF01, Si emens Healthineers) \nwas described in previous publications and is based  on axial T2- and diffusion-\nweighted imaging sequences. The results of the AI a lgorithm were compared \nwith those of human readers and the diagnostic perf ormance was determined. \nResults or Findings: The evaluation of 272 patients resulted in 436 targ et \nlesions. 135 patients (49.5%) had clinically signif icant prostate cancer (csPCa), \n35 (12.8%) had clinically insignificant prostate ca ncer (ISUP=1) and 102 \n(37.5%) were benign. Patient-level cancer detection  rates (CDRs) of csPCa for \nAI versus human reading were 11%/18% for PI-RADS ≤2, 24%/11% for PI-\nRADS 3, 54%/41% for PI-RADS 4, and 74%/92% for PI-R ADS 5. The accuracy \nof the AI was significantly better (0.74 versus 0.6 3 at a threshold of PI-RADS \n≥4, p <0.01). 62 patients with human reading PI-RADS  ≥3 were correctly \nclassified as true negative by AI. \nConclusion: The AI algorithm proved to be a reliable and robust  tool for lesion \ndetection and classification. Furthermore, the CDRs  and distribution of PI-\nRADS assessment categories of the AI are consistent  with the results of recent \nmeta-analyses, indicating precise risk stratificati on. \nLimitations: The limitations of our study are mainly its retrosp ective and \nmonocentric design. Additionally, the study design based on histopathological \nverification implies an under-representation of neg ative MRI scans and a \ncohort that is not fully representative of the wide r patient population. \nFunding for this study: The licence of the AI algorithm was part of an \nunrestricted collaboration agreement between Siemen s Healthineers and the \nDepartment of Radiology, Medical Center - Universit y of Freiburg, Faculty of \nMedicine, University of Freiburg. While Siemens pro vided technical support, \nthe study conception and design, as well as the ana lysis and interpretation of \nthe data, were conducted independently. August Sigl e received research \nsupport within the Berta-Ottenstein-Programme. Othe r than that, the authors \ndeclare that no funds, grants, or other support wer e received during the \npreparation of this manuscript. \nEthics committee - additional information: Approval was granted by the \nEthics Committee of the University of Freiburg (No.  20-1256). \nAuthor Disclosures:  \nTobias Krauß: Nothing to disclose \nBenedict Oerther: Nothing to disclose \nAugust Sigle: Nothing to disclose \nJakob Weiß: Nothing to disclose \nRobert Grimm: Nothing to disclose \nAndrea Nedelcu: Nothing to disclose \nMatthias Benndorf: Nothing to disclose \nHannes Engel: Nothing to disclose \nHeinrich Von Busch: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n14:00-15:30 Research Stage 1 \nResearch Presentation Session: Cardiac \nRPS 2003 \nCardiac applications of photon-counting \nCT \n \nModerator \nF. Catapano; Milan/IT  \n(federica.catapano@humanitas.it) \n \n \nUltrahigh-resolution photon-counting detector CT de tects a significantly \nlower coronary plaque volume than energy-integratin g detector CT \n*M. Vecsey-Nagy*¹, G. Tremamunno¹, C. Gnasso¹, E. Z sarnóczay²,  \nD. Kravchenko², B. Szilveszter², P. Maurovich-Horva t², A. Varga-Szemes¹,  \nT. S. Emrich¹; ¹Charleston, SC/US, ²Budapest/HU \n(vnagymilan@gmail.com) \n \nPurpose or Learning Objective: To evaluate the effect of photon-counting \ndetector (PCD)-CT on coronary plaque quantification  and characterization on \ncoronary CT angiography (CCTA) series compared to e nergy-integrating \ndetector (EID)-CT. \nMethods or Background: Consecutive patients undergoing clinically indicate d \nCCTA on EID-CT (192×0.6 mm collimation) were enroll ed for an ultrahigh-\nspatial-resolution (UHR) PCD-CT scan (120×0.2 mm co llimation) within 30 \ndays. PCD-CT was acquired using equivalent or lower  CT dose index and \nequivalent contrast media volume as the clinical sc an. Total, calcified, fibrotic, \nand low-attenuation coronary plaque volumes were qu antified and compared \nbetween scanners. Intra- and inter-reader reproduci bility was assessed on both \nsystems. \nResults or Findings: A total of 164 plaques from 48 patients were segmen ted \non both scans. Total plaque volume was lower on PCD -CT compared to EID-\nCT (723.5 [interquartile range: 500.6–1184.7] vs. 1 084.7 [IQR: 710.7–1609.8] \nmm3, p<0.001). UHR-based segmentations produced low er fibrotic plaque \nvolumes than EID-CT-based measurements (325.4 [IQR:  151.7–519.2] vs. \n627.7 [IQR: 385.8–795.1] mm3, p<0.001), while low-a ttenuation (72.1 [IQR: \n38.6–161.9] vs. 58.1 [IQR: 23.4–102.3] mm3, p=0.052 ) and calcified plaque \nvolumes (IQR: 344.5 [174.3–605.7] vs. 342.1 [IQR: 1 80.4–607.5] mm3, \np=0.50) did not differ significantly between PCD-CT  and EID-CT. Total, \ncalcified, and fibrotic plaque volumes demonstrated  excellent agreement \nbetween repeated measurements and between readers f or both PCD-CT and \nEID-CT (all intraclass correlation coefficients >0. 90). While low-attenuation \nplaque volume had strong intra- (ICC: 0.84 [95%CI, 0.57–0.94]) and inter-\nreader (ICC: 0.92 [95%CI, 0.81–0.97]) agreements fo r PCD-CT, EID-CT \nshowed only moderate (ICC: 0.62 [95%CI, 0.11–0.86])  and poor (ICC: 0.47 \n[95%CI, 0.01–0.79]) intra- and inter-reader reprodu cibility. \nConclusion: Compared to EID-CT, PCD-CT UHR imaging reduces segm ented \ncoronary plaque volume by nearly one-third and impr oves the reproducibility of \nlow-attenuation plaque measurements. \nLimitations: Lack of invasive reference. \nFunding for this study: The study was funded by a research grant from \nSiemens Healthineers. \nEthics committee - additional information: Local Ethics Committee \napproved the present study. \nAuthor Disclosures:  \nEmese Zsarnóczay: Nothing to disclose \nPál Maurovich-Horvat: Research/Grant Support: Sieme ns Healthineers \nMilán Vecsey-Nagy: Nothing to disclose \nGiuseppe Tremamunno: Nothing to disclose \nDmitrij Kravchenko: Nothing to disclose \nBálint Szilveszter: Nothing to disclose \nTilman Stephan Emrich: Research/Grant Support: Siem ens Healthineers \nChiara Gnasso: Nothing to disclose \nAkos Varga-Szemes: Research/Grant Support: Siemens Healthineers \n \n \nReducing Variability in Coronary Plaque Characteriz ation with Photon-\nCounting CT \nA. Choux¹, S. Sharma², S. Ross², R. Thompson², Z. Y u², *A. Pourmorteza*¹; \n¹Atlanta, GA/US, ²Vernon Hills, IL/US \n(amir.pourmorteza@gmail.com) \n \nPurpose or Learning Objective: Photon-counting detector CT (PCD-CT) has \ndemonstrated significant radiation dose reduction c apabilities compared to \nenergy-integrating detector CT (EID-CT), particular ly for coronary artery \ncalcium scoring (CACS). Alternatively, the dose sav ings from PCD-CT can be \n\n \n \nSaturday \nAbstract-based Programme \n \n 229  \nleveraged to improve the reproducibility of CACS by  reconstructing images \nwith thinner slice thicknesses to reduce variabilit y from partial volume effects \n(PVE). This ex vivo study aims to evaluate and quan tify the effect of PVE on \nreproducibility of calcified plaque volume measurem ents. \nMethods or Background: Six excised human hearts with varying degrees of \ncalcification were placed inside a chest phantom an d scanned on a CdZnTe-\nbased PCD-CT (120 kVp and 250 mAs). To simulate var iability in the scan and \nreconstruction ranges, 10 image volumes were recons tructed for each heart \nusing identical parameters, with the only variation  being the starting slice \nlocation, which was incremented by 1/10 of the slic e thickness for each \nreconstruction. Images were reconstructed using FBP  (soft kernel, 0.3 mm in-\nplane pixel size) at both 3-mm (recommended for CAC S) and 1.5-mm slice \nthicknesses. Calcium volume was measured using 130 HU threshold (as per \nSCCT guidelines). Variability was quantified as the  ratio of std deviation and \nmean of calcium volume measurements for each heart and a paired t-test \n(alpha=0.05) was used for establishing statistical significance. \nResults or Findings: Variability was found to be 36.2% for conventional 3-mm \nimages, which was significantly higher than the 16. 6% variability for 1.5-mm \nimages (p<0.001). \nConclusion: Slight variations in scan or reconstruction range l eads to \nsignificant variability in coronary plaque characte rization. Taking advantage of \nthinner slices provided by PCD-CT (matched in radia tion dose to thick EID \nslices) can mitigate this variability drastically. \nLimitations: This was a small sample size ex-vivo study. \nFunding for this study: Sponsored research agreement with Canon Medical \nResearch USA, Inc. \nEthics committee - additional information: Ex vivo study, did not require \nethics committee approval \nAuthor Disclosures:  \nZhou Yu: Employee: CMRU \nSteven Ross: Employee: CMRU \nRichard Thompson: Employee: CMRU \nArnaud Choux: Research/Grant Support: Canon Medical  Research USA, GE \nHealthCare \nAmir Pourmorteza: Grant Recipient: Canon Medical Re search USA, GE \nHealthCare \nShobhit Sharma: Employee: CMRU \n \n \nQuantification of Coronary Plaque Components with P hoton-counting \nCT: Analyzing Software Consistency and Variability \n*M. Gruber*, D. Beitzke, C. Loewe, D. Beitzke; Vien na/AT \n(manuel.gruber19@gmail.com) \n \nPurpose or Learning Objective: Coronary computed tomography \nangiography (CCTA) enables non-invasive quantificat ion of plaque burden and \ncomposition. The aim of this study was to evaluate the intra-reader and inter-\nplatform reproducibility of coronary plaque volume and composition \nmeasurements from two plaque analysis software solu tions using a first-\ngeneration Photon-counting CT (PCCT) system. \nMethods or Background: Twenty plaques from thirteen patients who \nunderwent CCTA with a slice thickness of 0.4 mm wer e analyzed. Plaque \nquantification was performed using two dedicated so ftware solutions (Software \n1: Syngo.via Frontier CT Coronary Plaque Analysis, Siemens Healthineers; \nSoftware 2: QAngio CT, Medis Medical Imaging System s). Volumes of \ncoronary lumen, total plaque and plaque components were assessed at two \ntime points with each software using the following attenuation thresholds for \nplaque classification in Hounsfield units (HU) : ne crotic core: -30 to 30 HU, \nfibrotic plaque: 31 to 350 HU and calcified plaque:  >351 HU. \nResults or Findings: Intraobserver variability, determined by the Pearso n \ncorrelation coefficient showed a strong positive co rrelation (r = 0.94, p< 0.001), \nindicating a high consistency in repeated measures.  Furthermore, the paired-\nsamples t-test showed no statistically significant difference between the two \ntime points for Software 1 t(19) = 0.77, p= 0.45 or  Software 2 t(19) = -0.81, p= \n0.43). A repeated-measures ANOVA revealed no signif icant main effect for the \nsoftware type, but differences were observed for lu men volume and calcified \nplaque volume with F(11.78, 33.85) =12.81, p< 0.001  and F(1.29, 24.56) = \n13.89, p < 0.001, respectively. \nConclusion: Quantification measurements of plaque with PCCT are  feasible \nand highly reproducible, indicating that software c omparisons should take into \naccount potential differences in measurements of sp ecific components. \nLimitations: Limitations of this study are the relatively small sample size, as \nwell as the reliance on a single reader. \nFunding for this study: Beitzke Daniela as a research radiographer is \nsupported by a research grant from Siemens Healthin eers. \nEthics committee - additional information: The study is approved by the \nlocal Ethics board. All patients gave written and i nformed consent. \nAuthor Disclosures:  \nManuel Gruber: Nothing to disclose \nDietrich Beitzke: Nothing to disclose \nChristian Loewe: Nothing to disclose \nDaniela Beitzke: Research/Grant Support: Siemens He althineers \nSemiquantitative Metrics of Coronary Artery Disease  Burden: Intra-\nIndividual Comparison between Ultrahigh-Resolution Photon-Counting \nDetector CT and Energy-Integrating Detector CT \n*G. Tremamunno*¹, A. Varga-Szemes², U. J. Schoepf²,  D. Kravchenko²,  \nM. T. Hagar², A. Laghi¹, T. S. Emrich², M. Vecsey-N agy²; ¹Rome/IT, \n²Charleston, SC/US \n(giuseppe.tremamunno@uniroma1.it) \n \nPurpose or Learning Objective: To assess the impact of ultrahigh-resolution \n(UHR) photon-counting detector (PCD)-CT on the semi quantitative evaluation \nof coronary artery disease (CAD) compared to energy -integrating detector \n(EID)-CT. \nMethods or Background: Patients undergoing coronary CT angiography \n(CCTA) on an EID-CT system were prospectively enrol led for UHR PCD-CT \nscan within 30 days. Both datasets were visually ev aluated using five \nestablished semiquantitative scores: Segment Involv ement Score (SIS), \nSegment Stenosis Score (SSS), Multivessel Aggregate  Stenosis Score \n(MVAS), CCTA-adapted Leaman score (CT-LeSc), and Co ronary Artery \nDisease Reporting and Data System (CAD-RADS). Addit ionally, the total \nnumber of detected plaques and high-risk features w ere reported (positive \nremodeling, spotty calcification, low-attenuation, and napkin-ring sign). \nResults or Findings: The cohort comprised 46 patients (37 men, 68.4±6.9 \nyears). When assessing stenosis severity, PCD-CT sh owed lower SSS (3.5 \n[1.3-5.0] vs 6.5 [3.0-9.8], p<0.001), MVAS (5.5 [4. 0-7.0] vs 7.0 [5.0-9.0], \np<0.001), and CT-LeSc (10.4 [8.5-13.9] vs 11.2 [8.8 -15.4], p=0.032). \nFurthermore, 52% (24/46) of patients were reclassif ied to a lower CAD-RADS \ncategory compared to EID-CT. In terms of CAD extent , PCD-CT demonstrated \nhigher SIS (8.0 [6.0-9.0] vs 7.0 [6.0-8.8], p=0.018 ) and plaque count (9.0 [7.0-\n13.8] vs 7.0 [7.0-9.8] p<0.001). Positive remodelin g was less frequent in PCD-\nCT datasets (2.0 [1.0-4.3] vs 1.0 [0.0-3.0], p=0.01 2), with no significant \ndifferences in other high-risk features. \nConclusion: The use of UHR PCD-CT results in the detection of l ess severe, \nbut more extensive CAD in the same patient compared  to EID-CT. The effect \nof such CCTA-based differences on individual risk s tratification needs further \ninvestigation. \nLimitations: Only a limited number of patients were enrolled and  the \ncharacterization and extent of plaques were not val idated against invasive \nreference. \nFunding for this study: This study received funding by Siemens Healthineers  \nEthics committee - additional information: Name of Institution: Medical \nUniversity of South Carolina \nCode/Number: Pro00108359 \nDate of Approval: 4/13/2021 \nAuthor Disclosures:  \nMuhammad Taha Hagar: Nothing to disclose \nMilán Vecsey-Nagy: Nothing to disclose \nUwe Joseph Schoepf: Research/Grant Support: Siemens  Healthineers \nGiuseppe Tremamunno: Nothing to disclose \nDmitrij Kravchenko: Nothing to disclose \nAndrea Laghi: Nothing to disclose  \nTilman Stephan Emrich: Research/Grant Support: Siem ens Healthineers \nAkos Varga-Szemes: Research/Grant Support: Siemens Healthineers \n \n \nCost-effectiveness of ultrahigh-resolution photon-c ounting detector \ncoronary CT angiography for the evaluation of stabl e chest pain \n*M. Vecsey-Nagy*¹, T. S. Emrich¹, G. Tremamunno¹, D . Kravchenko¹,  \nM. T. Hagar¹, B. Szilveszter², P. Maurovich-Horvat² , A. Varga-Szemes¹,  \nJ. A. Decker³; ¹Charleston, SC/US, ²Budapest/HU, ³A ugsburg/DE \n(vnagymilan@gmail.com) \n \nPurpose or Learning Objective: To simulate the cost-effectiveness of \nultrahigh-resolution (UHR) photon-counting detector  (PCD)-CT in stable chest \npain patients undergoing coronary CT angiography (C CTA). \nMethods or Background: A decision and simulation model was developed \nusing Monte Carlo simulations with 1,000 bootstrap resamples to estimate the \ncosts associated with PCD-CT in lieu of EID-CT for CCTA and the referral for \nsubsequent testing. The model was constructed using  the diagnostic accuracy \nmetrics of 55 coronary lesions of patients who unde rwent CCTA on both CT \nsystems and subsequent invasive coronary angiograph y (ICA). Sensitivity and \nspecificity were defined for each Coronary Artery D isease Reporting and Data \nSystem category. The aggregate healthcare expenditu res were derived from \nthe hospital billing system. \nResults or Findings: Assuming a projected cohort of 15,000 patients over  the \nlifetime of the PCD-CT, its implementation resulted  in a 18.9% reduction in the \nnumber of functional follow-up tests (6330.3 ± 59.5 vs. 5135.7 ± 60.6, \np<0.001), a 6.0% reduction in performed ICAs (1,447 .7 ± 36.2 vs. 1,360.2 ± \n34.7, p<0.001), and a 9.4% decrease in major proced ure-related \ncomplications. Over a 10-year expected life expecta ncy, PCD-CT led to an \naverage cost saving of $794.50 ± 18.50 per patient and an overall cost \ndifference of $11,917,500 ± 4,350,169. \n\n \n \nSaturday \nAbstract-based Programme \n \n 230  \nConclusion: PCD-CT has the potential to reduce the financial bu rden on \nhealthcare systems and procedure-related complicati ons for stable chest pain \npatients with coronary calcification when compared to EID-CT. \nLimitations: Limited number of lesions included. \nFunding for this study: The study was funded by a research grant from \nSiemens Healthineers. \nEthics committee - additional information: Local Ethics Committee \napproved the present study. \nAuthor Disclosures:  \nMuhammad Taha Hagar: Nothing to disclose \nPál Maurovich-Horvat: Research/Grant Support: Sieme ns Healthineers \nMilán Vecsey-Nagy: Nothing to disclose \nGiuseppe Tremamunno: Nothing to disclose \nDmitrij Kravchenko: Nothing to disclose \nBálint Szilveszter: Nothing to disclose \nTilman Stephan Emrich: Research/Grant Support: Siem ens Healthineers \nJosua A. Decker: Nothing to disclose \nAkos Varga-Szemes: Research/Grant Support: Siemens Healthineers \n \n \nIntra-individual differences in pericoronary fat at tenuation index \nmeasurements between photon-counting and energy-int egrating detector \ncomputed tomography \nG. Tremamunno¹, M. Vecsey-Nagy¹, M. T. Hagar¹, U. J . Schoepf¹,  \nJ. O'Doherty¹, J. A. Luetkens², A. Varga-Szemes¹, T . S. Emrich¹,  \n*D. Kravchenko*¹; ¹Charleston, SC/US, ²Bonn/DE \n \nPurpose or Learning Objective: Pericoronary adipose tissue (PCAT) fat \nattenuation index (FAI) predicts major adverse card iac events, but is known to \nbe influenced by multiple factors, such as kernel s harpness, slice thickness, \nand tube potential. The objective of this study was  to explore intra-individual \ndifferences in PCAT FAI between PCD- and energy-int egrating detector (EID)-\nCT. \nMethods or Background: Patients were prospectively enrolled for a PCD-CT \nresearch scan after EID coronary CT angiography. Re constructions were \nperformed using a Qr36 kernel at 0.6 mm slice thick ness (EID and PCD-down-\nsampled [DS]) and at 0.2 mm ultra-high resolution ( UHR) for the PCD-CT. Data \nwas processed either with no use of iterative recon struction, using a weighted \nfilter back projection, or set to a strength level of Advanced Modeled Iterative \nReconstruction 3 for the EID-CT and Quantum Iterati ve Reconstruction 4 for \nthe PCD-CT. PCAT FAI of the right coronary artery ( RCA), left anterior \ndescending artery, and left circumflex artery (LCX)  was measured \nautomatically using established thresholds of -190 to -30 HU at a set distance \nand radius. Statistical testing was performed using  repeated-measures ANOVA \nand Bonferroni’s multiple comparison tests (p<.003) . \nResults or Findings: 40 patients (mean age 68±8 years, 32 males [80%]) \nwere analyzed. Absolute FAI measurements differed s ignificantly for all vessels \nbetween all reconstructions in the ANOVA comparison  (all p<.001). The mean \nFAI when using iterative reconstruction did not dem onstrate significant \ndifferences on multiple comparisons (e.g. LCX: EID:  -65.7±8.5; PCD-DS: -\n66.0±7.4; PCD-UHR: -67.8±7.0 HU, respectively; all p >.05). \nConclusion: Intra-individual absolute PCAT FAI measurements dif fer \nsignificantly between EID- and PCD-CT when controll ing for reconstruction \nkernel and slice thickness. However, the use of ite rative reconstruction \nminimizes most differences in FAI, enabling inter-s canner comparability. \nLimitations: Small population size and not all patients underwen t both \nexaminations at the same tube potential. \nFunding for this study: In part supported by a research grant from Siemens \nHealthineers. \nEthics committee - additional information: This HIPAA compliant single-\ncenter study received approval from the Institution al Review Board of the \nMedical University of South Carolina. \nAuthor Disclosures:  \nJulian Alexander Luetkens: Nothing to disclose \nMuhammad Taha Hagar: Nothing to disclose \nMilán Vecsey-Nagy: Nothing to disclose \nJim O'Doherty: Employee: Siemens \nUwe Joseph Schoepf: Consultant: Keya Medical Consul tant: HeartFlow \nConsultant: Bracco Consultant: Bayer Grant Recipien t: Siemens Consultant: \nElucid Consultant: Guerbert \nGiuseppe Tremamunno: Nothing to disclose \nDmitrij Kravchenko: Speaker: Philips \nTilman Stephan Emrich: Speaker: Siemens \nAkos Varga-Szemes: Consultant: Elucid Grant Recipie nt: Siemens \n \n \n \n \n \n \n \nCorrelation between hemodynamically significant ste noses and spectral \nphoton-counting CT first-pass myocardial perfusion imaging compared \nwith dual-energy CT in very-high risk patients \n*G. Fahrni*¹, S. A. Si-Mohamed², R. Wiemker³, D. C.  Rotzinger¹, A. Houmeau², \nC. Prieur², P. C. Douek², S. Boccalini²; ¹Lausanne/ CH, ²Lyon/FR, ³Hamburg/DE \n \nPurpose or Learning Objective: To assess the capabilities of first-pass \nmyocardial perfusion imaging (MPI) with Spectral Ph oton Counting CT \n(SPCCT) Coronary Angiography to detect fractional-f low reserve (FFR) positive \ncoronary artery stenoses, compared with dual-energy  CT (DECT) in a very-\nhigh cardiovascular risk population. \nMethods or Background: 18 very-high cardiovascular risk patients referred \nfor CCTA after diagnostic invasive coronary angiogr aphy (ICA) were \nprospectively included (15 men, 3 women). ECG-gated  CCTA was performed \nwith DECT (IQon CT and CT7500, Philips Healthcare) and SPCCT (Philips \nclinical SPCCT prototype, Philips Healthcare), with in 3 days. First-pass \nperfusion images were reconstructed with high-resol ution (DECT) and ultra-\nhigh-resolution (SPCCT) parameters. Myocardial terr itories were visually \nlabelled as normal or showing hypoperfusion. These labels were then \ncompared to hemodynamic positivity on diagnostic IC A, indicated by either by \nsignificant stenosis (>70%) or FFR-positive stenosi s. Myocardial attenuation \nwas measured in normal and hypoperfusion territorie s. \nResults or Findings: A total of 54 coronary artery territories were incl uded. At \nICA, significant stenosis was found in 15 (26%) art eries (8 LAD, 3 CX, 4 RCA). \n11 lesions were significant stenoses, 4 were FFR-po sitive. Seventeen (31%) \ncoronary artery territories were labelled as hypope rfusion for DECT and 18 \n(33%) for SPCCT. There was a significant difference  between normal and \nhypoperfusion segments both for DECT (mean: 97±43 H U versus 80±35 HU, \np<0.001) and SPCCT (mean: 92±86 HU versus 76±74 HU, p<0.001). A \nsensitivity for hemodynamically significant stenose s detection of 73% and \nspecificity of 69% were found for DECT, versus 60% ad 72% for SPCCT. \nConclusion: In very-high cardiovascular risk population, SPCCT first-pass \nmyocardial perfusion imaging was comparable to DECT  in detecting \nmyocardial hypoperfusion in significant or FFR-posi tive coronary artery \nstenosis territories. \nLimitations: Small cohort of patients and predominance of LAD le sions. \nFunding for this study: This work was supported by the European Union \nHorizon 2020 grant No. 643694. G.F. is supported by  a research grand from \nthe Swiss Society of Radiology (SSR, Luzern, Switze rland) and Lausanne \nUniversity hospital (CHUV, Lausanne, Switzerland). \nEthics committee - additional information: Hospices Civils de Lyon, \napproval number: 2019-A02945–52, SPEQUA study \nAuthor Disclosures:  \nSalim Aymeric Si-Mohamed: Nothing to disclose \nCyril Prieur: Nothing to disclose \nAngele Houmeau: Nothing to disclose \nGuillaume Fahrni: Nothing to disclose \nDavid Christian Rotzinger: Nothing to disclose \nRafael Wiemker: Employee: Philips Innovative Techno logies, 22335 Hamburg, \nGermany \nPhilippe Charles Douek: Nothing to disclose \nSara Boccalini: Nothing to disclose \n \n \nMyocardial Extracellular Volume Using Photon-Counti ng Detector CT \nwith Synthetic Hematocrit Derived from Virtual Non- Contrast Images \n*V. Mergen*, N. Ehrbar, L. J. Moser, R. Manka, H. A lkadhi, M. Eberhard; \nZurich/CH \n(victor.mergen@usz.ch) \n \nPurpose or Learning Objective: To evaluate the accuracy of myocardial \nextracellular volume (ECV) calculation using a synt hetic hematocrit derived \nfrom virtual non-iodine images (VNI) and virtual no n-contrast images (VNC) \nwith photon-counting detector (PCD) CT. \nMethods or Background: In this retrospective study, 125 consecutive patien ts \nexamined by a PCD coronary CT angiography (CCTA) an d a cardiac late \nenhancement (LE) scan, and having a recent blood he matocrit were included. \nIn the derivation cohort (first 75 patients), CCTA and LE scans were \nreconstructed as VNI at 60, 70, and 80keV and as VN C with quantum iterative \nreconstruction (QIR) strengths 2, 3, and 4. Mean bl ood pool attenuation \n(BPmean) was correlated to the blood hematocrit. In  the validation cohort (next \n50 patients), BPmean served to calculate a syntheti c hematocrit. Myocardial \nECV was computed using the synthetic hematocrit and  compared with the ECV \nusing the blood hematocrit as reference. \n \n \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 231  \nResults or Findings: In the derivation cohort (49 men, mean age 79±8 yea rs, \nmean BMI 26±5 kg/m2), correlation between BPmean an d blood hematocrit \nranged from poor for VNI of CCTA at 80keV, QIR2 (R2 =0.12) to moderate for \nVNI of LE at 60keV, QIR4; 70keV, QIR3 and 4; and VN C of LE, QIR3 and 4 \n(all, R2=0.58). In the validation cohort (29 men, a ge 75±14 years, mean BMI \n26±5 kg/m2), BPmean from VNC of the LE scan with QIR3 served to calculate \nthe synthetic hematocrit. Median ECV was 26.9% (int erquartile range (IQR), \n25.5%,28.8%) using the blood hematocrit and 26.8% ( IQR, 25.4%,29.7%) \nusing the synthetic hematocrit (VNC, QIR3; mean dif ference,-0.2%; limits of \nagreement,-2.4%,2.0%; p=0.33). \nConclusion: Myocardial ECV calculation with PCD-CT, using a syn thetic \nhematocrit derived from VNC images, enables accurat e ECV measurements. \nLimitations: Limited number of patients and lack of comparison w ith the \nclinical reference standard cardiac MRI \nFunding for this study: None \nEthics committee - additional information: Kantonale Ethikkommission \nZürich \nAuthor Disclosures:  \nVictor Mergen: Research/Grant Support: institutiona l grants from Bayer, \nCanon, Guerbet, and Siemens \nMatthias Eberhard: Research/Grant Support: institut ional grants from Bayer, \nCanon, Guerbet, and Siemens \nNicolas Ehrbar: Research/Grant Support: institution al grants from Bayer, \nCanon, Guerbet, and Siemens \nLukas Jakob Moser: Research/Grant Support: institut ional grants from Bayer, \nCanon, Guerbet, and Siemens \nRobert Manka: Research/Grant Support: institutional  grants from Bayer, \nCanon, Guerbet, and Siemens \nHatem Alkadhi: Research/Grant Support: institutiona l grants from Bayer, \nCanon, Guerbet, and Siemens \n \n \nQuantification of extracellular volume (ECV) with P hoton Counting CT to \nidentify transthyretin-related cardiac amyloidosis \nA. Clemente, *A. Marcucci*, C. De Gori, M. Occhipin ti, M. Muca, F. Pignatelli, \nD. Cioni, E. Neri; Pisa/IT \n \nPurpose or Learning Objective: Cardiac amyloid deposition causes \ninterstitial expansion, thereby increasing myocardi al extracellular volume \n(ECV). The need for quantification of amyloid burde n in transthyretin-related \ncardiac amyloidosis (TTR-CA) is currently met in pa rt through semi-quantitative \nbone scintigraphy or with measurement of ECV throug h Cardiovascular \nMagnetic Resonance(CMR). Although Photon-counting C T (PCCT) provides \ncomprehensive spectral data with every acquisition of the heart, to date few \nstudies analysed ECV in cardiac amyloidosis by usin g it. We evaluated the \naccuracy of extracellular volume (ECV) quantificati on with PCCT in TTR-CA. \nMethods or Background: We prospectively enrolled 15 patients referred to \nour centre for suspected cardiac amyloidosis and at ypical symptoms who \nunderwent a complete diagnostic work-out including bone scintigraphy and \nPCCT. Iodine maps were created by using multienergy  late scan. Volumetric \nROIs of at least 2 cm3 were manually positioned usi ng a 17 segments model \nof the left ventricle and global ECV was calculated  as average of all segments. \nThen, the study population was divided into TTR-CA cases and non-amyloid \nheart disease cases (NCA). \nResults or Findings: One patient was excluded due to light chain (AL) \namyloidosis. Among the 14 patients included (77 yea rs±5;10 men) 7 were \ndiagnosed with TTR-CA. All TTR-CA cases showed posi tive scintigraphy \n(Perugini score ≥2), except for one, where the diagnosis was made th rough fat \ntissue biopsy and consistent cardiac CMR findings. Global ECV was \nsignificantly higher in patients with TTR-CA (40.50 ±8.97%) than in NCA \npatients (27.00±3.03%;P<0.01). The accuracy of myoc ardial global ECV to \nidentify occult TTR-CA was high (AUC=0.87; 95%CI, 0 .65-1.00). \nConclusion: Preliminary results on myocardial tissue characteri zation based \non ECV quantification with PCCT iodine maps show th is is a promising method \nto detect TTR-CA. \nLimitations: Preliminary results with small sample. \nFunding for this study: No funding received for this study. \nEthics committee - additional information: No \nAuthor Disclosures:  \nFrancesca Pignatelli: Nothing to disclose \nMariaelena Occhipinti: Nothing to disclose \nEmanuele Neri: Nothing to disclose  \nMatilda Muca: Nothing to disclose \nAlessandro Marcucci: Nothing to disclose \nAlberto Clemente: Investigator: study MYOAMY-CT \nCarmelo De Gori: Nothing to disclose \nDania Cioni: Nothing to disclose \n \n \n \n \nAccuracy of Iodine Maps from Photon-Counting Detect or CT for \nDetecting Myocardial Late Enhancement – A Compariso n to LGE-MRI \nG. Tremamunno¹, A. Varga-Szemes², D. Kravchenko³, A . Laghi¹, F. Bamberg⁴, \nM. Vecsey-Nagy⁵, T. S. Emrich⁶, *M. T. Hagar*⁴; ¹Rome/IT,  \n²Charleston, SC/US, ³Bonn/DE, ⁴Freiburg im Breisgau/DE, ⁵Budapest/HU, \n⁶Mainz/DE \n(taha.hagar@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: To assess the feasibility and determine the \ndiagnostic accuracy of iodine maps from photon-coun ting detector (PCD) CT in \ndetecting and characterizing myocardial late enhanc ement (LE), compared to \nlate gadolinium enhancement (LGE) MRI. \nMethods or Background: This IRB-approved retrospective analysis of a \nprospective study cohort included subjects who unde rwent cardiac MRI \nfollowed by late iodine enhancement (CT-LE) using a  PCD-CT system \n(NAEOTOM Alpha, Siemens Healthineers). CT-LE scans were performed 5 \nminutes after administering 100 mL of contrast medi a (Ultravist, 370 mg I/mL; \nBayer Healthcare) with an ECG-triggered sequential protocol, and full spectral \ncapabilities. Iodine maps were reconstructed at a s ection thickness of 1.0 mm, \nusing a quantitative kernel (Qr40), and iterative r econstruction (QIR level 3). \nTwo blinded and independent radiologists interprete d all images. Diagnostic \naccuracy was evaluated per-patient and per-segment using LGE-MRI as the \nstandard of reference. Inter-reader agreement was a ssessed with Cohen’s \nkappa. \nResults or Findings: The final cohort comprised 27 patients (52% female,  \nmean age 52.9 ± 17.2 years). Twelve patients (44%) had positive LGE on MRI \n(3 ischemic, 9 non-ischemic). Per-patient sensitivi ty was 100% and 91.7%, with \nspecificities of 73.3% and 80.0%, and accuracies of  85.2%, respectively for \nboth readers. Per-segment sensitivity was 74.7% and  66.7%, with specificities \nof 94.9% and 96.4%, and accuracies of 91.1% and 90. 7%. Cohen’s kappa was \n0.70 (patient level) and 0.63 (segment level), resp ectively. \nConclusion: Iodine maps from PCD-CT provide high diagnostic acc uracy for \nmyocardial LE detection with substantial inter-read er agreement. PCD-CT may \nserve as an alternative to LGE-MRI in selective cas es, such as a \ncontraindication to MRI or anxiety. \nLimitations: Our study's limited sample size mandates confirmato ry research. \nOur study design introduces a selection bias, so fu rther studies are needed to \nevaluate the diagnostic role across diverse patient  populations. \nFunding for this study: This study was partially supported by an unrestrict ed \nresearch grant from Siemens Healthineers. \nEthics committee - additional information: The study protocol, compliant \nwith the Health Insurance Portability and Accountab ility Act, received approval \nfrom the local institutional review board at the re spective academic medical \ncenter, and all subjects provided written informed consent. \nAuthor Disclosures:  \nMuhammad Taha Hagar: Speaker: Siemens Healthineers \nMilán Vecsey-Nagy: Nothing to disclose \nGiuseppe Tremamunno: Nothing to disclose \nFabian Bamberg: Nothing to disclose \nDmitrij Kravchenko: Nothing to disclose \nAndrea Laghi: Nothing to disclose \nTilman Stephan Emrich: Nothing to disclose \nAkos Varga-Szemes: Research/Grant Support: Siemens Healthineers \n \n \nUltra-low dose (0.4 mSv) coronary computed tomograp hy angiography \nusing photon-counting detector computed tomography \n*S. Araki*, S. Nakamura, M. Takafuji, Y. Ichikawa, H. Sakuma, K. Kitagawa; \nTsu/JP \n(s-araki@med.mie-u.ac.jp) \n \nPurpose or Learning Objective: Photon-counting detector computed \ntomography (PCD-CT), which allows for the exclusion  of electronic noise, \nshows promise for significant dose reduction in cor onary computed \ntomography angiography (CCTA). This study aimed to assess the radiation \ndose and image quality of CCTA using PCD-CT, combin ed with \nelectrocardiogram (ECG)-triggered prospective high- pitch helical scanning and \nan ultra-low tube potential of 70 kVp and to invest igate the effect of a sharp \nkernel on image quality and stenosis assessment in such an ultra-low dose \nCCTA setting. \nMethods or Background: Forty patients (65% male) with stable heart rates \nand no prior coronary interventions were included. Data on CT dose index \nvolume (CTDIvol) and dose-length product (DLP) were  collected, with effective \nradiation dose estimated using a conversion factor of 0.014. Images were \nreconstructed using kernels of Bv64 and Bv40 for im age quality and stenosis \nassessment. \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 232  \nResults or Findings: The mean CTDIvol, DLP, and effective dose of CCTA \nwere 1.72±0.38 mGy, 29.1±6.8 mGy·cm, and 0.41±0.09 mSv, respectively. \nImage quality was similar (p=0.75) between the two kernels, with over 95% of \nsegments achieving a rating of good image quality f or both kernels. \nAdditionally, 42% of non-calcified plaques showed a n increased stenosis \nseverity from Bv40 to Bv64 (p<0.001), while 25% of calcified plaques exhibited \na decreased severity (p<0.001). \nConclusion: PCD-CT technology with ECG-triggered prospective hi gh-pitch \nhelical scanning and the tube potential of 70kVp ca n provide CCTA with ultra-\nlow radiation exposure (0.4 mSv), offering a safer and more effective method \nfor diagnosing coronary artery disease. The noise r eduction capability of PCD-\nCT allows the use of a sharp kernel even in this lo w-dose CCTA setting without \ncompromising image quality, potentially improving t he evaluation of coronary \nartery stenosis. \nLimitations: There is no reference standard such as coronary ang iography. \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: Clinical Research Ethics Review \nCommittee of Mie University Hospital (approval No. H2019-207) \nAuthor Disclosures:  \nKakuya Kitagawa: Nothing to disclose \nHajime Sakuma: Nothing to disclose \nSuguru Araki: Nothing to disclose \nYasutaka Ichikawa: Nothing to disclose \nMasafumi Takafuji: Nothing to disclose \nSatoshi Nakamura: Nothing to disclose \n \n \nBenefits of photon counting CT for the assessment o f native heart valves \nC. Mayard, S. A. Si-Mohamed, A. Houmeau, L. Boussel , P. C. Douek,  \n*S. Boccalini*; Lyon/FR \n(sara.boccalini@yahoo.com) \n \nPurpose or Learning Objective: To assess the benefits of photon counting \nCT (PCCT) on image quality of cardiac valves as com pared to conventional CT \n(conv-CT). \nMethods or Background: Patients were prospectively included to undergo a \nclinically indicated coronary CT angiography with P CCT and conv-CT within \nthree days. All the components of each valve were s ubjectively scored by two \nobservers with a 4-point scale for sharpness and co nspicuity. The number of \nnodules and of mitral chordae was noted. The number  and the localisation of \nthe calcifications relative to the thickness of the  aortic leaflets were assessed. \nFurthermore, the full width at mid weight (FWMH) of  the attenuation profile of a \nline perpendicular to the commissure between the le ft and right coronary cusps \nof the aortic valve was calculated. \nResults or Findings: Thirty-three patients were included (88% men; 62±13  \nyears). 33 pairs of aortic and mitral valves and 18  pairs of pulmonary valves \ncould be analysed. Conspicuity of aortic, mitral, a nd pulmonary valvular \nstructures was increased with PCCT except for one c ommissure of the aortic \nand pulmonary valves (p=0.06 and p=0.07). Sharpness  was superior for all \nstructures of the aortic and mitral valves, and for  2/3 edges of the pulmonary \nvalve. More fine structures (nodules, chordae) and calcifications of the aortic \nand mitral valves were visible with PCCT. The preci se localisation of the \ncalcifications could be assessed with PCCT in most cases while it remained \ndoubtful in many cases with conv-CT (p=0.02). FWMH was lower with PCCT \n(1.7(IQ=1.1) vs 2.5mm (IQ=1.3); p<0.01). The radiat ion dose was lower with \nPCCT (567.8 ±67.7 vs 681.8 ±159.6 mGy*cm; p<0.01). \nConclusion: PCCT yielded better objective and subjective image quality of \ncardiac valves as compared to conv-CT and more comp onents of the valve \nstructures were visible. \nLimitations: Small cohort \nFunding for this study: European grant H2020 \nEthics committee - additional information: Approved \nAuthor Disclosures:  \nSalim Aymeric Si-Mohamed: Speaker: Philips \nAngele Houmeau: Nothing to disclose \nPhilippe Charles Douek: Speaker: Philips \nLoïc Boussel: Speaker: Philips \nSara Boccalini: Speaker: Philips \nCharles Mayard: Nothing to disclose \n \n \n \n \n \n \n \n14:00-15:30 Research Stage 2 \nResearch Presentation Session: \nMusculoskeletal \nRPS 2010 \nImaging of injuries and instabilities of the \nperipheral joints \n \nModerator \nA. J. Shah; Ahmedabad/IN  \n(drankur203@gmail.com) \n \n \nMRI findings of shoulder injury related to vaccine administration (SIRVA) \nfollowing COVID-19 vaccination: A cross-sectional s tudy \n*N. M. I. Obeidat*¹, R. Khasawneh¹, S. Bani Essa¹, M. Alkhatatba¹,  \nA. Abdel Kareem¹, M. Al-Na'Asan¹, Y. Alshgerat¹, M.  Aljarrah¹, L. Sawalha²; \n¹Irbid/JO, ²Amman/JO \n(nmobeidat8@just.edu.jo) \n \nPurpose or Learning Objective: To investigate the MRI findings of patients \npresenting with SIRVA after COVID-19 vaccination an d to assess the \nassociations between these findings and patient dem ographics, clinical \nsymptoms, and vaccine-related factors. \nMethods or Background: A retrospective cross-sectional study involved \npatients who reported shoulder disorders following COVID-19 vaccination \nbetween 1 May 2021 and 1 May 2022. Data collected i ncluded demographics, \nclinical symptoms, vaccination details, and MRI fin dings. Statistical analyses \nassessed associations between MRI findings and pati ent demographics, \nclinical symptoms, and vaccine-related factors. Chi -square tests and t-tests \nwere utilized, with statistical significance set at  p < 0.05 and trends noted at \np < 0.1. \nResults or Findings: 35 patients were diagnosed with SIRVA (3 had bilate ral \ninvolvement; mean age 53.6 ± 9.0 years; 54.3% females). The majority \ndeveloped symptoms within 24 hours post-vaccination  (88%), most of them \nresolved within a week (84%). Pain was the predomin ant symptom (51.4%). \nMRI findings were subacromial bursitis (89.5%), ent hesial erosions (63.2%), \ntendinopathy (47.4%), rotator cuff tears and change s of adhesive capsulitis \n(each 36.8%), effusion (23.7%), as well as muscle a nd bone marrow edema \n(10.5% and 7.9%, respectively). Statistically signi ficant associations were \nfound between tendinopathy and both, older age (p=0 .024) and AstraZeneca \nvaccine (p=0.033), subacromial bursitis with both f emale gender (p=0.013) and \nhigher BMI (p=0.023), and between changes of adhesi ve capsulitis and \nSinopharm vaccine (p=0.029). Non-diabetics, females , and patients with \ntendinopathy, were more likely to have persistent s ymptoms after 3 years \nfollowing vaccination. \nConclusion: SIRVA following COVID-19 vaccination results in dif ferent \nshoulder pathologies apparent on MRI, many of which  are strongly related to \npatient demographics and type of vaccine administer ed. Awareness of SIRVA \nchanges among radiologists is paramount, especially  in seasons of vaccination \ncampaigns such as in early winter (flu-vaccines). \nLimitations: Retrospective study and small patient sample. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Retrospective study. \nAuthor Disclosures:  \nLeen Sawalha: Nothing to disclose \nMohammad Al-Na'Asan: Nothing to disclose \nMohammad Alkhatatba: Nothing to disclose \nSuhaib Bani Essa: Nothing to disclose \nNaser Mohammad Issa Obeidat: Nothing to disclose \nMajed Aljarrah: Nothing to disclose \nRuba Khasawneh: Nothing to disclose \nAli Abdel Kareem: Nothing to disclose \nYahya Alshgerat: Nothing to disclose \n \n \nLow-dose four-dimensional ct in diagnosing wrist in stability \n*I. Blom*¹, N. Mathijssen², G. Kraan²; ¹Delft/NL, ² Zoetermeer/NL \n(i.blom@rdgg.nl) \n \nPurpose or Learning Objective: To assess the ability to lower the radiation \ndose in four-dimensional computed tomography (4DCT)  for assessment of \nscapholunate (SL) instability without compromising diagnostic quality. \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 233  \nMethods or Background: Scapholunate (SL) instability can be diagnosed \nusing 4DCT. However, CT comes with increased radiat ion exposure compared \nto other diagnostic imaging tools. Therefore, six r andomly chosen cadaveric \nhuman hand specimens were scanned using an automati c device simulating \nradial-ulnar deviation. Parameters affecting radiat ion dose—scanning time, \nscan range, tube current (mAs), and tube voltage (k Vp)—were varied and \ncompared to clinical settings. Effective dose (ED) and image noise were \nassessed for all performed scans. Image noise was m easured in soft tissue \nand cortical bone, since cortical bone was used for  segmentation. Three \nmedical specialists analyzed image quality and diag nostic value using a 5-point \nLikert scale. \nResults or Findings: The ED was 0.081 mSv under standard clinical settin gs. \nReducing the scan range to 80 mm (including all car pal bones) decreased the \nED to 0.038 mSv. Further reduction to 0.004 mSv was  achieved by shortening \nthe scanning time and lowering the tube current. Lo wering tube current \nincreased image noise in soft tissue, but reduced n oise in cortical bone. \nAdjusting scanning time and scan range did not affe ct image noise. Image \nquality was deemed diagnostically acceptable for lo w dose 4DCT by all \nobservers, even 22 percent of the scans were deemed  as excellent image \nquality. \nConclusion: A low-dose 4DCT protocol for SL instability seems f easible \nwithout compromising diagnostic image quality compa red to other imaging \ntools. Further research is needed to explore low-do se 4DCT for other \nindications. \nLimitations: The limitations of the study are a human cadaveric study and a \nquestionaire based on 1 question. \nFunding for this study: No fundings were received for this study. \nEthics committee - additional information: None. \nAuthor Disclosures:  \nNina Mathijssen: Nothing to disclose \nGerald Kraan: Nothing to disclose \nIan Blom: Nothing to disclose \n \n \nEnhancing hip replacement assessment: Integrating i terative metal \nartefact reduction (iMAR) algorithm with cinematic volume rendering \ntechnique (cVRT) in photon-counting CT \n*X. Liu*; Zhengzhou, Henan Province, China/CN \n(278459366@qq.com) \n \nPurpose or Learning Objective: To explore the potential of combining the \niterative metal artifact reduction (iMAR) algorithm  with cinematic volume \nrendering technique (cVRT) in photon-counting CT fo r assessing hip \nreplacements. \nMethods or Background: A retrospective study was conducted on 120 \npatients who underwent hip arthroplasty exams using  the photon-counting CT \nscanner (NAEOTOM Alpha). Reconstruction of CT image s employing \nconventional methods, volume rendering (VR), and cV RT, both with and \nwithout iMAR. Measurements of CT numbers and standa rd deviations (SDs) in \nregions of interest (ROIs) were obtained. Objective  image quality and \nsubjective scores were assessed using established s cales. Statistical analyses \nincluded paired T tests, Mann-Whitney U tests, and Kappa tests. \nResults or Findings: Compared with the non-iMAR group, the iMAR group \nshowed significantly decreased and increased CT num bers in hyperattenuating \nand hypoattenuating areas, respectively, as well as  lowered artifact and image \nnoise (p<.001). Qualitatively, the iMAR group showe d superiority to the non-\niMAR group in both image quality and diagnostic con fidence, with scores \nincreases of 2.70 and 2.88 points, respectively (p< 0.05). iMAR combined with \ncVRT received the highest subjective score (p<0.05)  among the four series of \npost-processing images, followed by iMAR with VR im ages(P<0.05), cVRT and \nVR images in the non-iMAR group both received the l owest scores. \nConclusion: The iMAR algorithm in photon-counting CT effectivel y reduces \nartifacts and image noise, enhancing both image qua lity and diagnostic \nconfidence in post-hip metal replacement assessment s. When combined with \ncVRT, it provides a more intuitive visualization of  metal implant stability and the \nrelationship between implants and adjacent tissues.  \nLimitations: Not applicable. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number 2021-KS-HNSR115 \nAuthor Disclosures:  \nXing Liu: Nothing to disclose \n \n \nComplications of hip prostheses \n*S. L. Chung*¹, M. S. Sait²; ¹Oxford/UK, ²Kings Uni versity Hospital/UK \n(auntminnie23@gmail.com) \n \nPurpose or Learning Objective: To familiarise the different types of hip \nprostheses. To be able to interpret normal imaging findings post surgery in \nmainstay modalities such as radiograph, ultrasound,  CT and MRI. To \nunderstand the limitations of each imaging modaliti es and when to consider \nfurther imaging such as SPECT-CT scans or PET/CT. T o be aware of the \npossible false positive findings such as positive t racer uptake in nuclear scans \nwithin 1 year post operation is still within normal  physiological limits. To \nincrease awareness the numerous types of complicati ons from most common \nto least and the time intervals it occurs. \nMethods or Background: Review of current literature of optimisation of \nimaging modalities to reduce artefact caused by hip  prostheses enabling better \nassessment of the joint. \nResults or Findings: Slice-encoding for metal artefact correction (SEMAC )-is \na relatively new MRI sequence particularly STIR and  T1-weighted SEMAC \nsequences help reduce artefacts caused by hip prost hesis best and can help \nexclude aseptic loosening. Understand the scopes of  SPECT-CT, PET/CT and \nMRI including when to use adjunct imaging to diagno se. \nConclusion: MRI is the best imaging modality for reviewing hip prosthesis but \naccessibility is dependent on institution. MR SEMAC  protocol should be \nroutinely applied to reduce artefacts and increasin g diagnostic assessment. \nLimitations: Limited articles on updated nuclear studies/protoco ls. \nFunding for this study: Nil \nEthics committee - additional information: Not Applicable \nAuthor Disclosures:  \nMohammed Saif Sait: Nothing to disclose \nSiok Li Chung: Nothing to disclose \n \n \nHow routine day-one radiographs affect patient mana gement after hip \nand knee arthroplasty or internal fixation \n*A. Jonkergouw*, P. Tukker, W. De Monye; Haarlem/NL  \n \nPurpose or Learning Objective: This study aims to evaluate the clinical \nconsequences of radiographs performed routinely on the first day after \narthroplasty or internal fixation of the hip and kn ee. \nMethods or Background: We conducted a retrospective search of our \nimaging database for day-one post-operative radiogr aphs after total hip \narthroplasty, hip hemi-arthroplasty, dynamic hip sc rew fixation, gamma nail \nfixation, total knee arthroplasty, and unicompartme ntal knee arthroplasty, \nstarting from 1 January 2023 until 500 radiographs for hip surgeries and 500 \nfor knee surgeries were included. For each case, th e radiological report was \nreviewed to determine if any immediate post-operati ve abnormalities were \ndetected. Additionally, we recorded the type of pro sthesis, the patient’s age \nand sex, and information on excessive pain from pat ient records. \nResults or Findings: Of the 500 patients who underwent hip surgery, 388 \n(77.6%) received total hip arthroplasty, 27 (5.4%) hemi-arthroplasty, 34 (6.8%) \ndynamic hip screw fixation, and 51 (10.2%) gamma na il fixation. Of the 500 \npatients who underwent knee surgery, 420 (84%) rece ived total knee \narthroplasty, and 80 (16%) unicompartmental knee ar throplasty. Across the \nentire cohort, post-operative abnormalities were no ted in 33 patients (3.3%). \nFive patients (0.5%) required additional medical ca re due to a fracture, all of \nwhom had undergone hip surgery. Three (0.3%) underw ent revision surgery, \nand 2 (0.2%) received weight-bearing restrictions. Of the 5 patients with a \nfracture, 4 (80%) reported significant pain prior t o undergoing radiological \nexamination. Gender did not influence the occurrenc e of post-operative \nabnormalities. \nConclusion: Given that only 0.5% of all patients required addit ional medical \ncare after detection of post-operative abnormalitie s, the necessity for routine \npost-operative radiographs appears limited. \nLimitations: Based on the available data, no definite predictive  factor has \nbeen identified in relation to radiological abnorma lities after hip or knee \nsurgery. \nFunding for this study: None \nEthics committee - additional information: The ACLU gave permission for \nthis study. This committee advises the Board of Dir ectors on new scientific \nstudies at Spaarne Gasthuis. \nAuthor Disclosures:  \nAnne Jonkergouw: Nothing to disclose \nWouter De Monye: Nothing to disclose \nPaul Tukker: Nothing to disclose \n \n \nIncreased Lateral Tibial Plateau Slope (LTPS) and e xtreme Intercondylar \nNotch Slope (INS): parameters to predict Anterior C ruciate Ligament \n(ACL) High-grade Injury (HgI) and Mucoid Degenerati on (MD)? \n*A. Cutaia*, R. Faletti, P. Fonio, E. La Paglia; Tu rin/IT \n(aldo.cutaia@unito.it) \n \nPurpose or Learning Objective: To assess the reliability of LTPS and INS to \npredict ACL HgI and MD. \nMethods or Background: 110 patients (Mean Age 46,1years) with MR \nevidence of pathological ACL were retrospectively s elected. Three groups: HgI \n(Mean age 31,58years), Low-grade injury (LgI) (Mean  age 45,85years) and MD \n(Mean age 59,33years). LTPS was measured on sagitta l T1-TSE images: \nFirstly selecting a slice comprehending tibial atta chment of posterior cruciate \nligament and intercondylar eminence, tracking the t ibial axis with craniocaudal \n\n \n \nSaturday \nAbstract-based Programme \n \n 234  \ncircles method. Secondly, in a slice immediately me dial to the head of the \nfibula, drawing a line perpendicular to the tibial axis. Thirdly, measuring the \nangle between this line and the one parallel to the  posterior tibial plateau. \nLTPS was reported as increased when greater than 10 °. INS was measured \ndrawing the Blumensaat Line, then evaluating in whi ch section of the tibial \nplateau it landed: anterior third type 1, middle th ird type 2, posterior third type \n3. INS was extreme when classified as group 2 or 3.  Statistical significance \nwas measured using Z and T test for quantitative va riables and ChiSquare for \nqualitative variables. \nResults or Findings: General Group (GG) mean LTPS was 6,87° +-3,09. \nPatients with an increased LTPS (Mean LTPS 10,69° + -0,73) showed \nsignificantly higher prevalence (66% vs.36%) of ACL  HgI compared to GG \n(p<0.005). HgI group’s mean LTPS was 7,83° +-3,02, significantly higher than \nLgI group (p<0.025). MD group showed significantly higher prevalence (55% \nvs.27%) of extreme INS compared to HgI (p<0.025). \nConclusion: Increased LTPS should be added in MR reports as it could \nhighlight a predisposition to ACL HgI. INS should b e assessed as it could play \na pivotal role in the onset of ACL MD on the basis of chronic impingement. \nLimitations: The study is retrospective and monocentric. \nFunding for this study: None. \nEthics committee - additional information: None. \nAuthor Disclosures:  \nAldo Cutaia: Nothing to disclose \nRiccardo Faletti: Nothing to disclose  \nErnesto La Paglia: Nothing to disclose \nPaolo Fonio: Nothing to disclose \n \n \nIs It Necessary To Add Soft Tissue Injury to the Cl assification in Tibial \nPlateau Fracture Management? \nM. Tunçez¹, I. Akan¹, F. Seyfettinoğlu², *H. Çetin Tunçez*¹, B. Dirim Mete¹,  \nC. Kazımoğlu¹; ¹Izmir/TR, ²Adana/TR \n(drhulyacetintuncez@gmail.com) \n \nPurpose or Learning Objective: Current classification systems have \ngenerally been developed based on the type of fract ure patterns. The most \ncommonly used is the Schatzker classification syste m, which includes six types \nof tibial plateau fractures. While this classificat ion evaluates the fracture in two \ndimensions, three-dimensional classifications were introduced after the spread \nof CT. To our knowledge, the classification system regarding soft tissue injuries \nbased on MRI findings has not been established yet.  Therefore, in this study, \nwe aimed to evaluate the usability of a new modifie d classification by \nevaluating soft tissue injuries in tibial plateau f ractures. \nMethods or Background: A total of 36 patients with tibial plateau fracture s \nwere included in the study. Patients’ age, gender, and affected sides were \nrecorded. Injuries to the medial meniscus, lateral meniscus, anterior cruciate \nligament, posterior cruciate ligament, medial colla teral ligament, and lateral \ncollateral ligament were examined with preoperative  magnetic resonance \nimaging. Soft tissue injuries were arranged accordi ng to the novel modified \nclassification based on the Schatzker classificatio n. \nResults or Findings: The mean age of the study participants was 45 (19-7 6) \nyears; 72% of the patients were men and 28% were wo men. Moreover, 44% \nand 56% of the patients had broken the right and le ft tibial plateaus, \nrespectively. At least one soft tissue injury was d etected in 29 (81%) patients. \nIn 14 (39%) patients, two or more soft tissue injur ies were observed. All \npatients were arranged according to the novel modif ied classification regarding \nligament and meniscus injuries. \nConclusion: With this novel modified classification system, we think that \nhaving better information about the preoperative co ndition of the soft tissue \ninjuries can change the surgical strategy in patien ts with tibial plateau \nfractures. \nLimitations: Patients with minimal displacement and those requir ing \nconservative treatment were excluded. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This retrospective study was \napproved by the Institutional Review Board of our h ospital. \nAuthor Disclosures:  \nHülya Çetin Tunçez: Author: Data collection Author:  STUDY DESIGN \nCemal Kazımoğlu: Author: literature search \nIhsan Akan: Author: data collection  \nMahmut Tunçez: Author: Study Design \nBerna Dirim Mete: Author: data collection  \nFırat Seyfettinoğlu: Author: literature search \n \n \n \n \n \n \n \n \nValidation of mRUST as an endpoint for fracture hea ling studies in lower-\nlimb fractures \nD. Deppe¹, *M. Gabriele*¹, E. K. Simşek², A. Ordas-Bayon³, M. Leskovar¹,  \nA. Trepczynski¹, S. Zachow¹, G. Duda¹, M. Heyland¹;  ¹Berlin/DE, ²Ankara/TR, \n³Madrid/ES \n(matteo.gabriele2@studio.unibo.it) \n \nPurpose or Learning Objective: Determining the best treatment for lower \nlimb fractures is challenging due to multiple facto rs affecting bone healing, \nrequiring precise monitoring for optimal care. A ra diographic assessment of the \nend point of bone healing is difficult, prompting t he use of scores such as the \nmodified Radiographic Union Score for Tibial fractu res (mRUST) that \nrepresents one of the most used scores to evaluate radiographic fracture \nhealing. However, the use of mRUST across different  locations and treatment \noptions has not yet been thoroughly demonstrated. T he objective of this study \nis to validate the robustness of mRUST as a reliabl e measure during follow-up \nin lower limb fractures using various treatment mod alities. \nMethods or Background: Six international investigators (five orthopaedic \nsurgeons and one radiologist) independently assesse d the mRUST in 166 \npatients with extra-articular lower-limb fractures for different follow-up \ntimepoints. Inter-rater reliability was assessed fo r location (femur/ tibia), \ntreatment option (plate fixation/ nail fixation) an d for different treatment options \nin different fracture locations using intraclass co rrelation coefficients (ICC). \nResults or Findings: 166 patients (63 femur fractures, 103 tibia fractur es; \n32.52% female) with a total of 1136 follow up time points were included. \nOverall inter-rater reliability for mRUST was good (0.77), regardless of fixation \nmethod (0.79, for both nail and plate fixation) or anatomical location (0.78 in \ntibia fractures, 0.81 in femur fractures). On corte x level, reliability varied for \ndifferent location within in the bone with highest inter-rater agreement for the \nmedial cortex (0.70-0.74) and lowest for the poster ior cortex (0.65-0.74) \nConclusion: The mRUST-Score proves to be a robust scoring tool for healing \nmonitoring in lower-limb fractures treated with dif ferent fixation methods in \ndifferent parts of the bone. \nLimitations: Images were presented in chronological order, which  limited the \nresults of this study. \nFunding for this study: None. \nEthics committee - additional information: Local ethics committee approval \nwas granted for this retrospective study (EA4/099/2 2). \nAuthor Disclosures:  \nGeorg Duda: Nothing to disclose \nAlejandro Ordas-Bayon: Nothing to disclose \nStefan Zachow: Nothing to disclose \nDominik Deppe: Nothing to disclose \nMark Heyland: Nothing to disclose \nMatteo Gabriele: Nothing to disclose \nMarko Leskovar: Nothing to disclose \nEkin Kaya Simşek: Nothing to disclose \nAdam Trepczynski: Nothing to disclose \n \n \nCorrelation of Osteochondral Lesions of the Talar D ome with Tears of the \nSuperior and Inferior Bands of the Anterior Talofib ular Ligament and the \nCalcaneofibular Ligament: A Retrospective Study \n*S. Rajan*, J. S. Chatha, H. Mahajan; New Delhi/IN \n(drsriramrajan@gmail.com) \n \nPurpose or Learning Objective: Osteochondral lesions of the talar dome \n(OLT) can result from recurrent ankle microinstabil ity, traumatic events, or \nanatomical abnormalities. This study aims to evalua te the correlation between \nOLT and single, double, or triple ligament tears wi thin the lateral collateral \nligament complex. \nMethods or Background: A retrospective review was conducted on 212 \nconsecutive ankle MRIs performed for pain or instab ility, sourced from the \nPACS server. Examination under anesthesia and arthr oscopic confirmation \nwere obtained in 23 cases. The data were anonymized  and analyzed by two \nradiologists with extensive experience. The status of the superior and inferior \nATFL bands, the calcaneofibular ligament (CFL), and  the superficial and deep \ndeltoid ligaments were assessed. Osteochondral lesi ons were classified using \nthe Anderson classification. Statistical significan ce was evaluated using the \nChi-square test to assess the relationship between ligament tears and OLT. \nResults or Findings: OLT was identified in 74 patients (36.6%), while 12 6 \npatients (62.4%) had no OLT. Among those with OLT, 72 had lateral collateral \nligament pathology, with only 2 cases showing no de finitive tears. Of the 74 \npatients with OLT, 26 had tears of the superior ATF L with inferior band \nscarring, 36 had tears of both ATFL bands, and 19 h ad tears of both ATFL \nbands and the CFL. Statistical analysis revealed a significant correlation (p < \n0.01) between the presence of ligament tears and th e occurrence of OLT. \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 235  \nConclusion: This study highlights a significant correlation bet ween \nosteochondral lesions of the talar dome and lateral  ligament tears, \nemphasizing the need for comprehensive assessment o f ligament integrity in \npatients with OLT. \nLimitations: Arthroscopic proof of ligament tears was not presen t in all cases \nFunding for this study: None \nEthics committee - additional information: Retrospective study \nAuthor Disclosures:  \nJagneet Singh Chatha: Nothing to disclose \nSriram Rajan: Nothing to disclose \nHarsh Mahajan: Nothing to disclose \n \n \nEvaluating peroneus tendon motion using dynamic mag netic resonance \nimaging – a pilot study \nK. Bokwa, D. G. Mocanu, N. Solidakis, *P. Szaro*; G othenburg/SE \n(pawel.szaro@vgregion.se) \n \nPurpose or Learning Objective: Previous studies indicate that peroneus \nbrevis instability and split tears may be missed in  up to half of patients clinically \ndue to unclear clinical signs and on conventional m agnetic resonance (MRI) \nbecause of its static nature. We hypothesize that d ynamic imaging may \nimprove the diagnosis of peroneus brevis instabilit y and split tears. However, \nno studies have evaluated whether dynamic MRI can v isualize peroneus \ntendon motion. The aim of this study is to assess w hether dynamic MRI can be \nused to evaluate the motion of the peroneal tendons . \nMethods or Background: Study design: observational pilot study. We \nperformed dynamic MRI using two small flexible coil s in a 3T machine, \nassessing the axial plane at the lateral malleolus with Dynamic Balanced Fast \nField Echo (BFFE). Ten participants received moveme nt training from a \nradiology nurse before imaging. Two radiologists (r aters) evaluated the image \nquality. We included only examinations with clear t endon outlines, visible \nmotion without artifacts. Raters assessed the posit ion of peroneus brevis \nrelative to peroneus longus in neutral, plantar, an d dorsal flexion, reaching a \nconsensus. Radiologists measured the distance betwe en the tendons' central \npoints in each position, reporting the mean values.  \nResults or Findings: All examinations were included in the analysis. \nPreliminary analysis revealed that dynamic MRI allo ws visualisation of \nperoneus tendons motion in good quality. The mean d istance between the \ntendons was 2.1 mm (SD 0.3 mm) in the neutral posit ion, 4.8 mm (SD 0.3 mm) \nin dorsal flexion, and 2.0 mm (SD 0.2 mm) in planta r flexion. \nConclusion: Dynamic magnetic resonance allows the evaluation of  peroneus \ntendon motion, offering a novel approach for evalua ting stability that may \nimprove the accuracy of peroneus diagnostics. \nLimitations: Small sample size, only dorsal and plantar flexion in the ankle. \nFunding for this study: The study was founded by Stiftelsen Tornspiran 934:  \n2023-12-01. \nEthics committee - additional information: The Swedish Ethical Review \nAuthority approved the study: 2023-07231-01. \nAuthor Disclosures:  \nDan Gheorghe Mocanu: Nothing to disclose \nNektarios Solidakis: Nothing to disclose \nKatarzyna Bokwa: Nothing to disclose \nPawel Szaro: Nothing to disclose \n \n \nTop five MRI findings of professional soccer player s in pre-season \nmedical examinations \nE. Höhne, I. Yel, A. Gökduman, S. Bernatz, *M. Dimi trova*, C. Booz, T. Vogl, \nS. Mahmoudi; Frankfurt/DE \n \nPurpose or Learning Objective: In professional soccer extensive \nmusculoskeletal assessments are conducted prior to player transfers to \nevaluate the current state and future risk of injur y. Magnetic resonance imaging \n(MRI) is essential in this process revealing muscul oskeletal findings even in the \nabsence of symptoms. This analysis presents the fiv e most frequent MRI \nfindings in a cohort of professional soccer players  and aims to improve \nunderstanding of the physical condition of elite at hletes. \nMethods or Background: This retrospective study included comprehensive \nmusculoskeletal 3T MRI scans obtained during medica l checks of professional \nsoccer players from August 2019 to September 2024. Clinical data were \nextracted from medical records and supplemented wit h further functional \ninformation. \nResults or Findings: MRI scans of 44 professional soccer players were \nanalysed. Among the players, five were left-footed,  one was two-footed, and \nthe remaining players were right-footed. The averag e age at the time of \nexamination was 22.8 years (± 4,3). The most common  finding was a \nsecondary cleft on the left side, observed in 18 pl ayers (40.9%). A bulging or \nprotruding disc at the L5/S1 level was found in 31. 8% of the cases. \nDegenerative changes in the labrum were identified in 27.3% of players on the \nleft side and in another 27.3% on the right side. C hondropathy of the left knee \nwas present in 27.3% of cases. \nConclusion: This retrospective analysis revealed several notabl e findings, \nparticularly given the young average age of the ath letes. The main findings \nincluded a secondary cleft and labral degeneration,  which may be associated \nwith the common occurence of groin pain in soccer p layers. A better \nunderstanding of these associations could enhance t he development of more \neffective diagnostic and preventive strategies for musculoskeletal injuries. \nLimitations: The limited sample size restricts the generalizabil ity of findings. \nFunding for this study: None. \nEthics committee - additional information: The local ethics committee has \napproved this retrospective study. \nAuthor Disclosures:  \nChristian Booz: Nothing to disclose \nIbrahim Yel: Nothing to disclose \nMirela Dimitrova: Nothing to disclose \nThomas Vogl: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nAynur Gökduman: Nothing to disclose \nSimon Bernatz: Nothing to disclose \nElena Höhne: Nothing to disclose \n \n \n14:00-15:30 Research Stage 3 \nResearch Presentation Session: Hybrid, \nMolecular and Translational Imaging \nRPS 2006 \nHybrid and molecular imaging in \noncology: clinical and translational \nstudies \n \nModerator \nP. M. Kazmierczak; Munich/DE  \n \n \nGa-68-FAPI PET/CT in Malignant Mesothelioma: Prospe ctive Single-\nCenter Observational Trial \n*L. Kessler*, B. M. Schaarschmidt, J. Siveke, L. Um utlu, M. Schuler,  \nM. Stuschke, K. Herrmann, W. Fendler, H. Hautzel; E ssen/DE \n(lukas.kessler@uk-essen.de) \n \nPurpose or Learning Objective: Mesothelioma are rare tumors mostly \naffecting the pleura and are associated with overal l poor prognosis. \nMesothelioma subtypes have shown to express fibrobl ast-activation-protein \n(FAP) in tumor cells, suggesting FAP as a promising  target for imaging and \ntherapy. Thus, novel radiolabeled FAP-inhibitors (F API) are of interest for \nfuture theranostic approaches. The FAPI-PET observa tional trial \n(NCT04571086) evaluates Ga-68-FAPI PET imaging in c ancer patients and \nhere we present data on Ga-68-FAPI in patients with  mesothelioma. \nMethods or Background: Forty-one patients underwent Ga-68 FAPI-PET \nimaging and F-18-FDG PET. The primary endpoint was correlation of Ga-68-\nFAPI-PET uptake (SUVmax and SUVpeak) with histopath ological FAP \nexpression. Secondary objectives included detection  rate and diagnostic \nperformance (sensitivity, specificity, positive/neg ative predictive values and \naccuracy) compared to F-18-FDG PET validated by his topathology or a \ncompound reference standard (histopathology, altern ative imaging or follow-up \nimaging). \nResults or Findings: SUVmax and SUVpeak values showed a significant \ncorrelation with histopathological FAP expression ( SUVmax r = 0.49, p = 0.04; \nSUVpeak r = 0.51, p = 0.03). Overall Ga-68-FAPI sho wed high diagnostic \nperformance (SE 98%, SP 81%, PPV 88% and NPV 97%). Ga-68-FAPI had \nsimiliar sensivity compared to F-18-FDG on both per -patient (100.0% vs. \n97.3%) and per-region (98.0% vs. 95.9%) basis but s howed increased \nSpecificity (81.1% vs. 36.8%) and positive predicti ve value (87.5% vs. 66.2%) \nin per-region analysis, indicating superior perform ance. This discrepancy was \nattributed to a higher number of false positive reg ions on F-18-FDG (FAPI, N = \n7 vs. FDG, N = 31). \nConclusion: This is the first study to show correlation between  Ga-68-FAPI \nuptake and histopathological FAP expression and sup erior diagnostic value \ncompared to F-18-FDG in mesothelioma patients. Thes e findings highlight the \npotential of Ga-68-FAPI as a potential tool in clin ical practice. \nLimitations: The limitations are single center, observational co hort with \nheterogenous patients. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: University Duisburg-Essen \npermits 19-8991-BO and 20-9485-BO \n\n \n \nSaturday \nAbstract-based Programme \n \n 236  \nAuthor Disclosures:  \nKen Herrmann: Consultant: Bayer, SOFIE Biosciences,  SIRTEX, Adacap, \nCurium, Endocyte, BTG, IPSEN, Siemens Healthineers,  GE Healthcare, \nAmgen, Novartis, ymabs, Aktis Oncology, Theragnosti cs, Pharma15, \nDebiopharm, AstraZeneca, Janssen. \nHubertus Hautzel: Consultant: Urenco Consultant: Ro che \nWolfgang Fendler: Consultant: SOFIE Biosciences, Ja nssen (consultant, \nspeaker), Calyx (consultant, image review), Bayer ( consultant, speaker, \nresearch funding) \nMartin Stuschke: Nothing to disclose \nBenedikt Michael Schaarschmidt: Research/Grant Supp ort: DFG \nJens Siveke: Consultant: AstraZeneca, Bayer, Boehri nger Ingelheim, Bristol-\nMyers Squibb, Immunocore, MSD, Novartis, Roche/Gene ntech and Servier \nBoard Member: Pharma15 Research/Grant Support: Abal os Therapeutics, \nBoehringer Ingelheim, Bristol-Myers Squibb, Celgene , Eisbach Bio, and \nRoche/Genentech \nLukas Kessler: Nothing to disclose \nLale Umutlu: Nothing to disclose \nMartin Schuler: Nothing to disclose \n \n \nDiagnostic accuracy and molecular characterization of endometrial \ncancer using fully hybrid [18F]FDG PET/MRI \n*T. Russo*, C. Bezzi, C. Sabini, G. Candotti, G. Ir oni, F. De Cobelli, P. Mapelli, \nA. Chiti, M. Picchio; Milan/IT \n(t.russo.1994@gmail.com) \n \nPurpose or Learning Objective: This study aims at evaluating the \neffectiveness of fully hybrid [18F]FDG PET/MRI in E C staging, assessing its \ndiagnostic accuracy and prognostic role in predicti ng features of EC \naggressiveness, including p53abn MMRd for the new m olecular classification. \nMethods or Background: This prospective study involved 80 patients with \nbiopsy-confirmed EC who underwent preoperative [18F ]FDG PET/MRI for \nstaging purposes. The PET/MRI scans were independen tly reviewed by a \nradiologist and a nuclear medicine physician, asses sing the diagnostic \naccuracy (ACC), sensitivity (SN), specificity (SP),  and positive and negative \npredictive value (PPV, NPV). Imaging and clinical p arameters were then \ninvestigated for their correlation (Spearman's rank  correlation) and analyzed \nthrough Fisher’s exact test, and ROC analysis. Kapl an-Meier survival curves, \nLog-rank tests and Cox proportional hazards models were used to evaluate the \nprognostic value of parameters for predicting tumor  relapse. \nResults or Findings: PET/MRI provided ACC=98.75%, SN=98.75%, and \nPPV=100% for primary tumor detection, and ACC = 92. 31%, SN = 84.62%, SP \n= 93.85%, PPV = 73.33%, NPV = 96.83% for LN detecti on. PET/MR \nparameters were able to predict LVSI (AUC= 80.16%),  deep MI, p53abn and \nMMRd (AUC>70%). Less accurate predictions were foun d for EC histotype \n(AUC=68.04%) and infiltration pattern (AUC=65.19%).  Finally, quantitative \nparameters could also predict both disease relapse (AUC=81.63%), with MTV \nand Size_CC showing the highest prognostic value, a nd the need to administer \npost-operative adjuvant therapy (AUC=74.63%). \nConclusion: [18F]FDG PET/MRI show good accuracy in the staging of EC \nprimary tumor and LN metastases. Moreover, PET and MRI-derived \nparameters have a potential role in the characteriz ation tumor aggressiveness \nand molecular alterations, as well as tumor recurre nce prediction, crucial \ninformation for an optimal patient treatment and ma nagement in clinical \npractice. \nLimitations: Molecular characterization not available for all pa tients. \nFunding for this study: None \nEthics committee - additional information: The study received approval \nfrom the Institution’s Ethics Committee (protocol n umber 85/INT/2019) and \ninformed consent was obtained from all patients in accordance with EC \nguidelines . All procedures were carried out in acc ordance with the Declaration \nof Helsinki (1964) and its later amendments. \nAuthor Disclosures:  \nMaria Picchio: Nothing to disclose \nGabriele Ironi: Nothing to disclose \nCarlotta Sabini: Nothing to disclose \nArturo Chiti: Nothing to disclose \nGiorgio Candotti: Nothing to disclose \nCarolina Bezzi: Nothing to disclose \nTommaso Russo: Nothing to disclose \nPaola Mapelli: Nothing to disclose \nFrancesco De Cobelli: Nothing to disclose \n \n \n \n \n \n \n \n \nImplmentation of Diffusion Weighted Imaging for who le body staging of \nlymphoma patients \n*A. Milosevic*¹, M. Chodyla¹, H. Steinberg¹, L. Kes sler¹, B. M. Schaarschmidt¹, \nL. Umutlu¹, J. Grueneisen²; ¹Essen/DE, ²Munich/DE \n(aleksandar.milosevic@uk-essen.de) \n \nPurpose or Learning Objective: To asses the feasability of Diffusion \nWeighted Imaging (DWI) in staging of lymphoma patie nts to establish a \nradiation-free alternative to FDG-PET. \nMethods or Background: A total of 181 lymphoma patients (mean age: 30.9 ± \n19.1 years. 75 female and 106 male) undergoing clin ically indicated 18F-FDG \nPET/MR examinations were retrospectivly assessed. 7 45 target lesions were \nassessed regarding Tracer-uptake (Standardized Upta ke Values, SUV), \ndiffusion restriction (Apparent Diffusion Coefficie nt, ADC), size and localization. \nEach of the target lesions was assigned a Deauville  score. SUVs and ADC \nvalues were then compared using Spearman's rank cor relation test. ROC-\nanalysis was employed in order to find appropriate thresholds to distinguish \nbetween vital (score 4-5) and non-vital (score 1-3)  manifestations in ADC-\nmeasurements, \nResults or Findings: Calculated mean values for the ADCmin and ADCmean \nof targets with a Deauville score of 4 and 5 were s ignificantly lower when \ncompared to those lesions with a score of 1-3. Acco rdingly, ADCmean \ndisplayed a strong inverse correlation with the SUV s (r = -0.83). Furthermore, \nROC analysis displayed an AUC of 0.91, 0.98 and 0,8 7 with a sensitivity of \n87%, 93%, and 80% for ADCmin, ADCmean and ADCmax, r espectively. \nConclusion: We highly recommend considering DWI an adjunct para meter for \nstaging and restaging of lymphoma patients. DWI can  be particularly helpful for \nindividuals suffering from subtypes with low avidit y to FDG and patient groups \nsusceptible to radiation. \nLimitations: Lack of proper gold standard for reference tissue i n ADC \nmeassurements. Thus, threshold levels were calculat ed manually. \nFunding for this study: No funding. \nEthics committee - additional information: committee of university of \nDuisburg-Essen \nAuthor Disclosures:  \nJohannes Grueneisen: Nothing to disclose \nBenedikt Michael Schaarschmidt: Nothing to disclose  \nLukas Kessler: Nothing to disclose \nLale Umutlu: Nothing to disclose \nAleksandar Milosevic: Nothing to disclose \nMichal Chodyla: Nothing to disclose \nHannah Steinberg: Nothing to disclose \n \n \nIn pursuit of an appropriate use criteria for the u se of 18-F FES PET CT in \nthe management of ER positive breast cancer : work in progress \n*P. S. Choudhury*, S. Chowdhury, M. Gupta, R. Kumar ; Delhi/IN \n(pschoudhary@hotmail.com) \n \nPurpose or Learning Objective: Oestrogen receptor (ER) is highly expressed \nin 70-80% of breast malignancies (BC). ER expressio n or absence plays a \ncentral role in its oncogenesis and is a prognostic  and predictive biomarker. \nMolecular imaging with 18-F Fluroestradiol (FES) PE T-CT targets ER and may \nhave higher incremental value in guiding management  by increasing \nspecificity. \nMethods or Background: We enrolled 57 female and 1 male breast cancer \npatient during initial staging and restaging as a p art of an ongoing prospective \nstudy and performed 18-F FDG and 18-F FES within 1 week. Whole body FDG \nand FES PET-CT scan done from base of skull to mid thigh. Image of the \nbreasts performed in prone position by hanging tech nique. The study was \napproved by scientific committee (Res/SCM/53/2022/6 7) and IRB \n(RGCIRC/IRB-BHR/112/2022). Lesion detection sensiti vity was compared for a \ntotal number of lesions by McNemar test. FES was ta ken as reference in \nindeterminate lesions. Incremental value was report ed by identifying FES \nexclusive lesions. Spearman rank test was used to c o-relate ER expression \nand SUV max. \nResults or Findings: FDG was more sensitive in lesion detection (80.3% v s \n61.2% p<0.001) However FES detected more lesions in  lobular variety (81.5% \nvs 56.2% p0.09). Significant co-relation seen betwe en ER+ve and FES uptake. \nSignificant incremental value of FES seen in 27% of  patients with \nindeterminate lesions characterised by FES. Overall  change in management \nnoted in 21.1% (5.2% surgical and intent of managem ent 15.8%). \nConclusion: Potential clinical applications of FES PET CT could  be to select \nappropriate patients for hormonal therapies, resolv ing ER status of lesions \nnon-invasively, solving clinical dilemmas when resu lts of other investigations \nare inconclusive, systemic staging of breast cancer s with low metabolic activity \nand selecting optimal doses for current or novel ER  targeted therapies. \nLimitations: Work in progress \nFunding for this study: Radiopharmaceuticals were procured by the \ninstitution and the equipments used belongs to the institution. No other source \nof funding was used \n\n \n \nSaturday \nAbstract-based Programme \n \n 237  \nEthics committee - additional information: Ethics Committee letter \nreference: IRB (RGCIRC/IRB-BHR/112/2022) \nAuthor Disclosures:  \nRajiv Kumar: Nothing to disclose \nManoj Gupta: Nothing to disclose \nSuchita Chowdhury: Nothing to disclose \nPartha S Choudhury: Nothing to disclose \n \n \n[1-11C]acetate PET/CT distinguishes aggressive crib riform Gleason score \n7 prostate cancer and is mechanistically informed b y spatial \nmetabolomics \n*N. Sushentsev*¹, G. Hamm¹, R. Manavaki¹, D. Solovi ev², D. Lewis², L. Aloj¹, \nR. Goodwin¹, F. A. Gallagher¹, T. Barrett¹; ¹Cambri dge/UK, ²Glasgow/UK \n(ns784@medschl.cam.ac.uk) \n \nPurpose or Learning Objective: We aimed at identifying a clinical metabolic \nimaging technique to differentiate Gleason score 7 (GS7) prostate tumours \nwith dominant cribriform and non-cribriform Gleason  pattern 4 (GP4) based on \ntheir comparative metabolic pathway enrichment anal ysis (MPEA). \nMethods or Background: 28 prostate cancer (PCa) patients with n=39 GS7 \nlesions on prostatectomy were recruited, of which n =27 and n=12 harboured \nnon-cribriform and cribriform GP4, respectively. Th e patients were divided into \nthree sub-cohorts (A, B, and C), each encompassing n=13 GS7 lesions (n=9 \nnon-cribriform; n=4 cribriform). In cohort A, n=39 fresh-frozen tumour samples \nwere used for spatial metabolomics imaging to enabl e comparative MPEA \nbetween cribriform and non-cribriform GP4 epitheliu m. In cohort B, formalin-\nfixed-paraffin-embedded samples were immunohistoche mically stained for fatty \nacid synthase (FASN) to corroborate the findings fr om cohort A. In cohort C, \nwe determined standardised uptake value (SUVbw) for  [1-11C]-acetate \nPET/CT in cribriform and non-cribriform GS7 lesions  as a marker of fatty acid \nsynthesis. \nResults or Findings: In cohort A, MPEA highlighted fatty acid biosynthes is as \nthe most significantly enriched pathway in cribrifo rm GP4 epithelium compared \nto non-cribriform glands (fold change 4.2; Padj<0.0 001). In cohort B, this \naligned with a significantly increased expression o f FASN in cribriform GS7 \nlesions compared to non-cribriform tumours (P=0.001 ). In cohort C, this \ncorresponded to a significant increase in mean SUVb w of cribriform lesions \ncompared to non-cribriform tumours (P<0.05 for all timepoints up to 60min \npost-injection). Conversely, the comparison of tumo ur-to-urine 1H-MRI \nADCratio derived from the whole cohort showed no di fference between the two \nGS7 phenotypes (P=0.56). \nConclusion: Clinical imaging of lipid metabolism is a biologica lly informed way \nof characterising cribriform and non-cribriform GS7  PCa, which is a challenge \nfor 1H-MRI. \nLimitations: Modest sample size dictated by study complexity. \nFunding for this study: Prostate Cancer UK, Cancer Research UK, \nAstraZeneca \nEthics committee - additional information: National Research Ethics \nService Committee East of England, Cambridge South,  Research Ethics \nCommittee; study numbers: 16/EE/0205, 03/018. Cambr idge University \nHospitals Local Ethics Committee (CUH/15/EE/0213), and the Administration \nof Radioactive Substances Advisory Committee (ARSAC , certificate reference \nRPC/83/400/33606). \nAuthor Disclosures:  \nLuigi Aloj: Nothing to disclose \nRichard Goodwin: Employee: AstraZeneca \nRoido Manavaki: Nothing to disclose \nNikita Sushentsev: Nothing to disclose \nDavid Lewis: Nothing to disclose \nGregory Hamm: Employee: AstraZeneca \nTristan Barrett: Nothing to disclose \nFerdia Aidan Gallagher: Nothing to disclose \nDmitry Soloviev: Nothing to disclose \n \n \nThe Role of [ 68 Ga]Ga FAPi PET/CT in Staging and R estaging in Breast \nCancer with Low FDG Uptake \nN. Alan Selcuk, *G. Beydagi*, K. Akcay, B. B. Oven,  S. Celik, L. Kabasakal; \nIstanbul/TR \n(gamzebeydagi@gmail.com) \n \nPurpose or Learning Objective: The aim of this study is to assess the \npotential efficacy of [68Ga]Ga FAPi PET/CT in stagi ng and restaging in breast \ncancer patients with FDG-negative or low FDG uptake  lesions. \nMethods or Background: Between October 2020 and February 2024, 25 \nfemale patients with breast cancer were prospective ly enrolled. These patients \nunderwent [68Ga]Ga-FAPi and [18F]-FDG PET/CT imagin g within one week \nfor staging or restaging. The maximum standard upta ke values (SUVmax) of \nthe primary tumor areas and metastases in the [68Ga ]Ga-FAPi and [18F]-FDG \nPET/CT images were recorded and statistically compa red using the paired t-\ntest. \nResults or Findings: 25 female patients with suspicious primary malignan cy \nrecurrence or metastases but low FDG affinity were imaged with [68Ga]Ga-\nFAPi PET/CT. The mean age was 57.1±11.7 years. Hist opathologic \nexamination available for 20 patients revealed lobu lar carcinoma in 10 cases, \nductal carcinoma in 8 cases, signet ring cell carci noma in one patient and \nsquamous cell carcinoma in one patient. In six pati ents (24%), neither the \n[18F]-FDG nor the [68Ga]Ga-FAPi PET/CT revealed any  findings indicating \nrecurrence or metastasis. Disease stage increased i n 36% (n=9) of patients \nafter [68Ga]Ga-FAPi PET/CT imaging, with 8 of them showing no pathologic \nfindings on [18F]-FDG PET/CT. 60% (n=6) of the lobu lar carcinomas were \nupstaged after [68Ga]Ga-FAPi PET/CT. The detection of lymph nodes and \ndistant metastases in lobular carcinoma was higher with [68Ga]Ga-FAPi \nPET/CT than with [18F]-FDG PET/CT. Furthermore, [68 Ga]Ga-FAPi PET/CT \nshowed a higher SUVmax in primary tumor foci and me tastases (p<0.05). \nConclusion: [68Ga]Ga-FAPi PET/CT has been shown to be superior for \nstaging in breast cancer, especially for lobular ca rcinoma with low FDG affinity. \nIt is anticipated that [68Ga]Ga-FAPi PET/CT will be  included in future \nguidelines for staging in breast cancer patients, e specially in patients with \nlobular carcinoma. \nLimitations: None \nFunding for this study: None \nEthics committee - additional information: Ethics committee approval no: \n1576 \nAuthor Disclosures:  \nLevent Kabasakal: Nothing to disclose \nBala Basak Oven: Nothing to disclose \nKaan Akcay: Nothing to disclose \nGamze Beydagi: Nothing to disclose \nSerkan Celik: Nothing to disclose \nNalan Alan Selcuk: Nothing to disclose \n \n \nCan 18F-FDG-PET CT Predict Mediastinal Lymph Node M etastases in \nNSCLC Patients Undergoing Neo-Adjuvant Immuno-Chemo therapy? \n*D. Kifjak*, M. J. Hochmair, A. Korajac, S. Pochepn ia, R-I. Milos, K. Sinn, \nA. Hoda, H. Prosch, L. Beer; Vienna/AT \n \nPurpose or Learning Objective: To evaluate the association between 18F-\nFDG-PET CT quantitative imaging markers of surgical ly resected mediastinal \nlymph nodes and histopathologic results in patients  with operable non-small \ncell lung cancer (NSCLC) treated with neo-adjuvant combined immuno-\nchemotherapy. \nMethods or Background: This preliminary analysis of a prospective, single-\ncenter study included 20 patients (8 male, 12 femal e) with NSCLC who were \ntreated with neo-adjuvant combined immune-chemother apy and underwent \npre-operative 18F-FDG-PET-CT. We semi-automatically  extracted the \nfollowing parameters for each mediastinal lymph nod e station: metabolic tumor \nvolume (MTV) and total lesion glycolysis (TLG) at p re-operative scans. The \nhistological results of mediastinal lymph nodes wer e retrieved from patients’ \nrecords. A Mann-Whitney-U-Test was calculated to co mpare MTV/TLG and \nhistological results for each mediastinal lymph nod e station. \nResults or Findings: A total of 191 mediastinal lymph nodes were surgica lly \nremoved. Four mediastinal lymph node metastases wer e found in three \npatients. In contrast 17 patients showed no tumor c ells in their mediastinal \nlymph nodes. The median MTV for positive lymph node s was 4.2 (range: 0-7.4) \ncompared to 0 (range 0-4.3) for negative lymph node s. The median TLG for \npositive lymph nodes was 6.3 (range 0-42) and for n egative lymph nodes it \nwas 0 (range 0-24). A statistically significant ass ociation between MTV and \nmediastinal lymph node metastases was observed, p=0 .046. While there was a \nnoticeable trend, no statistically significant asso ciation was identified between \nTLG and lymph node metastases, p=0.052. \nConclusion: High MTV values and to a lesser degree TLG are asso ciated with \nresidual lymph node metastases. However, both false  positive (inflammation) \nand negative findings (micrometastatis) occur, thus  limiting the ability of 18F-\nFDG-PET CT to predict histopathological response. \nLimitations: Single-center study. Selection bias, only patients with \nhistologically confirmed lymph nodes were part of t he study. \nFunding for this study: The Austrian Federal Ministry for Digital and \nEconomic Affairs, the National Foundation for Resea rch, Technology and \nDevelopment and the Christian Doppler Research Asso ciation is gratefully \nacknowledged. \nEthics committee - additional information: Local IRB # 1521/2015 \nAuthor Disclosures:  \nKatharina Sinn: Nothing to disclose \nMaximilian Johannes Hochmair: Nothing to disclose \nHelmut Prosch: Nothing to disclose \nLucian Beer: Nothing to disclose \nAida Korajac: Nothing to disclose \nDaria Kifjak: Nothing to disclose \nRuxandra-Iulia Milos: Nothing to disclose \nAlireza Hoda: Nothing to disclose \nSvitlana Pochepnia: Nothing to disclose \n\n \n \nSaturday \nAbstract-based Programme \n \n 238  \nIntegrated diagnostics for survival prediction in p atients with GEP-NET \nundergoing PRRT \n*F. Herr*, C. A. Dascalescu, M. P. Fabritius, M. Br endel, C. Auernhammer,  \nC. Spitzweg, J. Ricke, M. Heimer, C. C. Cyran; Muni ch/DE \n(felix.herr@med.uni-muenchen.de) \n \nPurpose or Learning Objective: Integrated biomarkers of survival for patients \nwith gastroenteropancreatic neuroendocrine tumors ( GEP-NETs) receiving \npeptide receptor radionuclide therapy (PRRT) are st ill limited. This study aims \nto identify predictors of progression-free survival  (PFS) in patients with GEP-\nNETs undergoing PRRT. \nMethods or Background: This single-center retrospective study included 178  \npatients with GEP-NETs (G1 and G2) who received at least two consecutive \ncycles of PRRT with [177Lu]Lu-DOTA-TATE and underwe nt somatostatin \nreceptor (SSTR) PET/CT before and after therapy. At  baseline, an assessment \nwas conducted in accordance with the Krenning score , and clinical parameters, \nincluding chromogranin A (CgA), neuron-specific eno lase (NSE), hemoglobin, \nKi-67, erythrocytes, C-reactive protein (CRP) and a lbumin were also collected. \nPFS was defined by a GEP-NET multidisciplinary team  assessment. \nResults or Findings: In univariate analysis at baseline, Krenning score 3, \nelevated levels of CgA (> 200 ng/dl) and NSE (>25 n g/dl) were significantly (p \n< 0.05) associated with shorter PFS. Ki-67 index > 5 %, primary tumor in the \npancreas, levels of erythrocytes > 4 Mio/ µl, CRP >  1 mg/dl and albumin < 4.1 \ng/dl at baseline were also significantly (p<0.05) c orrelated with a shorter PFS. \nIn multivariate analysis, Krenning score 3, CgA > 2 00 ng/ml, NSE > 35 ng/ml, \nand Ki-67 index > 5 % at baseline were significantl y (p < 0.05) associated with \nshorter PFS. Including the Krenning score at baseli ne leads to a significant \nimprovement of the cox regression model (p<0.05). O nly the Ki-67 index \n(z=2.55) showed a higher z-score than the Krenning score at baseline (z = \n2.41). \nConclusion: This study demonstrates the additional prognostic v alue of the \nKrenning score in conjunction with clinical paramet ers for patients with GEP-\nNET undergoing PRRT. \nLimitations: Limitations of this study are its retrospective sin gle-center design \nand the lack of multimodality imaging biomarkers. \nFunding for this study: Wilhelm Vaillant Stiftung \nEthics committee - additional information: Ethics Committee of LMU-\nMunich – Project number: 20-1077 \nAuthor Disclosures:  \nMatthias Philipp Fabritius: Nothing to disclose \nClemens C. Cyran: Nothing to disclose \nChristine Spitzweg: Nothing to disclose \nMaurice Heimer: Nothing to disclose \nFelix Herr: Nothing to disclose \nChristoph Auernhammer: Nothing to disclose \nMatthias Brendel: Nothing to disclose  \nChristian Alexander Dascalescu: Nothing to disclose  \nJens Ricke: Nothing to disclose \n \n \nMulti-radiotracer PET/CT for the evaluation of caro tid atherosclerotic \nplaque vulnerability: A systematic review \n*T. R. Readford*, P. Kench, M. Ugander, S. Patel, N . Giannotti; Sydney/AU \n(thomas.readford@sydney.edu.au) \n \nPurpose or Learning Objective: Carotid atherosclerosis is a major contributor \nto the burden of cerebrovascular diseases. Conventi onal imaging is limited in \ninterrogating the biological and functional charact eristics that may increase \nplaque vulnerability. Multiple positron emission to mography/computed \ntomography (PET/CT) radiotracers provide novel diag nostic insights into \nplaque vulnerability and identify patients at highe r risk of cerebrovascular \nevents. This systematic review investigated the cli nical role of PET radiotracers \nin identifying vulnerable carotid atherosclerotic p laque. \nMethods or Background: A systematic review of the existing literature was \nperformed using the following search strategy: ‘car otid’, ‘PET’, \n‘atherosclerosis’, ‘plaque’ and ‘vulnerability’. On ly original research articles \nwere included. Covidence was used for data screenin g and data extraction. \nResults or Findings: Thirty-nine studies were included that used 18F-\nfluorodeoxyglucose (18F-FDG), 18F-sodium fluoride ( 18F-NaF), 18F-\nfluoromisonidazole (18F-MISO), 68Ga-DOTATATE, 68Ga- Pentixafor and 11C-\nAcetate to target plaque metabolism-related inflamm ation, microcalcification, \nhypoxia, activated macrophages, C-X-C motif chemoki ne receptor4 and fatty \nacid synthesis, respectively. Seven studies used du al PET radiotracers with \ntime intervals between scans ranging from one day t o 4.8 months. Correlation \nbetween PET imaging and histology post-carotid enda rterectomy was available \nin 17 studies. Authors noted agreement between macr ophage-driven plaque \ninflammation by PET/CT and vulnerability-related mo rphological changes by \nMRI, suggesting complementary roles of combined MRI  and PET/CT in \ndetecting vulnerable plaque. Significant variabilit y was observed in reported  \n \n \nPET/CT acquisition techniques, injected radiotracer  dose [18F-FDG: 185-\n925MBq, 18F-NaF: 125-370 MBq, 68Ga-DOTATATE: 148-15 7 MBq] and \nuptake times [18F-FDG: 50-180min ,18F-NaF: 60-180mi n, 68Ga-DOTATATE: \n60-120min]. \nConclusion: The use of multiple PET radiotracers may provide ad vanced \ndiagnostic insights into carotid atherosclerotic pl aque vulnerability. Further \nresearch is necessary to establish consensus on wha t constitutes a standard \napproach for the evaluation of vulnerable carotid p laque by PET/CT. \nLimitations: Further quantitative analysis was limited by the va riability of \nimaging parameters used across studies in this revi ew. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This was a systematic review of \nexisting literature. \nAuthor Disclosures:  \nPeter Kench: Nothing to disclose  \nMartin Ugander: Nothing to disclose  \nNicola Giannotti: Nothing to disclose \nThomas Ramsay Readford: Nothing to disclose \nSanjay Patel: Nothing to disclose \n \n \nLymphoma Therapy Response Assessment with Low-Dose [18F]FDG \nTotal-Body PET/CT \n*C. Mingels*¹, K. J. Chung¹, H. Nalbant¹, A. Willey ¹, L. K. Shiyam Sundar²,  \nY. G. Abdelhafez¹, R. Badawi¹, B. A. Spencer¹, L. N ardo¹;  \n¹Sacramento, CA/US, ²Vienna/AT \n \nPurpose or Learning Objective: Our aim was to identify the lower limit of \ninjected dose for [18F]FDG Total-Body (TB) PET/CT i n lymphoma therapy \nresponse assessment. \nMethods or Background: In this prospective study 24 patients with biopsy-\nproven lymphoma were enrolled for interim or end-of -treatment TB PET/CT \nafter 1h and 2h of the injection of ~3.0MBq/Kg [18F ]FDG. Lower injected \nactivities (1.0 MBq/kg, 0.5MBq/kg, 0.25MBq/kg, 0.12 5MBq/kg) were simulated. \nLesions were segmented by an artificial intelligenc e-aided software and \nconfirmed by an expert. Standardized-uptake values (SUVmax/mean/peak), \nmetabolic tumor volume (MTV) and total-lesion glyco lysis (TLG) were \ncalculated. Additionally, total MTV (TMTV) was asse ssed for each patient. \nLiver and mediastinal blood-pool were used to calcu late tumor-to-background \nratio (TBR) and contrast-to-noise level (CNR). Ther apy response assessment \nwas performed by Deauville criteria. \nResults or Findings: In total, 182 lymphoma lesions were analyzed. \nSUVmax/mean/peak, MTV, TLG, TBR and TMTV were not s ignificantly \ndifferent between reference standard and low-dose i mages. Image noise \nincreased significantly with lower doses. CNR decre ased significantly. Clinical \ntherapy response assessment by Deauville Score was significantly lower \nbetween 0.125MBq/kg and reference standard (p<0.01)  for 1h p.i. imaging. All \nother low-dose reconstructions revealed no signific ant differences. For 2h p.i. \nthere was a significant difference in Deauville Sco re for 0.5MBq/kg, \n0.25MBq/kg and 0.125MBq/kg compared to the referenc e standard (p<0.01). \nDeauville Scores for 1MBq/kg at 2h were not signifi cantly different to the \nreference standard (p=0.16). \nConclusion: Dose reduction in therapy response assessment with TB PET/CT \nis possible to a lower limit of 0.25MBq/kg for 1h p .i. imaging and 1.0MBq/kg for \n2h p.i. TB PET/CT. However, lower injected activiti es are at risk to \nunderestimate the metabolic activity of the lymphom a lesions due to higher \nnoise levels. TMTV and TLG were not different in ul tra-low-dose [18F]FDG TB \nPET/CT. \nLimitations: Small cohort, simulated low dose images \nFunding for this study: The work was also supported by the In Vivo \nTranslational Imaging Shared Resources with funds f rom NCI P30CA093373 \nand by the Fred and Julia Rusch Foundation for Nucl ear Medicine Research \nand Education. Hande Nalbant’s funding is partially  provided by United \nImaging Health’s UIH Fellowship Gift. \nEthics committee - additional information: This study was approved by the \nUC Davis institutional review board (IRB1470016). W ritten informed consent \nfor inclusion was obtained. The study was performed  in accordance with the \nDeclaration of Helsinki. \nAuthor Disclosures:  \nHande Nalbant: Nothing to disclose \nKevin J. Chung: Nothing to disclose \nClemens Mingels: Nothing to disclose \nRamsey Badawi: Nothing to disclose \nLorenzo Nardo: Nothing to disclose \nAndrew Willey: Nothing to disclose \nLalith Kumar Shiyam Sundar: Nothing to disclose \nBenjamin A. Spencer: Nothing to disclose \nYasser Gaber Abdelhafez: Nothing to disclose \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 239  \n14:00-15:30 Research Stage 4 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 2001 \nShining a spotlight on HCC and liver \ntransplant \n \nModerator \nW. K. Jeong; Seoul/KR  \n(jeongwk@gmail.com) \nAuthor Disclosures:  \nWoo Kyoung Jeong: Consultant: Guerbet; Research Gra nt/Support: GE \nHealthcare; Speaker: GE Healthcare \n \n \nThe predictive value of iodine map histogram analys is of non-tumorous \nhepatic parenchyma for post-hepatectomy Liver Failu re after narrow-\nmargin hepatectomy in hepatocellular carcinoma \n*Y. Xu*, J. Liu, J. Zhou; Lanzhou City/CN \n(xuyuan961030@163.com) \n \nPurpose or Learning Objective: Post-hepatectomy liver failure (PHLF) is a \nsevere postoperative complication with a high incid ence and mortality rate, \nparticularly in patients with narrow-margin (NM). T his study aims to predict \nPHLF in NM-hepatocellular carcinoma (HCC) using iod ine map histogram \nanalysis of non-tumorous hepatic parenchyma. \nMethods or Background: A retrospective analysis was conducted on the \nclinical and imaging data of 107 patients with NM-H CC who underwent \nhepatectomy, divided into those with PHLF (n=45) an d without PHLF (n=62). \nHistogram parameters of non-tumorous hepatic parenc hyma were calculated \nfrom iodine map derived from the portal venous phas e of spectral CT, including \nMin, Max, Mean, SD, Skewness, Kurtosis, Entropy, an d percentiles (V10-V95), \nalong with the future liver remnant volume (FLV) an d standardized future \nresidual liver volume ratio (SFLV%). Logistic regre ssion analyse was used to \nidentify independent predictors of PHLF, and a comp rehensive model was \ndeveloped. The performance of the new comprehensive  model was assessed \nusing ROC curve analysis and was compared with ALBI  and MELD scores. \nResults or Findings: Significant intergroup differences were observed in  the \niodine map histogram analysis of non-tumorous hepat ic parenchyma for \nSkewness, Kurtosis, V50, V75, V90, V95, FLV, and SF LV% (P < 0.01 to P = \n0.04). Multivariate logistic regression analysis re vealed that V95, Kurtosis, and \nSFLV% were independent risk factors for predicting PHLF. The comprehensive \nmodel (ModelALL), developed by combining these inde pendent risk factors, \nexhibited the highest predictive efficacy for PHLF,  with an AUC of 0.77 (0.67-\n0.87), outperforming both the ALBI and MELD scores,  which had AUCs of 0.70 \n(0.58-0.81) and 0.62 (0.49-0.74), respectively. \nConclusion: The model which combines the iodine map histogram p arameters \nof non-tumorous hepatic parenchyma (V95 and Kurtosi s) with SFLV%, aids in \nthe preoperative prediction of PHLF in NM-HCC patie nts and outperforms \nconventional scoring systems. \nLimitations: Not applicable \nFunding for this study: This study has received funding by grants of Natura l \nScience Foundation of China (82260361, 82371914), L anzhou University \nSecond Hospital Second Hospital “Cuiying Technology  Innovation Plan” \n(CY2022-QN-A10), Lanzhou University Second Hospital  \"Cuiying Postgraduate \nInstructor Cultivation Program Project (CYDSPY20200 3) and Outstanding \nYoung Talents and Backbone Talents Project of Gansu  Provincial Health \nIndustry Research Program (GSWSQN2023-04). \nEthics committee - additional information: The present study was approved \nby our hospital ethics committee (No. 2022A-112) an d performed according to \nthe ethical guidelines of the 1975 Declaration of H elsinki. \nAuthor Disclosures:  \nJunlin Zhou: Nothing to disclose \nYuan Xu: Nothing to disclose \nJianli Liu: Nothing to disclose \n \n \n \n \n \n \n \n \n \nDiagnostic performance of an abbreviated magnetic r esonance protocol \nfor surveillance of hepatocellular carcinoma in can didates for liver \ntransplant \n*M. Mattone*, F. Quintarelli, A. Napoli, C. Catalan o; Rome/IT \n(monicamattone95@gmail.com) \n \nPurpose or Learning Objective: To evaluate the per-patient diagnostic \nperformance of an abbreviated magnetic resonance pr otocol for hepatocellular \ncarcinoma (HCC) surveillance in cirrhotic liver. \nMethods or Background: Retrospective study including cirrhotic patients wi th \nHCC who are candidates for liver transplant enrolle d in a surveillance program \nbased on the use of MRI with hepatobiliary contrast  agent. Two different sets \nof images for each patient are provided to two radi ologists, who interpret the \nimages independently, to simulate an abbreviated pr otocol. Interobserver \nvariability was assessed using Cohen's kappa coeffi cient. A reference standard \nbased on histologic examination or radiologic crite ria (LIRADS at least equal to \n4) was used to determine the diagnostic accuracy of  each set of images. \nResults or Findings: A total of 200 patients who underwent MRI for HCC \nsurveillance were included in this study. One set o f images consisted of the \ncomplete protocol (Set1) and one set of images cons isted of T2WI with fat \nsaturation, DWI and hepatobiliary sequences (Set2). The sensitivity, specificity, \nand accuracy of Set of readers 1 and 2 were 91.5%/9 0.5%, 88.6.4%/87.6% \nand 87.5%/85.0%, respectively. The sensitivities of  the sets were not \nsignificantly different. Inter-reader agreement was  substantial. Ascites, \nprevious surgical and interventional radiology trea tments, and small tumor size \nare associated with lower sensitivity. \nConclusion: An abbreviated MRI protocol including T2WI with fat  saturation, \nDWI and hepatobiliary sequences is highly sensitive  and can be a valid method \nfor HCC surveillance in cirrhotic liver in patients  candidates to liver transplant. \nLimitations: Low number of readers \nFunding for this study: No founds were used \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMonica Mattone: Nothing to disclose \nFabio Quintarelli: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nAlessandro Napoli: Nothing to disclose \n \n \nImpact of CT-defined sarcopenia on survival in pati ents undergoing \northotopic liver transplant \n*D. Tore*¹, C. Guarnaccia¹, A. Depaoli², C. Gaetani ¹, M. Anna Pia¹,  \nM. Visciano¹, M. Dini¹, F. Tandoi³, P. Fonio¹; ¹Tur in/IT, ²Ivrea/IT, ³Bari/IT \n \nPurpose or Learning Objective: To evaluate the impact of CT-defined \nsarcopenia on survival in patients undergoing ortho tropic liver transplant (OLT). \nMethods or Background: Monocentric retrospective study. 440 patients (70 \nfemales, 370 males) who underwent OLT at our Instit ution within 01.01.2014 \nand 31.12.2019 with an abdomen CT scan acquired wit hin six month before \nsurgery were selected. Skeletal muscles segmentatio ns at the level of L3 and \nL4 were performed using open source software 3D Sli cer using thresholds (-29 \nto 150 HU); psoas muscle area (PMA), skeletal muscl e area (SMA), psoas \nmuscle index (PMI) and skeletal muscle index (SMI) were calculated at both \nlevels. Optimal cut-offs to dichotomize between sar copenic and non-\nsarcopenic patients divided in females and males we re calculated using \nYouden's J statistic and ROC curves. Survival analy sis was performed using \nKaplan-Meyer's curves. \nResults or Findings: Male patients classified as sarcopenic according to  SMI \ncriterion at L3 level presented an increased risk o f mortality compared to non-\nsarcopenic with an hazard ratio of 1.63 (p=0.03). M ale patients classified as \nsarcopenic according to SMA criterion at L3 level p resented an increased risk \nof mortality compared to non-sarcopenic with an haz ard ratio of 1.87 (p=0.008). \nIn the female group none of the sarcopenia definiti ons tested highlighted an \nincreased risk of mortality with p values always >0 .05. \nConclusion: PMI and SMA CT-defined sarcopenia at L3 level repre sented a \nnegative prognostic factor for male patients surviv al after OLT. The use of such \ndefinitions of sarcopenia may identify more fragile  subjects in this setting and \nimprove care and, possibly outcomes, for such patie nts. The absence of \nstatistically significant differences between sarco penic and non-sarcopenic \nfemale subjects may be due to the relatively limite d sample size for such group \nof patients. \nLimitations: Monocentric retrospective study. \nFunding for this study: No funds were received for this work. \nEthics committee - additional information: Not applicable, retrospective \nstudy \n \n \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 240  \nAuthor Disclosures:  \nMicheletti Anna Pia: Nothing to disclose \nAlessandro Depaoli: Nothing to disclose \nFrancesco Tandoi: Nothing to disclose \nCarla Guarnaccia: Nothing to disclose \nMassimiliano Dini: Nothing to disclose \nDavide Tore: Nothing to disclose \nPaolo Fonio: Nothing to disclose \nClara Gaetani: Nothing to disclose \nMichele Visciano: Nothing to disclose \n \n \nHepatic functional reserve score based on two-dimen sional shear wave \nelastography for evaluation of preoperative hepatic  functional reserve \nand prediction of post-hepatectomy liver failure in  HCC \n*X. Zhong*; Guangzhou/CN \n(zhongx35@mail3.sysu.edu.cn) \n \nPurpose or Learning Objective: To establish a Hepatic Functional Reserve \n(HFR) score based on liver stiffness (LS) and liver  volume, and to explore its \nfeasibility in evaluating preoperative hepatic rese rve and predicting \npostoperative remnant hepatic function in hepatocel lular carcinoma (HCC). \nAdditionally, the study aimed to determine a safe f uture liver remnant (FLR) \nvolume threshold to prevent post-hepatectomy liver failure (PHLF). \nMethods or Background: The study included 345 HCC patients, with 265 in \nthe training group and 80 in the test group. Preope rative LS was measured \nusing two-dimensional shear-wave elastography. Tota l liver volume (TLV), \ntumor volume (TuV), and FLR were simulated using th e IQQA system. The \nmedian LS of patients without significant fibrosis (7.6kPa) and standard liver \nvolume (SLV) were used as normal references for liv er quality and quantity. \nThe preoperative HFR score (HFR-PRE) was calculated  as 7.6/LS×(TLV-\nTuV)/SLV, and its correlation with liver function i ndicators was assessed. The \npostoperative HFR score (HFR-POST) was calculated a s 7.6/LS×FLR/SLV, \nand its effectiveness in predicting PHLF was evalua ted. The optimal FLR \nthreshold was determined using the maximum Youden i ndex. \nResults or Findings: HFR-PRE showed significant correlation with \npreoperative liver function indicators such as Chil d-Pugh, MELD, ALBI scores, \nand ICG-R15 (all p<0.05). The AUC for HFR-POST in p redicting PHLF was \n0.864 in the training group and 0.857 in the test g roup. HFR-POST \noutperformed ALBI, MELD, and ICG-R15 scores (all p< 0.05) in predicting \nPHLF. The minimum FLR/SLV threshold was 5/76*LS to prevent PHLF, with a \nsensitivity of 81.5% and specificity of 77.3%. \nConclusion: The HFR score effectively evaluates preoperative li ver function \nand predicts PHLF in HCC patients. It also helps de termine a safe residual \nliver volume to reduce the risk of PHLF. \nLimitations: LS measurement acquired from one liver segment migh t not \nrepresent the quality of the whole liver. \nFunding for this study: None \nEthics committee - additional information: The study protocol received \napproval from the Institutional Review Board of The  First Affiliated Hospital of \nSun Yat-sen University (IRB approval number: [2019] 046). \nAuthor Disclosures:  \nXian Zhong: Nothing to disclose \n \n \nComparative Analysis of Semi-Automated CT Volumetry  and \nIntraoperative Graft Weight in Living Donor Liver T ransplantation in a \ntertiary care centre \n*A. Garg*, J. Chowdhary, P. K. Sukhani, S. P. Joshi ; Jaipur/IN \n(ashnagarg9@gmail.com) \n \nPurpose or Learning Objective: To evaluate the accuracy of semi-automated \nCT volumetry (CTV) in estimating liver volume in po tential living donor liver \ntransplant (LDLT) donors, compared to the gold stan dard of intraoperative graft \nweight measurement. Aim: 1. Inform transplant surge ons and radiologists \nabout the reliability of semi-automated CTV in LDLT  donor evaluation. 2. \nHighlight the potential benefits of using CTV in pr eoperative planning, \nimproved donor selection and reduced surgical compl ications. 3. Discuss the \nimplications of accurate volume assessment on LDLT outcomes and patient \nsafety. \nMethods or Background: Background: Living Donor Liver Transplantation \n(LDLT) is a life-saving procedure for patients with  end-stage liver disease. \nAccurate liver volume assessment is crucial in LDLT  to ensure sufficient graft \nvolume for recipient survival, prevent small-for-si ze graft syndrome and reduce \ndonor morbidity and mortality. Methods: This retros pective study included 70 \npotential LDLT donors who underwent: Semi-automated  CTV using AW \nVolumeShare 7 on 128 slice - GE Healthcare, Optima.  Inclusion Criteria: Adult \nLDLT donors (>18 years) Exclusion Criteria: Previou s liver surgery/disease \nVariant vascular/biliary anatomy, not allowing safe  resection \nResults or Findings: Primary Outcome: Strong correlation (r = 0.92, p < \n0.001) Sensitivity: 93.2% (95% CI: 85.1-97.5) Speci ficity: 90.5% (95% CI: 81.2-\n95.8) Mean difference between CTV and intraoperativ e graft weight: 25.6 ± \n57.8 grams Secondary Outcomes: Bland-Altman analysi s showed good \nagreement between CTV and intraoperative graft weig ht, with 95% limits of \nagreement (-89.2 to 140.4 grams) CTV accurately pre dicted graft weight within \n10% of actual weight in 85% of cases (n = 70) \nConclusion: Semi-automated CTV demonstrates high accuracy and r eliability \nin estimating liver volume in LDLT donors. CTV accu rately predicts graft weight \nwith good agreement with intraoperative measurement s thus helping in \npreoperative evaluation and surgical planning for L DLT. \nLimitations: Single-center experience, Operator dependence, Intr aoperative \ngraft weight measurement variability \nFunding for this study: No external funding \nEthics committee - additional information: Approved by the Institutional \nEthics Committee, Mahatma Gandhi Medical College an d Hospital, India \nAuthor Disclosures:  \nSwati Purohit Joshi: Nothing to disclose \nJai Chowdhary: Nothing to disclose \nParesh Kumar Sukhani: Nothing to disclose \nAshna Garg: Nothing to disclose \n \n \nLI-RADS Nonradiation Treatment Response Algorithm V ersion 2024: \nDiagnostic Performance and Added Value of Ancillary  Features \n*S. Zhou*, Y-C. Wang; Nanjing/CN \n \nPurpose or Learning Objective: We aimed to evaluate the diagnostic \nperformance of the MRI-based Nonradiation LR-TRA v2 024 and the value of \nincorporating ancillary features (AFs) on improving  diagnostic precision. \nMethods or Background: This retrospective study included patients with HCC  \nwho underwent local-regional therapy (LRT) followed  by curative treatments \nbetween January 2017 and December 2022. Each treate d lesion was \nevaluated according to the LR-TRA v2024, LR-TRA v20 17, and modified \nResponse Evaluation Criteria in Solid Tumors (mRECI ST) criteria, with \npathologic response serving as the reference. The s ensitivity, specificity, and \naccuracy of different treatment response criteria w ere compared using the \nMcNemar test. \nResults or Findings: A total of 231 patients (198 males; median age, 56 \nyears; IQR, 50-63 years) with 306 treated lesions ( 249 incomplete pathologic \nnecrosis) were evaluated. LR-TRA v2024-Viable (with out AFs) exhibited \ncomparable sensitivity and accuracy than LR-TRA v20 17-Viable and mRECIST \n(sensitivity: 81.1% [95%CI: 75.8, 85.8], 79.5% [95%  CI: 74.0, 84.4], and 80.3% \n[95% CI: 74.8, 85.1]; accuracy: 80.1% [95% CI: 75.1 , 84.4], 79.4% [95% CI: \n74.4, 83.8] and 75.2% [95% CI: 69.9, 79.9], respect ively) for predicting \nincomplete pathologic tumor necrosis. Both LR-TRA v 2024 (without AFs)-\nViable and v2017-Viable exhibited significantly hig her specificity than \nmRECIST (both P ≤ .001). LR-TRA v2024 (with AFs)-Viable incorporatin g \nancillary features exhibited the highest sensitivit y (85.9% [95% CI: 81.0, 90.0]) \nand accuracy (83.7% [95% CI: 79.0, 87.6]) among dif ferent treatment response \ncriteria, and showed a statistic difference compare d to LR-TRA v2017-Viable \n(both P = .006), without sacrificing specificity (7 3.7% [95% CI: 60.3, 84.5]). \nConclusion: LR-TRA v2024 (without AFs) demonstrates good diagno stic \nperformance and ease of use. AFs significantly enha nce diagnostic sensitivity \nand accuracy through category adjustments, without sacrificing specificity. LR-\nTRA v2024 combined with AFs is strongly recommended  for use in clinical \npractice. \nLimitations: Selection bias; only based on MRI \nFunding for this study: This study has received funding by four National \nNatural Science Foundation of China (NSFC, No. 8227 1978, 92359304, \n82330060, 823B2040); Zhongda Hospital Affiliated to  Southeast University, \nJiangsu Province High-Level Hospital Pairing Assist ance Construction Funds \n(No. zdyyxy09). \nEthics committee - additional information: This study was approved by our \ninstitutional review board, and the requirement for  written informed consent \nwas waived for the retrospective data [No. 2022ZDSY LL410-P01]. \nAuthor Disclosures:  \nYuan-Cheng Wang: Nothing to disclose \nShuwei Zhou: Nothing to disclose \n \n \nShort MRI Surveillance (SMS) for hepatocellular car cinoma screening: \nfirst results on image quality of the SMS-HCC study  \n*C. Van De Braak*, F. Willemssen, F. Smits, R. De M an, A. Van Der Lugt,  \nD. Bos, R. S. Dwarkasing; Rotterdam/NL \n(c.vandebraak@erasmusmc.nl) \n \nPurpose or Learning Objective: Current guidelines recommend biannual US \nscreening for patients with high risk of developing  hepatocellular carcinoma \n(HCC). It was reported that the sensitivity of US f or detecting early-stage HCC \nin these patients is merely 47%. Our aim is to vali date a Short MRI \nSurveillance (SMS)-protocol in current surveillance  patients and compare it to \nUS in a prospective, multicentre study. Here, we pr esent our first results on \nimage quality. \n\n \n \nSaturday \nAbstract-based Programme \n \n 241  \nMethods or Background: From November 2023, patients from the current \nHCC surveillance programme were invited to undergo paired US-MRI \nscreening. The MRI was performed on 1.5/3.0-T syste ms using a dedicated 8-\n16 channel range body coil. The protocol consisted of T1W in-out phase, T2W \nwith fat saturation and DWI. One radiologist evalua ted the US- and MRI \nimages, while a second and third radiologist solely  evaluated the MRI images. \nUS was reported according to the LIRADS-US surveill ance classification. The \ncurrent analyses were restricted to the first 50 st udy participants and revolve \naround descriptive analyses of the image quality an d presence of lesions. \nResults or Findings: The most common indication for the HCC surveillance  \nprogramme was Hepatitis B (35/50, 70%), followed by  cirrhosis (16/50, 32%). \nImage quality was rated as good in 64% of the US ex aminations and 94% of \nthe MRI examinations. Based on US, a total of 16 le sions were found in eight \npatients, whilst on MRI this total was 109 lesions in 21 patients. Reported \nlesions were benign (e.g. cysts), except for one le sion that was noted on SMS, \nbut undetected on US, which proved to be LIRADS-3 o n subsequent contrast-\nenhanced MRI. \nConclusion: Our preliminary results show the potential of MRI f or HCC \nsurveillance, through better image quality and a hi gher detection rate of focal \nlesions. \nLimitations: No limitations were identified. \nFunding for this study: Funding was provided by a grant of the Dutch Cancer  \nSociety (grant number: 2021-2/13803). \nEthics committee - additional information: This study was approved by the \ninstitutional review board and written consent was obtained from all \nparticipants. \nAuthor Disclosures:  \nAad Van Der Lugt: Nothing to disclose \nRob De Man: Nothing to disclose \nRoy S. Dwarkasing: Nothing to disclose \nFokko Smits: Nothing to disclose \nFrançois Willemssen: Nothing to disclose \nDaniel Bos: Nothing to disclose \nCéline Van De Braak: Nothing to disclose \n \n \nEnhanced prediction of microvascular invasion in he patocellular \ncarcinoma: a comparative study between intraoperati ve ultrasound and \nCT radiomics \n*F. Rizzetto*, S. Tortora, E. Rondi, P. Carboni, M.  M. B. Barabino, A. Vanzulli; \nMilan/IT \n(Francesco.rizzetto@unimi.it) \n \nPurpose or Learning Objective: Intraoperative ultrasound (IOUS) can identify \nsigns of microvascular invasion (MVI) in hepatocell ular carcinoma (HCC) \nduring surgery. We evaluated whether preoperative C T-derived radiomics can \noffer a less invasive alternative. \nMethods or Background: All patients who underwent surgical resection with \nIOUS for histologically confirmed HCC lesions were retrospectively included. \nFor those with available preoperative triphasic CT scans, HCC nodules were \nsegmented across the arterial, venous, and delayed phases. Using \nPyRadiomics, radiomic features (RFs) were extracted  from each segmentation. \nAfter dimensionality reduction, the selected RFs fr om each phase were used to \ntrain and validate various predictive models based on Support Vector Machine \n(SVM) algorithm, with histopathological confirmatio n of MVI as ground truth. \nThe best performing model was tested on an independ ent dataset. Qualitative \nIOUS features were selected through multivariate re gression and used to build \na corresponding SVM model. Performance for detectin g MVI was assessed \nusing the area under the Receiver Operating Charact eristic curve (AUC-ROC), \nwith model comparisons made using the DeLong test. \nResults or Findings: A total of 124 patients, each with a single HCC les ion, \nwere selected, with preoperative CT scans acquired in nearly 30 different \ninstitutions. Of them, 86 patients were assigned to  the training and validation \ndataset (80:20 split), while the remaining 38 serve d as independent test \ndataset. The most performing radiomic model include d 10 RFs extracted from \narterial phase. ROC analysis in the independent tes t yielded an AUC-ROC of \n70% (95% confidence interval[CI]: 52-88%; p=0.030) for radiomic model and \n70% (95%CI: 51-87%; p=0.020) for IOUS model, with n o statistical difference \n(p=1.00). \nConclusion: Preoperative CT-derived radiomics demonstrated equi valent \npredictive performance to IOUS in assessing MVI in HCC, offering a non-\ninvasive preoperative alternative for MVI risk stra tification. \nLimitations: Retrospective design is the main study limitation. \nFunding for this study: No fundings were received for this study. \nEthics committee - additional information: Institutional Review Board \napproved the retrospective data collection in anony mous form \n \n \n \n \n \n \nAuthor Disclosures:  \nFrancesco Rizzetto: Nothing to disclose \nPierluigi Carboni: Nothing to disclose \nAngelo Vanzulli: Nothing to disclose \nSilvia Tortora: Nothing to disclose \nElisa Rondi: Nothing to disclose \nMatteo Mario Bruno Barabino: Nothing to disclose \n \n \nA CT-based radiomics model to predict P53-mutated h epatocellular \ncarcinoma \n*Y. Shi*, M. Li, Y. Pei, W. Li; Changsha/CN \n(sytsyt1998@163.com) \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of a \nCT-based radiomics model for predicting P53-mutated  hepatocellular \ncarcinoma (HCC). \nMethods or Background: In this retrospective single-center study, patients  \nwith histopathologically confirmed HCC who underwen t preoperative \ncontrasted-enhanced CT examination and surgery betw een November 2017 \nand July 2022 were recruited. HCC was classified in to P53-mutated HCC and \nnon-P53-mutated HCC using the gene sequencing. Radi ological features were \nanalyzed and clinical information were collected. R adscore was based on \nradiomics features extracted from the plain scan, a rterial phase and portal vein \nphase images using the random forest method. Univar iable and multivariable \nlogistic regression analyses were used to identify variables that were \nsignificantly and independently associated with P53 -mutated HCC, which were \nfurtherly to develop a model. The model performance  was evaluated with the \narea under the receiver operating characteristic cu rve (AUC), and the log-rank \ntest was used to analyze recurrence-free survival ( RFS). \nResults or Findings: A total of 109 patients were enrolled, and assigned  \nrandomly (8:2) into training (87 patients) and vali dation sets (22 patients). The \nunivariable analysis showed that the presence of en hancing capsule and \nintratumoral artery, and the radscore were signific ant risk factors for P53-\nmutated HCC. Further multivariable analysis identif ied that only the radscore \nwas the independent predictor (odds ratio (OR), 3.0 5 [95% CI, 2.00–4.67], \np<0.001) for P53-mutated HCC, and was used to devel op the radiomics model. \nThe model showed excellent performance in predictin g P53-mutated HCC, with \nAUC of 0.899 (95% CI: 0.836-0.962) in the training set and 0.744 (95% CI: \n0.523-0.964) in the validation set, Patients with p redicted P53-mutated HCC \nhas shorter RFS than those with predicted non-P53-m utated HCC (p=0.03). \nConclusion: A CT-based radiomics model could accurately predict  P53-\nmutated HCC. \nLimitations: A single-center retrospective study with a small sa mple. \nFunding for this study: Funding: This work was supported by the National \nNatural Science Foundation of China (82071895, Wenz heng Li; 82271984, \nWenzheng Li), the Natural Science Foundation of Hun an Province \n(2023JJ30903, Wenzheng Li; 2022JJ30950, Yigang Pei) , the Natural Science \nFoundation for Youth of Hunan Province (2024JJ6665,  Mengsi Li; \n2023JJ40970, Wenguang Liu), the Youth Science Found ation of Xiangya \nHospital (2023Q06, Mengsi Li) and Postdoctoral Fell owship Programof CPSF \n(GZC20242047, Mengsi Li). \nEthics committee - additional information: Institutional Review Board \napproval was abtained. \nAuthor Disclosures:  \nWenzheng Li: Nothing to disclose \nYuting Shi: Nothing to disclose \nYigang Pei: Nothing to disclose \nMengsi Li: Nothing to disclose \n \n \nThe role of gadoxetic acid-enhanced MRI-derived res idual relative \nenhancement index (RREI) in quantifying liver funct ion in hepatocelulary \ncarcinoma patients \n*U. Eryürük*, M. N. Tasdemir, E. Cakir, S. Aslan; G iresun/TR \n(uluhaneryuruk@gmail.com) \n \nPurpose or Learning Objective: To investigate the efficacy of the residual \nrelative enhancement index (RREI), derived from gad oxetic acid-enhanced \nMRI, in estimating liver function in patients with hepatocellular carcinoma \n(HCC), by validating with the albumin-bilirubin (AL BI) grade. \nMethods or Background: We retrospectively analyzed 41 patients with HCC \nwho underwent gadoxetic acid-enhanced MRI. Enhancem ent ratio (ER) was \ncalculated using the formula ER=(SI_HBP20-SI_pre)/S I_pre, where SI_HBP20 \nrepresents the signal intensity in the hepatobiliar y phase and SI_pre \nrepresents the pre-contrast signal intensity. RREI was calculated as \nRREI=residual liver volume (RLV)×ER. Receiver opera ting characteristic curve \nanalysis was performed to determine optimal cut-off  values of RLV, ER, and \nRREI for predicting ALBI grades. Spearman's rank co rrelation was used to \nevaluate correlations between RLV, ER, RREI, and AL BI scores. Intraclass \ncorrelation coefficient (ICC) was used to assess in tra-reader reliability and \ninter-reader agreement for RLV, ER, and RREI measur ements. \n\n \n \nSaturday \nAbstract-based Programme \n \n 242  \nResults or Findings: ROC analysis showed that the optimal RREI cut-off f or \npredicting ALBI grade 1 was 680-698, and for ALBI g rade 3 was 537-496, for \nreader 1 and reader 2, respectively. RREI demonstra ted good performance in \npredicting ALBI grade 1, with accuracy of 85.3%-95. 1%, sensitivity of 89.4%-\n94.7%, and specificity of 81.8%-95.4%. In different iating ALBI grade 3 from \nother grades, RREI showed excellent performance, wi th accuracy of 92.7%-\n97.6%, sensitivity of 90.6%-96.9%, and specificity of 100%. Strong correlations \nwere observed between RREI and ALBI scores, with co rrelation coefficients of \n-0.852 and -0.839 for both readers. Intra-reader an d inter-reader reliability was \nalmost perfect, with ICC values of 0.975 and 0.937,  respectively. \nConclusion: RREI exhibited a strong correlation with ALBI score s for \nassessing liver function in HCC patients and showed  good accuracy in \npredicting ALBI grades, indicating its potential as  a reliable radiological tool for \nevaluating liver function. \nLimitations: This was a retrospective study. \nFunding for this study: None. \nEthics committee - additional information: The study was approved by the \nlocal ethics commitee. \nAuthor Disclosures:  \nUluhan Eryürük: Nothing to disclose \nMerve Nur Tasdemir: Nothing to disclose \nErtugrul Cakir: Nothing to disclose \nSerdar Aslan: Nothing to disclose \n \n \nSarcopenia improves after transjugular intrahepatic  portosystemic shunt \n(TIPS) placement in patients with cirrhosis \n*J. Kittinger*, T. Müllner-Bucsics, L. Hartl, L. Re ider, F. Wolf, M. Trauner,  \nM. Mandorfer, T. Reiberger, K. Lampichler; Vienna/A T \n \nPurpose or Learning Objective: Transjugular intrahepatic portosystemic \nshunt (TIPS) is used to treat complications of port al hypertension in patients \nwith cirrhosis. Sarcopenia in cirrhosis has been li nked to worse patient \noutcomes but may improve after TIPS. We aimed to ev aluate the prevalence of \nsarcopenia in patients undergoing TIPS and its evol ution after TIPS. \nMethods or Background: Retrospective analysis of the Vienna TIPS cohort \nfrom 01/2004 to 03/2022. Transversal psoas muscle t hickness (TPMT) was \nevaluated on cross-sectional abdominal imaging both  (i) prior to TIPS (time \ninterval < 3 months) and (ii) at follow-up time poi nts (FU 1: 3-18 months; FU 2 \n> 18 months) – by two independent radiologists. Sar copenia was defined by \npreviously published height-corrected TPMT cut-offs  at the level of L3 (men < \n12mm/m; women < 8mm/m). \nResults or Findings: 199 patients were included (mean age 55.7±11 years,  \n69.3% male; median MELD: 11 (9-18) ; history of var iceal bleeding 33.2% and \nascites 79.9%). FU imaging was available in 70 pati ents (FU 1) and 57 patients \n(FU 2), respectively. Interrater reliability of pso as muscle measurements was \nexcellent (κ=0.985). Sarcopenia was highly prevalent in patient s undergoing \nTIPS (42.7%, 85/199). 4/45 (8.9%) of non-sarcopenic  patients at baseline \ndeveloped sarcopenia after TIPS implantation at FU 1 and 5/35 (14.3%) at FU \n2, respectively. Transplant-free survival (TFS) of patients without sarcopenia \nwas favourable (median: 59 months, IQR: 15 months –  ∞), while mortality was \nhigh among patients with sarcopenia at baseline (TF S: median 21, IQR 1.8 – \n106 months; p < 0.001). Resolution of sarcopenia wa s linked to a significant \nsurvival benefit (HR 0.121; p<0.001; adjusted to ag e and MELD score). \nConclusion: Sarcopenia is prevalent in TIPS patients and linked  to worse \noutcome. \nLimitations: CT-based diagnosis of sarcopenia without functional  testing. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Approval of local ethics \ncommittee (Medical University of Vienna) was obtain ed: EK 1760/2014, 14-\n264-VK, EK 1943/2017. \nAuthor Disclosures:  \nMattias Mandorfer: Speaker: AbbVie, Gilead, Collect ive Acumen, and W. L. \nGore & Associates, Takeda Advisory Board: AbbVie, G ilead, Collective \nAcumen, and W. L. Gore & Associates, Takeda \nTheresa Müllner-Bucsics: Nothing to disclose \nMichael Trauner: Speaker: Bristol-Myers Squibb (BMS ), Falk Foundation, \nGilead, Intercept and Merck Sharp & Dohme (MSD) Adv isory Board: from \nAbbvie, Albireo, Boehringer Ingelheim, BiomX, Falk Pharma GmbH, GENFIT, \nGilead, Hightide, Intercept, Janssen, MSD, Novartis , Phenex, Pliant, Regulus \nand Shire; travel grants from AbbVie, Falk, Gilead,  Intercept and Janssen; and \nresearch grants from Albireo, Alnylam, CymaBay, Fal k, Gilead, Intercept, MSD, \nTakeda and Ultragenyx \nKatharina Lampichler: Nothing to disclose \nJakob Kittinger: Nothing to disclose \nThomas Reiberger: Speaker: AbbVie, Bayer, Boehringe r Ingelheim, Gilead, \nIntercept, MSD, Roche, Siemens, and W. L. Gore & As sociates \nFlorian Wolf: Nothing to disclose \nLukas Hartl: Nothing to disclose \nLukas Reider: Nothing to disclose \n \n \n14:00-15:30 Room G1 \nResearch Presentation Session: \nRadiographers \nRPS 2014 \nGreen innovations: radiographers \npioneering sustainable healthcare \n \nModerators \nG. D'Anna; Milan/IT  \n(gennaro.danna@gmail.com) \nM. F. McEntee; Cork/IE \n(mark.mcentee@ucc.ie) \n \n \nSustainability in medical imaging and radiotherapy education and \npractice: a survey of the student perspectives in a  Portuguese Allied \nHealth School \n*J. M. Saude*, N. Adubeiro, L. Nogueira, I. Ribeiro , A. Xavier, C. Carvalhais; \nPorto/PT \n(miguelsaude@ess.ipp.pt) \n \nPurpose or Learning Objective: Higher education institutions play an \nimportant role as they prepare the professionals, d ecision-makers, and \ndemocratic citizens of the future. In Allied Health  education, it is crucial to \ndevelop a professional that, in addition to the tec hnical and clinical content, \nalso have a critical sense regarding environmental and social aspects. This \nstudy aims to survey medical imaging and radiothera py (MIR) students’ \nperceptions and knowledge about environmental susta inability in MIR \neducation and practice. \nMethods or Background: A cross-sectional study was carried out, using a \nself-designed questionnaire, partially adapted and translated to Portuguese \nfrom previous studies. The final version of the que stionnaire, after a pilot test, \nincluded twenty-six questions and was distributed a mongst MIR bachelor’s \ndegree students attending 1st, 2nd, 3rd and 4th aca demic years in September \nand October of 2024. \nResults or Findings: A total of 175 students participated in the study. Results \nshowed that almost everyone believes they possess g eneral knowledge about \nSustainability, however 23,0% demonstrate a lack of  awareness about the \nnegative impact that clinical practice has on the e nvironment. Data also \nrevealed that students considered there is a need f or more teaching about \nsustainability in the course (49,0%). In general, p articipants demonstrated little \nknowledge about sustainable practices in the profes sion (34,0%), although, in \ndaily life reported high adoption of sustainable pr actices (85,0%). Strong \nconnection between their course and SDGs 3 (91,4%),  4 (53,7%), and 8 \n(38,8%), was reported. \nConclusion: Our findings revealed that there is room for curric ula adjustments. \nEarly exposure to sustainability concepts has the p otential to promote the \ndevelopment of environmentally conscious profession als. By integrating \nsustainability into their education will potentiall y increase the knowledge and \nskills to make informed decisions that can reduce t he negative environmental \nimpacts of clinical practices. \nLimitations: No limitations were identified. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is educational and \nreceived a favourable statement from the institutio nal Data Protection Officer. \nAuthor Disclosures:  \nLuisa Nogueira: Nothing to disclose \nNuno Adubeiro: Nothing to disclose \nJose Miguel Saude: Nothing to disclose \nInês Ribeiro: Nothing to disclose \nCarlos Carvalhais: Nothing to disclose \nAna Xavier: Nothing to disclose \n \n \nEuropean Radiology Departments: Radiographers' Pers pectives on \nEnvironmental and Energy Sustainability \n*A. Roletto*¹, D. Catania¹, L. Rainford², A. Savio³ , M. Zanardo¹, G. R. Bonfitto¹, \nS. Zanoni³; ¹Milan/IT, ²Dublin/IE, ³Brescia/IT \n(rolettoandrea@yahoo.it) \n \nPurpose or Learning Objective: Energy consumption of radiology equipment, \nlife cycle of consumables, waste generation, and CO 2 emissions caused by \nstaff travel are factors that influence the environ mental impact of radiology \ndepartments. Through an international survey, the p erception and knowledge \n\n \n \nSaturday \nAbstract-based Programme \n \n 243  \nof European radiographers on environmental sustaina bility issues was \ninvestigated. \nMethods or Background: In March 2024, an online survey was developed \nand shared with radiographers and therapeutic radio graphers across Europe. \nThe survey consisted of 43 questions, covering demo graphic information and \nparticipants' views and actions related to environm ental sustainability, energy \nconsumption and waste generation in healthcare. \nResults or Findings: In total, 253 responses were collected from 27 Euro pean \ncountries. Most respondents highlighted the importa nce of environmental \nsustainability in healthcare. According to 63.6% (n =161) of participant, the \nmain source of environmental impact in radiology co mes from energy \nconsumption of radiological equipment. Additionally , 44.7% (n=113) suggested \nthat conducting diagnostic exams remotely could red uce the CO2 emissions \ncaused by staff commuting. Regarding workplace prac tices, over 70% (n=192) \nreported that they turn-off devices after use. Amon g the possible obstacles to \nturning off radiological equipment, respondents ide ntified long shutdown and/or \nrestart times, loss of clinical/technical data and non-applicability to the \nradiographers’ role. Although recycling is practice d, it is limited to paper, \nplastic, and glass. Lack of environmental sustainab ility strategies in the \nworkplace was reported by 66% (n=167). Meanwhile, 9 6.1% (n=243) believe \nthat radiographers could play an active role in pro moting environmental \nsustainability in their departments. \nConclusion: This study offers a comprehensive analysis of Europ ean \nradiographers' awareness and opinions on environmen tal sustainability. \nAlthough radiographers recognize the importance of creating eco-friendly \nradiology departments, there are still gaps in impl ementation of sustainable \npractices within radiology practices. \nLimitations: Potential biases may have been introduced in relati on to the \ncomplexity of the topics covered. \nFunding for this study: Not applicable \nEthics committee - additional information: Ethical exemption was granted \nby the host institution University College Dublin f or this anonymous survey (LS-\nLR-24-25-Catania-Rainford) \nAuthor Disclosures:  \nAnna Savio: Nothing to disclose \nDiego Catania: Nothing to disclose \nLouise Rainford: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nMoreno Zanardo: Nothing to disclose \nGiuseppe Roberto Bonfitto: Nothing to disclose \nSimone Zanoni: Nothing to disclose \n \n \nState-of-Play in Artificial Intelligence Sustainabi lity in Medical Imaging: a \nscoping review \nM. Champendal¹, B. Lokaj², J. Zaghir³, V. Durand De  Gevigney², C. Lovis³,  \nH. Müller⁴, J. Schmid², *R. S. T. Ribeiro*¹; ¹Lausanne/CH, ²G enève/CH, \n³Geneva/CH, ⁴Sierre/CH \n(ricardo.ribeiro@hesav.ch) \n \nPurpose or Learning Objective: To synthesize the existing literature on how \nthe environmental sustainability of artificial inte lligence (AI) in medical imaging \nis being addressed and to identify specific strateg ies that have been used. \nMethods or Background: A scoping review was conducted following the \nJoanna Briggs Institute methodology. Comprehensive literature search was \nperformed in MEDLINE, Embase, CINAHL, and Web of Sc ience, targeting \npublications from 2014 to 2024 in English or French . The search used a \ncombination of keywords and MeSH terms related to e nvironmental \nsustainability, AI, and medical imaging modalities.  Three independent \nreviewers screened abstracts, titles and full texts  for eligibility. \nResults or Findings: The search identified 2812 results, of which 11 met  the \ninclusion criteria. The selected papers comprised 8  research articles, 3 \nreviews. Three key themes emerged: energy consumpti on (n=10), carbon \nfootprint (n=4), and computational resources (n=4).  The metrics CO2 \nequivalent, carbon intensity, training time, power use effectiveness, equivalent \ndistance travelled by car were proposed to assess p otential AI impact on the \nenvironment. Most energy-efficient techniques invol ved data, AI modelling and \ntraining such as data augmentation, data quantisati on, lightweight model \ndevelopment, reduction of parameters. Identified st rategies to enhance \nefficiency and reduce environmental impact include (i)integrating energy and \ncarbon metrics in AI evaluation in addition to accu racy assessments, \n(ii)developing an ecolabel for AI tools, (iii)transitioning to cloud computing and \n(iv)developing lightweight AI models. \nConclusion: This review identified critical metrics and actiona ble strategies \nused to assess and improve sustainable practices in  AI for medical imaging \nwhich include the integration of specific sustainab ility-related metrics, cloud \ncomputing adoption and development of efficient AI models. \nLimitations: The limitations of this review include not assessin g the quality of \nthe articles, which is standard practice in scoping  reviews, and restricting the \nsearch to only four databases. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nRicardo Silva Teresa Ribeiro: Nothing to disclose  \nJamil Zaghir: Nothing to disclose \nHenning Müller: Nothing to disclose \nChristian Lovis: Nothing to disclose \nJerome Schmid: Nothing to disclose \nValentin Durand De Gevigney: Nothing to disclose \nBelinda Lokaj: Nothing to disclose \nMélanie Champendal: Nothing to disclose \n \n \nA life-cycle assessment framework for quantifying t he carbon footprint of \ndiagnostic imaging \n*A. Roletto*¹, A. Savio², B. Marchi², S. Zanoni²; ¹ Milan/IT, ²Brescia/IT \n(rolettoandrea@yahoo.it) \n \nPurpose or Learning Objective: Environmental sustainability topic is \nincreasingly relevant in the radiology sector, whic h accounts for about 10% of \nhealthcare sector's carbon footprint. Life-cycle as sessment (LCA) is one of the \nmain tools for analysing the environmental impact o f processes. This study \naims to review existing LCA approaches in radiology  available in literature, \nidentify their characteristics and limitations to p ropose a more comprehensive \nLCA framework for diagnostic imaging. \nMethods or Background: Through a literature review on the topic of LCA in \ndiagnostic imaging, performed according to the PRIS MA statement, a novel \nLCA framework specifically for diagnostic imaging w as developed. The \nframework includes all the features described by th e various selected articles \nand overcoming their limitations. \nResults or Findings: Regarding the literature review, an LCA approach wa s \ndescribed by 5 articles. All studies electricity co nsumption in their framework. \nUsage of consumables and auxiliary equipment was in cluded in 80% (4/5) of \nthe studies. Equipment production, staff commuting,  and waste generation \nwere included in only 40% (2/5). Only two articles have considered a cradle-to-\ngrave system approach. Subsequently, a novel LCA fr amework was designed \nto overcome the limitations by covering all aspects  that can be included in the \nthree phases: 1) Preclinical, production of imaging  equipment and \nconsumables, staff commuting; 2) Clinical, use of d evices and consumables \nduring diagnostic imaging; 3) Post-clinical, waste generation and data storage \nfor the images produced. \nConclusion: This literature review provides an overview of stud ies focusing on \nLCA methodology for diagnostic imaging. The results  culminated in the \nproposal of a comprehensive framework, which aims t o overcome the identified \nlimitations by providing a complete analysis of the  environmental footprint of \nradiological procedures. \nLimitations: Still sparse reporting of LCA in Diagnostic imaging  limited the \nstrength of the conclusions of this study. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nAnna Savio: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nBeatrice Marchi: Nothing to disclose \nSimone Zanoni: Nothing to disclose \n \n \nLeading Radiographers into a Greener Future: A Syst ematic Review of \nGreen Transformational Leadership and Sustainable P ractices \n*L. Federico*¹, A. Roletto², D. Catania², S. Zanoni ³, S. Durante¹; ¹Bologna/IT, \n²Milan/IT, ³Brescia/IT \n(laura.federico01@gmail.com) \n \nPurpose or Learning Objective: The healthcare system requires many \nresources to maintain the standards necessary to ca re for the population. It is \nthe task of effective leadership to ensure the envi ronmental sustainability \nclinical activities. The aim of this study is to ex amine the role of Green \nTransformational Leadership (GTL) in radiography, f ocusing on the mitigation \nstrategies adopted by radiographer managers and the ir impact on promoting \nenvironmental sustainability in radiology departmen ts. \nMethods or Background: A systematic literature was conducted in \naccordance with the PRISMA statements in several da tabases, targeting \nstudies on leadership in radiography, green innovat ion, and environmental \nresponsibility. The search strategy employed keywor ds such as \"green \ntransformational leadership,\" \"radiographer manager ,\" and \"environmental \nleadership in healthcare.\" Two reviewers independen tly screened titles and \nabstracts to select studies based on inclusion crit eria. \nResults or Findings: Among the retrieved articles, 6 met the inclusion c riteria. \nThe analysis highlighted how GTL can influence the implementation of \nsustainable practices in radiography. Radiographer managers using GTL play \na key role in promoting sustainable practices, such  as reducing energy \nconsumption, encouraging use of reusable imaging eq uipment, and minimising \nwaste generation. Interdisciplinary collaboration c ould be decisive in increasing \neffectiveness of green initiatives. Studies reviewe d highlighted the importance \n\n \n \nSaturday \nAbstract-based Programme \n \n 244  \nof successful communication, measurable goals, and staff training for adopting \ngreen practices. Radiography service managers with emotional intelligence \nwho fostered a climate of trust were more successfu l in promoting a culture of \nsustainability. \nConclusion: Radiography leaders who prioritise green strategies  can \nsignificantly reduce the environmental impact of ra diology departments. \nInvesting in leadership development, promoting emot ional intelligence, and \nencouraging interdisciplinary collaboration are cri tical steps in creating a green \nhealthcare system that aims towards environmental s ustainability without \nreducing quality patient care. \nLimitations: Sparce literature of GTL limit the strength of the conclusions of \nthis study. \nFunding for this study: Not needed \nEthics committee - additional information: Not needed \nAuthor Disclosures:  \nDiego Catania: Nothing to disclose \nStefano Durante: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nSimone Zanoni: Nothing to disclose \nLaura Federico: Nothing to disclose \n \n \nSustainable practices in nuclear medicine: A scopin g review \n*D. Fonseca Ribeiro*¹, K. Borg Grima², A. Geão³, C.  Andersson⁴, S. Murphy⁵, \nP. S. Costa⁶, C. Baun⁷, A. Karangelis⁸, M. Champendal⁹; ¹London/UK, \n²Naxxar/MT, ³Montijo/PT, ⁴Uppsala/SE, ⁵Dublin/IE, ⁶Porto/PT, ⁷Odense/DK, \n⁸Patra/GR, ⁹Lausanne/CH \n \nPurpose or Learning Objective: Sustainable development seeks to balance \neconomic growth, environmental impact, and social i nclusion, which is critical \nin high-tech fields like nuclear medicine. This stu dy reviewed published \nliterature on sustainable practices in nuclear medi cine, focusing on the three \nmain pillars of sustainability: environmental, soci al, and economic. \nMethods or Background: The scoping review was conducted in accordance \nwith the Joanna Briggs Institute methodology. The s earch was performed on \nPubMed, Embase, Cinhal and Web of Science in Novemb er 2023 and included \nstudies in English. The research equation combined keywords and Medical \nSubject Heading terms (MeSH) related to sustainabil ity in Nuclear Medicine. \nThree independent review authors screened all abstr acts and titles, and four \nreviewers conducted the data extraction and analysi s. \nResults or Findings: Thirty-two studies met the inclusion criteria for t his \nscoping review, with most articles having been publ ished in 2022 and 2023. \nSpain contributed the highest number of publication s. Studies were \ncategorised according to procedure type (n = 32; 31 % therapy, 9% PET, 35% \ndiagnostic & therapy, 19% SPECT) and study location  (hospital based, non-\nhospital based, or both). The studies primarily foc used on strategies related to \nthe three sustainability pillars, with an emphasis on the environmental impact. \nConclusion: This review highlighted the growing interest in sus tainability \nwithin the nuclear medicine field, especially in re lation to environmental factors. \nHowever, significant knowledge gaps emerged on the impact of economic and \nsocial factors within nuclear medicine practices. R ecommended strategies \nincluded proper radioactive waste management, resou rce optimisation, and \nfostering collaborative environments to ensure sust ainable nuclear medicine \npractices. \nLimitations: The limitations of the study are the exclusion of a rticles not \navailable in the English language and the quality o f the articles included was \nnot assessed according to the methodology of a scop ing review. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: This study was a scoping review. \nAuthor Disclosures:  \nDaniela Fonseca Ribeiro: Nothing to disclose \nKaren Borg Grima: Nothing to disclose \nApostolos Karangelis: Nothing to disclose \nAna Geão: Nothing to disclose \nCamilla Andersson: Nothing to disclose \nChristina Baun: Nothing to disclose \nPedro Silva Costa: Nothing to disclose \nShauna Murphy: Nothing to disclose \nMélanie Champendal: Nothing to disclose \n \n \nTowards sustainable radiography: exploring percepti ons and overcoming \nbarriers \n*M. A. Rawashdeh*¹, A. England², M. F. Mcentee², M.  Ali¹, M. E. S. Zakaria¹; \n¹Ajman/AE, ²Cork/IE \n(marawashdeh@just.edu.jo) \n \nPurpose or Learning Objective: Previous research has examined the \nattitudes and behaviors of various professions rega rding environmental \nsustainability. However, there is a paucity of stud ies specifically addressing the \nperspectives of radiographers. This study aims to i nvestigate radiographers' \nperceptions, practices, and barriers to change rela ted to environmental \nsustainability within the field of radiology. \nMethods or Background: Ethical approval was obtained and data collection \nwas conducted using Google Forms. The survey target ed 104 practicing \nradiographers across multiple countries. Questions were organized into five \ndomains to gather insights into demographics, train ing related to global \nwarming and climate change. Data analysis employed both descriptive and \ninferential statistics. \nResults or Findings: A total of 104 radiographers completed the survey. \nFemale respondents exhibited a significantly higher  participation rate in \nenvironmental protection campaigns (P=0.01). The ma jority of participants \n(68%) acknowledged their awareness of climate chang e and its impact on the \nnatural environment. Furthermore, 74% of respondent s indicated a belief in the \nnecessity to enhance sustainability practices. The most frequently employed \nstrategies to reduce energy consumption and emissio ns included low-energy \nlighting (60%), real-time power monitoring tools (4 1%), and energy-efficient \nheating systems (32%). A notable concern regarding sustainability emerged \namong respondents, with time constraints (50%) and a lack of leadership \n(48%) identified as prevalent barriers. \nConclusion: Participants recognize the significance of environm ental \nsustainability in the field of radiology; however, obstacles such as inadequate \nleadership, support, authority, and facility limita tions impede the adoption of \nsustainable practices. \nLimitations: The present study has several limitations. It relie d on voluntary \nparticipation from radiographers and employed snowb all sampling, which may \nintroduce self-selection bias. Additionally, the on line survey format may limit \ngeographical or technical access, thereby affecting  the generalizability of study \nfindings. \nFunding for this study: Not applicable \nEthics committee - additional information: IRB from Gulf Medical University \nAuthor Disclosures:  \nMark F. Mcentee: Nothing to disclose \nMohamed El Sayed Zakaria: Nothing to disclose \nMagdi Ali: Nothing to disclose \nAndrew England: Nothing to disclose \nMohammad Ahmmad Rawashdeh: Nothing to disclose \n \n \nConsiderations on the environmental sustainability of using ChatGPT in \nradiography \n*E. Scaramelli*, G. R. Bonfitto, A. Roletto, S. V. Fasulo, L. Bombelli,  \nD. Catania; Milan/IT \n(es.elena.scaramelli00@gmail.com) \n \nPurpose or Learning Objective: ChatGPT has gained credit among \nradiographers, given its contribution to training a nd clinical activity. However, \nits environmental impact is overlooked. The increas ing use of AI is contributing \nto rising Greenhouse gas emissions, water waste and  energy consumption. \nThis research aims to explore the environmental imp act of using ChatGPT in \nradiography and raise awareness of its sustainabili ty in daily clinical practice, \nunderscoring the need for radiographers to engage, given the gap in AI's \nenvironmental impact studies within radiology. \nMethods or Background: A literature review across databases (Google \nScholar, PubMed) was conducted according to the PRI SMA statement to find \nrecent evidence on environmental impact of ChatGPT in radiography and, \nmore generally, in healthcare. \nResults or Findings: The literature shows limited evidence on the \nsustainability of AI in radiology, leading to an an alysis of the broader \nhealthcare sector to estimate its impact and develo p insights for radiology. \nFrom this search, only 16 articles were found to be  relevant to healthcare, and \njust one focused on radiology. The results reveal a  dual nature of AI: while it's \nincredibly useful in healthcare, it has a significa nt environmental impact. \nIndeed, companies like Google, Microsoft, and Meta consume 2.2 billion cubic \nmeters of water to cool their servers and produce e lectricity, marking a 48% \nincrease in water usage. GPT-3 produces 8.4 metric tons of CO2 annually and \nrequires 700,000 Liters of water for training. AI i s also responsible for 1-2% of \nglobal electricity consumption, raising important q uestions about its long-term \nsustainability. \nConclusion: Although tools such ChatGPT show usefulness in clin ical \npractice, including radiography, it’s essential to use them more responsibly, \ngiven their substantial environmental impact on pla net's resources. \nLimitations: The lack of evidence in the literature, limits the strength of the \nconclusions of this study. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nLuca Bombelli: Nothing to disclose \nSimone Vito Fasulo: Nothing to disclose \nDiego Catania: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nElena Scaramelli: Nothing to disclose \nGiuseppe Roberto Bonfitto: Nothing to disclose \n\n \n \nSaturday \nAbstract-based Programme \n \n 245  \nGreen Imaging Revolution: Pioneering Sustainable Pr actices in \nRadiology for a Healthier Planet \n*A. A. Bherwani*¹, M. Sudds², L. Townsend-Sanders³,  A. Prieto Valero⁴; \n¹Orlando, FL/US, ²Wymondham/UK, ³Tampa, FL/US, ⁴Madrid/ES \n(anand.bherwani@gmail.com) \n \nPurpose or Learning Objective: This study highlights the critical need to \nincorporate sustainability in radiology, reducing t he environmental impact \nwithout compromising diagnostic accuracy or patient  care. Focusing on \nenergy-efficient practices, waste reduction, and te chnological innovations in \nmodalities like CT, MRI, and general radiography, t he objective is to integrate \neco-friendly strategies into radiological departmen ts as part of broader \nhealthcare sustainability goals. \nMethods or Background: A comprehensive literature review of over 90 \nsources, including peer-reviewed articles, books, a nd case studies, was \nconducted. This research explores innovations such as energy-saving modes \nin MRI and CT, digital radiography's reduction of c hemical waste, and \nsustainable contrast media practices. Case studies from several healthcare \ninstitutes globally are analyzed, illustrating real -world applications of green \nradiology practices. \nResults or Findings: Adopting energy-efficient imaging systems and \noptimizing resource use led to significant environm ental and financial benefits. \nFor example, Singapore General Hospital reduced ene rgy consumption by \n30%, saving $1.2 million annually, while Mayo Clini c enhanced wastewater \ntreatment, minimizing environmental contamination f rom contrast media. These \npractices and many others from various other instit utions not only reduced \ngreenhouse gas emissions but also operational costs . \nConclusion: Incorporating sustainable practices in radiology is  vital for \nreducing healthcare’s environmental footprint while  maintaining high standards \nof patient care. This research provides a scalable framework for radiology \ndepartments worldwide, illustrating how integrating  sustainability can enhance \noperational efficiency and contribute to global hea lth equity and environmental \nconservation. \nLimitations: This study is based on case studies from select ins titutions, which \nmay not reflect the diverse resources and regulator y environments of all \nhealthcare systems. Additionally, the long-term imp acts of implementing these \nsustainable radiology practices require further inv estigation. \nFunding for this study: Not Applicable \nEthics committee - additional information: Not Applicable \nAuthor Disclosures:  \nLaurie Townsend-Sanders: Nothing to disclose \nAlicia Prieto Valero: Nothing to disclose \nAnand Ashok Bherwani: Employee: GE Healthcare Pharm aceutical \nDiagnostics \nMichael Sudds: Nothing to disclose \n \n \nCentargo: Insights from the CIMROD Experiment on Co ntrast \nConsumption, Efficiency, Patient Care and environme ntal impact \nY. Anquetil¹, *T. Leturgez*², J. E. Jacquin¹, F. Kr uta³, F. Jambon¹; \n¹Perigueux/FR, ²La Garenne Colombes/FR, ³Paris/FR \n(thibaut.leturgez@bayer.com) \n \nPurpose or Learning Objective: With the transition in France's contrast \nproduct supply model on March 1, 2024, the function alities of CT injectors have \nbeen re-evaluated. This study assesses the impact o f Centargo injectors on \ncontrast media volume, patient care, preparation ti me, and environmental \nimpacts at the CIMROD center, which acquired two Ce ntargo injectors. \nMethods or Background: To collect pertinent data, we conducted three \nmeasurements and utilized five data sources. The pr ivate hospital, equipped \nwith two CT machines, transitioned from a single-us e Stellant injector to a \nmulti-use system, ultimately adopting the Centargo injector with enhanced \nfeatures. Data sources included the volume of contr ast media administered per \npatient, the time radiographers allocated to variou s tasks in their routines, \nprotocol preparation and traceability times, and wa ste weight measurements. \nThe evaluation period spanned from April to Septemb er and incorporated \ndifferent injector configurations as well as a tagg ing system, Ubudu, to monitor \nworkflow efficiency among radiographers. \nResults or Findings: CIMROD aimed for an average contrast volume of 80 \nmL using Ultravist 370. The Centargo injector succe ssfully achieved an \naverage of 72.08 mL per injection across 1,099 proc edures, and after further \noptimization, this volume decreased to 68.31 mL, re sulting in a 15% reduction \ncompared to the initial target. Workflow improvemen ts included a 50% \nreduction in injector preparation time, a 14% incre ase in patient care time, and \na 6% increase in image preparation time. Time savin gs of 16% per exam were \nnoted, with waste production decreasing by 66% comp ared to the single-use \nStellant, equating to an annual reduction of 1.4 to nnes. \nConclusion: The findings confirm the benefits of the Centargo i njector, \nespecially when combined with Smart Protocol and Na utilus software, RIS-\nInjector connectivity, and kVp optimization, leadin g to enhanced operational \nefficiency and reduced environmental impact. \nLimitations: Single center evaluation. \nFunding for this study: Bayer funded \nEthics committee - additional information: The experimentation isn't \ncollecting patient level data so we didn't submitte d it to an ethical committee. \nAuthor Disclosures:  \nJerome Elisabeth Jacquin: Nothing to disclose \nFrançois Kruta: Equipment Support Recipient: Ubudu \nYohan Anquetil: Equipment Support Recipient: CIMROD  \nThibaut Leturgez: Employee: Bayer \nFrançois Jambon: Equipment Support Recipient: CIMRO D \n \n \nSuccess factors for implementing an intervention us ing return letters for \nlow-value MRIs \n*I. Ø. Brandsæter*, E. Kjelle, E. R. Andersen, B. M . Hofmann; Gjøvik/NO \n(Ingrid.o.brandsater@ntnu.no) \n \nPurpose or Learning Objective: The study aimed to investigate key \nstakeholders’ experiences with and reflections on s uccess factors for \nimplementing an intervention using return letters t o reduce the use of three \nlow-value magnetic resonance imaging (MRI) examinat ions. \nMethods or Background: An intervention to reduce low-value MRI was \ndesigned and implemented in private imaging centres  in Norway in October \n2022. The intervention used return letters based on  Choosing Wisely \nrecommendations for poor referrals for MRI of the l ower back, brain and knee \nsent to imaging centres. Two semi-structured indivi dual interviews were \nconducted with the medical directors of the two inc luded imaging providers \n(radiologists) and two focus group interviews with nine managers from the \nvarious private imaging centres (radiographers) ope rated by the two imaging \nproviders were conducted. Inductive content analysi s in three steps was used \nto analyse the data. \nResults or Findings: The analysis resulted in five categories: general \nexperience, anchoring, organisation, return letter procedure and outcome. In \ngeneral, the intervention was well received. Suffic ient information, anchoring \nand support from the organisation’s leaders were id entified as crucial success \nfactors. However, there were some barriers to the i mplementation, e.g. the \nmedical directors in charge of the implementation f ound it hard to be hands-on \nand distribute information to the radiogpraphers at  the imaging centres. \nAdditionally, some Choosing Wisely recommendations were found vague and \ndifficult to use by the radiographers doing the ref erral assessment. \nConclusion: This study provides insights into the practical and  crucial details \nof implementing interventions to reduce low-value i maging using Choosing \nWisely recommendations. The intervention was genera lly well received, and \nseveral key success factors were identified. \nLimitations: The limitations of the study are that only managers  and medical \ndirectors were included in the study, interviewing referrers and patients would \nhave provided other important perspectives. \nFunding for this study: Funding was provided by the Research Council of \nNorway (Project number 302503). \nEthics committee - additional information: Ethical approval is unnecessary \naccording to national regulations in Norway (LOV-20 08–06-20–44), and this \nstudy was not submitted to the regional committees for medical and health \nresearch ethics. The Norwegian Agency for Shared Se rvices in Education and \nResearch approved the processing and storage of per sonal information in this \nstudy (Ref. 974188) \nAuthor Disclosures:  \nIngrid Øfsti Brandsæter: Nothing to disclose \nBjørn Morten Hofmann: Nothing to disclose \nElin Kjelle: Nothing to disclose \nEivind R. Andersen: Nothing to disclose \n \n \nQuantifying Energy Savings in Radiology: A Simple A pproach to Make \nthe Radiology Department More Sustainable \n*K. Iaccarino*, D. Fazzini, S. Papa, M. Ali; Milan/ IT \n(iaccarinokatia@gmail.com) \n \nPurpose or Learning Objective: To quantify the energy consumption of \nradiology reporting stations and explore a hypothet ical scenario to mitigate \nenergy waste in a healthcare setting. \nMethods or Background: We monitored 10 reporting stations over a period of  \n90 days at the Centro Diagnostico Italiano (Milan, Italy). An energy logger was \ninstalled on each station to measure real-time powe r consumption, capturing \ndata on both active and idle states. Stations were configured to enter stand-by \nmode after 4 hours of inactivity. We conducted a si mulation to assess the \nimpact of shutting down the stations after 1 hour o f inactivity instead of allowing \nthem to remain in stand-by mode. \nResults or Findings: The overall power consumption of the 10 reporting \nstations was approximately 16,615.65 kWh, which is equivalent to about 6 \nhouseholds in Italy (average annual consumption of 2,700 kWh per \nhousehold). We identified three main power consumpt ion patterns: mainly-off, \nmainly-on, and always-off. The estimated on-mode co nsumption was \n12,738.44 kWh per year, while stand-by consumption was 3,128.13 kWh and \n\n \n \nSaturday \nAbstract-based Programme \n \n 246  \noff-mode consumption was 749.06 kWh. By implementin g the hypothetical \nscenario of shutting down after 1 hour of inactivit y, we estimated potential \nenergy savings of around 2,346.12 kWh/year. Conside ring that CDI has a total \nof 53 reporting stations across all sites, the estimated total annual energy \nsavings would be approximately 12,434.44 kWh. \nConclusion: Optimizing energy usage in radiology departments is  essential for \npromoting sustainability. Simple configuration chan ges can lead to significant \nreductions in energy waste, enhancing the environme ntal responsibility of \nhealthcare facilities. \nLimitations: The sample size of 10 reporting stations may not be  \nrepresentative of the entire radiology department, and the 90-day monitoring \nperiod might not account for seasonal variations in  usage. The hypothetical \nscenario assumes uniform behavior across stations, which may not reflect \nactual workflows. \nFunding for this study: None \nEthics committee - additional information: The ethics commettee approval \nis not applicable for this study \nAuthor Disclosures:  \nMarco Ali: Consultant: Bracco Imaging S.p.A. \nSergio Papa: Nothing to disclose \nDeborah Fazzini: Nothing to disclose \nKatia Iaccarino: Nothing to disclose \n \n \n16:00-17:30 Research Stage 1 \nResearch Presentation Session: Head and \nNeck \nRPS 2108 \nA journey through thyroid imaging \n \nModerator \nE. Gotsiridze; Tbilisi/GE  \n(gotsiridze.elene@gmail.com) \n \n \nAdditional Value of Pertechnetate Scintigraphy to A CR-TIRADS and EU-\nTIRADS for Thyroid Nodule Classification in Euthyro id Patients \nL. Sollmann, M. Eveslage, M. Danzer, M. Schäfers, B . Heitplatz, D. Hescheler, \nB. Riemann, *B. Noto*; Münster/DE \n(benjamin.noto@ukmuenster.de) \n \nPurpose or Learning Objective: Thyroid nodules are a highly prevalent, \npredominantly benign finding, yet their accurate ev aluation remains \nchallenging. While ultrasound and TIRADS are now wi dely accepted as \nstandard in evaluation, the utility of thyroid scin tigraphy in euthyroid patients \nremains debated. Previous studies have investigated  the diagnostic potential of \nTIRADS or radionuclide scanning in isolation, but a n integrated approach has \nnot been explored so far. This study aimed to evalu ate if pertechnetate \nscintigraphy enhances the diagnostic value of TIRAD S in a multimodal \nframework. \nMethods or Background: The diagnostic capabilities of ACR-TIRADS, EU-\nTIRADS, pertechnetate scintigraphy, and multimodal models were \nretrospectively analyzed for 322 nodules (231 benig n, 91 malignant) in 208 \neuthyroid patients undergoing thyroidectomy. Statis tical analysis employed \ngeneralized estimating equations. \nResults or Findings: Thyroid scintigraphy demonstrated an AUC of 0.6 \n(95%CI:0.55-0.66), ACR-TIRADS of 0.84 (95%CI:0.79-0 .89) and EU-TIRADS \nof 0.78 (95%CI: 0.72-0.83). Integrating thyroid sci ntigraphy with ACR-TIRADS \nenhanced diagnostic accuracy, yielding an AUC of 0. 86 (p=0.039). Similarly, \ncombining thyroid scintigraphy with EU-TIRADS resul ted in an AUC of 0.80 (p \n=0.008), surpassing the individual TIRADS performan ces. Furthermore, the \nintegration of thyroid scintigraphy adjusted the ma lignancy probability among \nTIRADS categories. Iso- or hyperfunctioning nodules  in ACR-TIRADS TR4 and \nhypofunctioning nodules in TR3 exhibited comparable  probabilities of \nmalignancy. Similarly, iso- or hyperfunctioning nod ules in EU-TIRADS 4 \nshowed similar malignancy probabilities to hypofunc tional nodules in EU-\nTIRADS 3, indicating the potential for more refined  risk stratification. \nConclusion: This study demonstrates enhanced diagnostic perform ance \nachieved by integrating thyroid scintigraphy with A CR—and EU-TIRADS for \nclassifying thyroid nodules in euthyroid patients. Such a multimodal approach \ncould improve risk stratification and management de cisions, particularly in \ncomplex scenarios like multinodular goiter. Further  research is warranted to \nvalidate these findings and explore their clinical implications. \nLimitations: Retrospective design \nFunding for this study: The Medical Faculty, University of Münster, Germany  \nsupported B.N. as a clinician scientist. \nEthics committee - additional information: The study protocol was approved \nby the ethics committee of the University of Munste r and performed in \naccordance with the ethical standards as laid down in the 1964 Declaration of \nHelsinki and its later amendments. \nAuthor Disclosures:  \nMichael Schäfers: Nothing to disclose \nMoritz Danzer: Nothing to disclose \nBarbara Heitplatz: Nothing to disclose \nLea Sollmann: Nothing to disclose \nBurkhard Riemann: Nothing to disclose \nDaniel Hescheler: Nothing to disclose \nBenjamin Noto: Nothing to disclose \nMaria Eveslage: Nothing to disclose \n \n \nSubtraction ultrasound microangiography for assessm ent of \nmicrovascularity patterns in diffuse thyroid diseas e \n*A. Borlea*, D. I. Stoian; Timisoara/RO \n(andreea.brl3@gmail.com) \n \nPurpose or Learning Objective: To evaluate the use of Subtraction \nUltrasound Microangiography (SUMA) in the quantitat ive assessment of \nmicrovascularity patterns in diffuse thyroid diseas es and to compare these \nfindings with normal thyroid tissue. \nMethods or Background: Conventional color Doppler ultrasound is limited by  \nits subjective nature in assessing thyroid vascular ity. SUMA provides a \nquantitative approach through color pixel percentag e (CPP) measurement. A \ncohort of 220 subjects was studied, including 90 wi th autoimmune thyroiditis, \n20 with Graves' disease, and 110 normal controls. S UMA was used to quantify \nCPP, and results were analyzed to identify differen ces between the groups. \nThe relationship between TSH levels and CPP was als o explored. \nResults or Findings: Significant differences in CPP were observed across  the \ngroups. Normal controls (euthyroid) demonstrated a median CPP of 26% (IQR \n18-37%), Graves' disease patients with clinical and  subclinical hyperthyroidism \nhad a median CPP of 75% (IQR 55-82%), and those wit h untreated \nhypothyroid Hashimoto thyroiditis (hypothyroid) had  a median CPP of 63% \n(IQR 49-75%). A U-shaped relationship was found bet ween TSH levels and \nCPP, indicating alterations in vascularity in both hypo- and hyperthyroid states. \nConclusion: SUMA allows for a more objective and quantitative a ssessment \nof thyroid microvascularity, revealing distinct dif ferences between normal, \nautoimmune thyroiditis, and Graves' disease tissues . The technique may \nimprove diagnostic accuracy compared to conventiona l Doppler ultrasound. \nLimitations: Further studies are needed to validate SUMA’s diagn ostic \npotential, including its reproducibility across dif ferent ultrasound systems and \nsettings, as well as its sensitivity in detecting disease progression or treatment \nresponse. \nFunding for this study: The study received no external funding \nEthics committee - additional information: Victor BAbes University of \nMedicine and Pharmacy \nAuthor Disclosures:  \nDana I Stoian: Nothing to disclose \nAndreea Borlea: Nothing to disclose \n \n \nUltrasound-guided microwave ablation versus thyroid ectomy for the \ntreatment of solitary nodular retrosternal goiter \n*Y. Li*, Y. Luo, M. Zhang; Beijing/CN \n(Lyy19900325@sina.com) \n \nPurpose or Learning Objective: The purpose of this study was to compare \nthe clinical outcomes of microwave ablation versus thyroidectomy for patients \nwith solitary nodular retrosternal goiter (RSG). \nMethods or Background: This retrospective study evaluated 243 patients wit h \nsolitary nodular RSG treated with MWA (M group, n=1 11) or thyroidectomy (T \ngroup, n=132). Complications, thyroid function, and  treatment variables, \nincluding procedure time, estimated blood loss, hos pitalization, and cost, were \ncompared. The volume, volume reduction rate, sympto ms, and cosmetic score \nwere also evaluated in the W group. \nResults or Findings: The rate of overall complications, transient RLN in jury \nand hypothyroidism was 35.6%, 9.1%, and 10.6% in th e T group, respectively, \nbut these complications were 7.2%, 2,7%, and 0 in t he M group (all P<.05). \nPatients in the M group had a significantly shorter  procedure time (median, \n6.48 versus 95.0 minutes, P<.0001), less estimated blood loss (0 versus 50 ml, \nP<.0001), and lower cost (US ＄1541.25 versus ＄2839.40, P<.0001) than \nthose treated by thyroidectomy. After MWA, the VRR was 76.31% and 89.37% \nat the 12 months and last follow-up time, respectiv ely. Of all the nodules \ntreated by MWA, 50 (45%) received additional ablati on. The symptom and \ncosmetic scores were both significantly reduced at the last follow-up. \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 247  \nConclusion: MWA is an effective and safe treatment for solitary  nodular RSG. \nMoreover, MWA is associated with a faster recovery,  fewer complications, and \nsuperior esthetic results relative to thyroidectomy  and it may be a potential \nalternative to surgery in selected patients, especi ally for those who are \nineligible or unwilling to receive surgical treatme nt. \nLimitations: This study was carried out in a single center, and multicenter \nstudies are warranted. \nFunding for this study: This study recieved no funding. \nEthics committee - additional information: Ethical approval was obtained \nfrom the Institutional Ethics Committee of the Chin ese PLA General Hospital. \nAuthor Disclosures:  \nYingying Li: Nothing to disclose \nMingbo Zhang: Nothing to disclose \nYukun Luo: Nothing to disclose \n \n \nMid-term thyroid function alterations as predictors  of long-term \noutcomes in radiofrequency ablation of benign thyro id nodules \n*Y-H. Chen*¹, P-L. Chiang¹, Y-H. Chang², C-K. Chou² , W-C. Lin¹; \n¹Kaohsiung/TW, ²Kaohsiung City/TW \n(yh19970718@gmail.com) \n \nPurpose or Learning Objective: Radiofrequency ablation (RFA) has gained \nrecognition as a highly effective treatment for ben ign thyroid nodules. While \nthyroid function alterations have been observed dur ing post-RFA follow-up, this \nstudy specifically focuses on the potential link be tween changes in thyroid \nfunction and volume reduction ratio of treated nodu les. Additionally, it seeks to \nevaluate whether fluctuations in thyroid function a t mid-term follow-up can \nserve as early indicators for the development of lo ng-term hypothyroidism \nfollowing RFA. \nMethods or Background: In this retrospective study, 50 euthyroid individua ls \n(mean age = 47.0 years; 43 females, 7 males) with a  total of 72 benign thyroid \nnodules (median volume = 4.61 mL) undergoing RFA we re evaluated. \nComprehensive assessments, including clinical exami nations, ultrasound \nimaging, and blood tests, were conducted at specifi c intervals (pre-RFA, and at \n6 months, 12 months, and annually post-RFA). \nResults or Findings: The mean follow-up period was 22.3 months. Both \nmedium-term and long-term follow-ups revealed signi ficant reductions in T3 \nlevels (p<0.001, p=0.005) and elevations in TSH lev els (p<0.001, p<0.001) \ncompared to baseline measurements. A negative corre lation was found \nbetween medium-term T3 levels and long-term volume reduction ratio (r=-\n0.332, p=0.005). Furthermore, patients with lower T 3 levels during medium-\nterm follow-up demonstrated a significantly higher long-term volume reduction \nratio compared to those with higher T3 levels (0.92  vs. 0.77, p=0.017). \nConclusion: Following radiofrequency ablation, notable unexpect ed \nalterations in thyroid function were observed, with out meeting the criteria of \nhypothyroidism. Additionally, a lower mid-term T3 l evel may be indicative of a \nbetter volume reduction ratio during long-term foll ow-up. \nLimitations: The study is limited by biases associated with its retrospective \ndesign and a lack of short-term data within a six-m onth period. To address \nthese limitations, further prospective studies with  increased focus on short-term \ndata are necessary. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study received approval \nfrom the institutional review board under the refer ence number 202200189B0. \nAuthor Disclosures:  \nWei-Che Lin: Nothing to disclose \nYen-Hsiang Chang: Nothing to disclose \nPi-Ling Chiang: Nothing to disclose \nYi-Han Chen: Nothing to disclose \nChen-Kai Chou: Nothing to disclose \n \n \nComparision of K-TIRADS, EU-TIRADS and ACR-TIRADS G uidelines for \nMalignancy Risk Determination of Thyroid Nodules \n*E. Tobcu*, E. Karavaş, G. Taşova Yılmaz, Z. Tobcu; Balıkesir/TR \n(etobcu@bandirma.edu.tr) \n \nPurpose or Learning Objective: To evaluate the performances of three \ninternationally recognized thyroid imaging reportin g and data systems \n(TIRADS) for risk stratification of malignancy in c omparison to one another. \nMethods or Background: A total of 225 thyroid nodules with definitive FNAB  \ncytology or histopathological diagnoses were includ ed in the study. Various \nultrasound (US) features were classified into categ ories based on three \nTIRADS editions. The guidelines were assessed regar ding sensitivity, \nspecificity, predictive values, and diagnostic accu racy to compare diagnostic \nvalue. \n \n \n \n \nResults or Findings: American College of Radiology (ACR)-TIRADS \ndemonstrated the best diagnostic accuracy (63.1%), the highest specificity \n(58.7%), and positive predictive value (36.3%) amon g three different TIRADS \nsystems. Korean (K)-TIRADS exhibited the highest se nsitivity (94.2%), \nnegative predictive value (96.1%), and the most fav orable negative likelihood \nratio (0.13). The European (EU)-TIRADS had a sensit ivity of 90.4%, specificity \nof 48.6%, and diagnostic accuracy of 58.2%, ranking  between the other two \nguidelines across most parameters. \nConclusion: The rigorous use of the guidelines established by e ach of the \nthree TIRADS systems would have markedly reduced th e number of FNABs \nperformed. The comparison of the three guidelines i n our study indicated that \nthey are effective screening methods for identifyin g malignant thyroid nodules. \nAmong them, K-TIRADS showed the most effective diag nostic performance in \nsensitivity, while ACR-TIRADS yielded the best spec ificity. \nLimitations: The main limitation of our study was its single-cen ter design. \nAnother limitation of our study was the lack of his topathological diagnoses for \nall nodules. All nodules with malignant cytological  results underwent surgery \nthat enables us to reach histopathological diagnose s, but only two nodules with \nbenign cytology underwent surgery due to the clinic ian's discretion and the \npatient's preference. Cytology results were used as  a reference standard \ndiagnosis in the remaining nodules. \nFunding for this study: No funding \nEthics committee - additional information: The study was performed in \naccordance with the ethical guidelines of the Helsi nki Declaration and \napproved by the local ethics review committee (2024 -3). \nAuthor Disclosures:  \nEren Tobcu: Nothing to disclose \nErdal Karavaş: Nothing to disclose \nGülden Taşova Yılmaz: Nothing to disclose \nZeynep Tobcu: Nothing to disclose \n \n \nThyroid nodule characterization: interobeserver eva luation of different \nTIRADS with and without AI software \n*C. Di Bella*¹, E. David², C. Solito¹, V. Dolcetti¹ , P. Pacini¹, G. Del Gaudio¹,  \nM. Renda¹, C. Catalano¹, V. Cantisani¹; ¹Rome/IT, ² Catania/IT \n(chiaradibella30@gmail.com) \n \nPurpose or Learning Objective: To evaluate the diagnostic performance of \nCAD compared with TI-RADS systems and to compare th e performance of TI-\nRADS when used by operators with different levels o f experience. \nMethods or Background: Three operators with different levels of experience  \nevaluated 484 thyroid nodules and the diagnostic ac curacy of three risk \nstratification systems (ACR-, EU-, K-TIRADS) and CA D software (S-Detect) in \ncharacterizing the nodules. Nodules were characteri zed and stratified by using \nthe three TIRADS systems; then S-detect software wa s applied and the data \nwere compared with each other and with the gold sta ndard (citology). \nResults or Findings: The sensitivity of the human operator measurement a nd \nthe negative predictive value (NPV) is 100%, for al l three types of TIRADS. \nThe positive predictive value (PPV) is 50%. The spe cificity is 78.4% (EU), \n85.7% (ACR), 89% (K); this implies a certain propor tion of \"false positives\", \nespecially in the use of the tirads EU. The sensiti vity of the measurement of s-\ndetect alone is 66.7%, for all TRAIDS. This estimat e implies a low certainty of \nthe negative result. The NPV is 96.2% (ACR, K) and 96.3% (EU). The PPV is \n50% (ACR-K) and 66.7% (EU). The specificity is high er than the sensitivity: \n92.7% (ACR - K) and 96.3% (EU). \nConclusion: S-DETECT combined with EU-TIRADS has similar result s as S-\nDETECT with ACR- and K- TIRADS in terms of sensitiv ity, specificity and NPV. \nHowever, it has a slightly better PPV, suggesting g reater accuracy in correctly \ndiagnosing positive cases than the ACR- and K-class ification systems. S-\nDetect cannot yet be considered a substitute for th e human operator but is a \nvaluable tool for characterizing thyroid nodules, w hen integrated with \nradiologist evaluation and for support tool for les s experienced operators and in \ndoubtful cases \nLimitations: Reduced patient sample. Ultrasound's operator depen dence \nFunding for this study: None \nEthics committee - additional information: Ethics committee \"Sapienza\" \nAuthor Disclosures:  \nGiovanni Del Gaudio: Nothing to disclose \nMaurizio Renda: Nothing to disclose \nPatrizia Pacini: Nothing to disclose \nChiara Di Bella: Nothing to disclose  \nVincenzo Dolcetti: Nothing to disclose \nCarmen Solito: Nothing to disclose \nVito Cantisani: Nothing to disclose \nEmanuele David: Nothing to disclose \nCarlo Catalano: Nothing to disclose \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 248  \nIs Shear-Wave Elastography an Accurate Tool for Eva luating Nodules in \nPatients with Autoimmune Thyroiditis? \n*D. I. Stoian*, A. Borlea; Timisoara/RO \n(stoian.dana@umft.ro) \n \nPurpose or Learning Objective: To evaluate the diagnostic accuracy of \nShear Wave Elastography (SWE) in differentiating be nign from malignant \nnodules in patients with chronic autoimmune thyroid itis. \nMethods or Background: SWE has been widely studied in assessing thyroid \nnodules, but the background of thyroid lymphocitic infiltration and fibrosis may \nimpact diagnostic performance. This study enrolled 130 subjects aged 18-84 \nyears with a previous diagnosis of autoimmune thyro iditis and thyroid nodules. \nSWE measurements were performed on both thyroid par enchyma and \nnodules, assessing elasticity indices (EIs), includ ing mean and maximum \nvalues, and the nodule-to-thyroid (N/T) ratio. Conv entional ultrasound risk \nassessment using TIRADS was also evaluated. The rel ationship between \nelasticity indices and biochemical parameters and t hyroid volume was \nexamined. \nResults or Findings: There were no statistically significant differences  in \nthyroid function or autoimmunity parameters between  benign and malignant \nnodules. Significant differences were found in TIRA DS scores (p < 0.0001), \nmean nodule EI (p < 0.0001), and N/T shear wave rat io (p < 0.0001). The \nmean nodule EI (47.2 kPa for malignant vs. 18.1 kPa  for benign nodules) had \nthe highest diagnostic performance, significantly o utperforming both the \nmaximum EI (p = 0.0360) and the N/T ratio (p = 0.01 30). The mean nodule EI \nalso demonstrated superior diagnostic accuracy comp ared to TIRADS (p = \n0.0025). \nConclusion: SWE demonstrates high diagnostic accuracy in evalua ting \nnodules also in context of autoimmune thyroiditis. The mean nodule EI is the \nmost reliable elastographic parameter, outperformin g other elasticity indices \nand TIRADS in distinguishing malignant from benign nodules. \nLimitations: This study did not include a control group of nodul es without \nautoimmune thyroiditis, limiting comparisons of SWE  performance in different \nthyroid backgrounds. Further studies are needed to assess reproducibility \nacross other clinical settings. \nFunding for this study: There was no funding for the study \nEthics committee - additional information: The studies involving humans \nwere approved by The Ethics Committee of the Victor  Babes University of \nMedicine and Pharmacy. The studies were conducted i n accordance with the \nlocal legislation and institutional requirements. T he participants provided their \nwritten informed consent to participate in this stu dy. \nAuthor Disclosures:  \nDana I Stoian: Nothing to disclose \nAndreea Borlea: Nothing to disclose \n \n \nFollow up or FNAB?: Malignancy Rates of ACR-TIRADS 4 and 5 Thyroid \nNodules <10mm in Diameter \n*E. Tobcu*, E. Karavaş, Z. Tobcu, G. Taşova Yılmaz; Balıkesir/TR \n(etobcu@bandirma.edu.tr) \n \nPurpose or Learning Objective: The purpose of this study is to ascertain the \nrate of malignancy in nodules that are <10 mm in di ameter in the TR-4 and TR-\n5 categories, as defined by the American College of  Radiology Thyroid \nImaging and Reporting Data System (ACR-TIRADS) 2017  whitepaper. \nMethods or Background: Assessment of thyroid nodules was conducted in \naccordance with the 2017 whitepaper of the ACR. The  fine-needle aspiration \nbiopsy (FNAB) procedure was performed under the gui dance of ultrasound. \nLesions designated as Bethesda group 3, 4, 5, and 6  were classified as \"non-\nbenign group,\" while nodules defined as Bethesda gr oup 2 were classified as \n\"benign group”. All patients underwent surgery, exc ept for those with benign \ncytology. \nResults or Findings: A total of 60 nodules were included in the study. B ased \non cytological analysis, 12 of the 15 nodules class ified in the TR-4 category \nwere determined to be benign thyroid nodules, while  3 nodules were classified \nas \"suspicious for malignancy\" (Bethesda-V) (%20) a fter cytological \nassessment. In the assessment of 45 TR-5 nodules, 3 0 were identified as non-\nbenign (66.6%), while the remaining 15 were classif ied as benign thyroid \nnodules after cytological evaluation. \nConclusion: Our investigation demonstrated that <10mm thyroid n odules \nclassified in categories TR-4 and TR-5, according t o ACR-TIRADS 2017 \nguideline, have a malignancy rate of 55%. We think that FNAB should be \nperformed prior to the decision of active surveilla nce, in order to determine the \nPMTCs that exhibit aggressive cytologic features an d to identify those with \nbenign cytology to avoid an unnecessary active surv eillance process. \nLimitations: The research was conducted with a small number of p atients and \nwas completed at a single center. Second, we did no t exclude patients with \nthyroiditis that may have an impact on the accuracy  of FNAB. \nFunding for this study: The authors state that this study has not received any \nfunding \nEthics committee - additional information: This prospective study was \napproved by the institutional review board of our i n¬stitution (decision number: \n2024/3-40), and any requirement of informed consent  was waived. \nAuthor Disclosures:  \nEren Tobcu: Nothing to disclose \nErdal Karavaş: Nothing to disclose \nGülden Taşova Yılmaz: Nothing to disclose \nZeynep Tobcu: Nothing to disclose \n \n \nEvaluating the diagnostic value of arterial enhance ment fraction from \ndual-layer spectral detector CT in lymph node metas tasis of papillary \nthyroid carcinoma \n*L. L. Ye*¹, X. Zheng¹, Y. Liao²; ¹Dongguan/CN, ²Gu angzhou/CN \n(13412192434@163.com) \n \nPurpose or Learning Objective: To assess the diagnostic value of the arterial \nenhancement fraction (AEF) derived from dual-layer spectral detector CT \nscans in detecting lymph node metastasis in papilla ry thyroid carcinoma (PTC). \nMethods or Background: Preoperative spectral CT images of 58 lymph \nnodes from 25 PTC patients (7 males, 18 females; ag ed 28-74) confirmed by \nsurgery and pathology were analyzed. All patients u nderwent lymph node \ndissection. Lymph nodes were classified into metast atic (N = 24) and non-\nmetastatic (N = 34) groups based on pathology. AEF (defined as iodine \nconcentration in the arterial phase / iodine concen tration in the venous phase) \nwas measured. The Mann-Whitney U test was used to c ompare AEF between \ngroups. ROC analysis evaluated AEF's predictive per formance for lymph node \nmetastasis. \nResults or Findings: AEF was significantly higher in metastatic lymph no des \nthan in non-metastatic ones (p<0.05). In terms of d ifferentiating lymph node \nmetastasis, the AUC of AEF is 0.737 (95%CI 0.582-0. 891), with an accuracy of \n0.810, a sensitivity of 0.583, and a specificity of  0.971. \nConclusion: AEF from dual-layer spectral detector CT demonstrat es \nsignificant differences between metastatic and non- metastatic lymph nodes in \nPTC and has diagnostic potential for identifying me tastatic lymph nodes, \nproviding a basis for precise preoperative treatmen t planning. \nLimitations: Not applicable \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nXiaolin Zheng: Consultant: mentor \nLi Li Ye: Speaker: speaker, main person in charge o f abstract \nYuting Liao: Other: statistical researcher, unpaid \n \n \nDiagnostic of intravoxel incoherent motion diffusio n-weighted imaging \nhistogram parameters in distinguishing between beni gn and malignant \nthyroid nodules \n*X. Li*, Y. Yue, J. Ren; Beijing/CN \n(lxp17634986228@163.com) \n \nPurpose or Learning Objective: To explore the diagnostic efficacy of \nhistogram parameters of IVIM in differentiating ben ign and malignant thyroid \nnodules. \nMethods or Background: A total of 51 patients with thyroid nodules were \nretrospectively included from March 2017 to Septemb er 2022, including the \nbenign group 24 cases and malignant group 27 cases . All results were \nconfirmed by surgical pathology. multiple b-value s mall field diffusion weighted \nimages were collected to generate true diffusion co efficient(D), pseudo \ndiffusion coefficient(D*), and perfusion fraction(f ) images. At the same time, an \napparent diffusion coefficient(ADC) image with a b- value of 990s/mm2 was \ngenerated. Manually outline the volume of interest( VOI) of the entire lesion. \nUsing a self-designed program in Matlab, calculate the histogram parameters \nof the D, D *, and f at VOI, and calculate the mean  ADC at VOI. Compare the \nintergroup differences between IVIM histogram param eters and ADC mean. \nUsing multiple logistic regression, further select the optimal parameters and \nestablish a multi parameter joint model and verify and compare the diagnostic \nperformance of different models. \nResults or Findings: The statistically significant differences between b enign \nand malignant thyroid nodules are as follows(P<0.05 ): gender, ADC mean, 5th, \n15th, 85th, and 95th percentile, mean, skewness, an d root mean square \ndifference in the D-plot, 5th and 15th percentile p ercentile, mean, skewness, \nand coefficient of variation in the D * plot. The A UC values for skewness, 15th \npercentile, and root mean square difference of the D-plot combined with \ngender were 0.94, with a sensitivity of 88.46% and a specificity of 90.91%. The \nAUC value of the mean ADC combined with gender is 0 .86, with a sensitivity of \n77.78% and a specificity of 87.50%. \nConclusion: Histogram parameters of IVIM has good diagnostic va lue in \ndistinguishing benign and malignant thyroid nodules . \nLimitations: Not applicable. \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 249  \nFunding for this study: Not applicable. \nEthics committee - additional information: This is a retrospective study. \nAuthor Disclosures:  \nYunlong Yue: Nothing to disclose \nXingpeng Li: Nothing to disclose \nJie Ren: Nothing to disclose \n \n \nNew thyroid imaging reporting and data system (TI-R ADS) based on \nultrasonography features for follicular thyroid neo plasms: A multicenter \nstudy \n*Y. Zhang*; Shanghai/CN \n(zhangyifeng@tongji.edu.cn) \n \nPurpose or Learning Objective: This study aimed to establish a new risk \nstratification system for FTN and new methods for n on-invasive and practical \npreoperative evaluation of thyroid follicular tumor s to reduce missed diagnoses \nand unnecessary biopsies. \nMethods or Background: 535 FTNs of 535 patients from four hospitals were \nincluded in this retrospective study. All the nodul es were randomly divided into \ntest and validation groups. FTN-TIRADS was establis hed based on the results \nof univariate analysis and logistic regression of u ltrasonography features in the \ntest group. Each nodule was evaluated and classifie d by existing risk \nstratification systems (EU-RSS, ATA-RSS, ACR-TIRADS , Chinese TIRADS [C-\nTIRADS]) and FTN-TIRADS. The diagnostic value of FT N-TIRADS in the \nvalidation group was verified and compared with the  test group and the other \nfour risk stratification systems. The unnecessary r ates of fine needle aspiration \n(FNA) of FTN-TIRADS and the other four risk stratif ication systems were \ncompared, too. \nResults or Findings: Test group and validation group included 370 patien ts \nand 165 patients. The following features were indep endent risk factors and \nincluded in FTN-TIRADS: nodule composition, echogen icity, calcifications, halo \nsign, and indistinct boundary with thyroid capsule.  The AUC of FTN-TIRADS \nwas 0.855, statistically higher than EU-RSS, ATA-RS S, ACR-TIRADS, and C-\nTIRADS (0.759, 0.759, 0.753, 0.677, respectively). The FTN-TIRADS of the \nvalidation group had a similar diagnostic performan ce. The unnecessary FNA \nrate of the FTN-TIRADS was 26.0%, which was signifi cantly lower than that of \nEU-RSS (79.9%), ATA-RSS (92.5%), ACR-TIRADS (55.8%)  and C-TIRADS \n(62.2%). \nConclusion: FTN-TIRADS achieved better differential diagnosis o f FTN than \ncurrent risk stratification systems and significant ly reduced the rate of \nunnecessary FNA. \nLimitations: The study was performed retrospectively; a more con clusive \nvalidation could be achieved using a prospective de sign that better captured \nthe diverse characteristics of actual clinical case s. \nFunding for this study: This work was supported by the National Natural \nScience Foundation of China (Grants No. 81927801, 8 1725008, 81772849, \n82171942, and 82371971) \nEthics committee - additional information: The institutional review board of \nthe university-affiliated hospital approved this re trospective study (approval \nnumber 22K82) \nAuthor Disclosures:  \nYifeng Zhang: Nothing to disclose \n \n \n16:00-17:30 Research Stage 2 \nResearch Presentation Session: Oncologic \nImaging \nRPS 2116 \nAdvances in imaging genitourinary cancer \n \nModerator \nT. Akbas; Istanbul/TR  \n(tugana.akbas@acibadem.com) \n \n \nTotal bone diffusion volume on whole-body diffusion -weighted imaging is \na strong prognostic marker of disease survival in m etastatic castration-\nresistant prostate cancer \n*L. D'Erme*¹, A. Candito², G. Avesani¹, S. Bottazzi ¹, R. Emsley², D. Meo²,  \nJ. Carmichael², N. Tunariu², D-M. Koh²; ¹Rome/IT, ² London/UK \n(lucaderme@icloud.com) \n \nPurpose or Learning Objective: To investigate the relationship between total \nbone diffusion volume (tBDV) and global apparent di ffusion coefficient (gADC) \nderived from whole-body diffusion-weighted MRI (WBD WI); as well as \nautomatic bone scan index (aBSI) derived from bone scintigraphy with disease \noverall survival (OS) in metastatic castration-resi stant prostate cancer \n(mCRPC). \nMethods or Background: In this IRB approved study, we retrospectively \nreviewed 302 mCRPC patients (Jan 2015 - Dec 2023) w ho underwent baseline \nWBDWI before systemic anticancer treatment. Segment ation masks of bone \ndisease on b900 WBDWI images were generated by an a utomated tool, and \nrefined by a 3-year experienced oncological radiolo gist, to derive the tBDV and \ngADC values. The aBSI was derived in 265 patients w ith available baseline \nbone scintigraphy. Kaplan-Meier survival curves and  log-rank tests for 5-years \nOS, including their hazard ratios (HR) and 95% conf idence intervals (CI), were \ncompared between the three groups, stratified by th e median values of tBDV \n(89 mL), gADC (0.83), and aBSI (0.021). Significanc e was set at p < 0.05. \nResults or Findings: Patients with tBDV < 89 mL demonstrated significant ly \nlonger OS compared to those with tBDV ≥ 89 mL (40.8 vs 23.7 months; p < \n0.0001; HR 2.28, 95%CI 1.76–2.94). No significant s urvival difference was \nobserved between the gADC groups (32.0 vs 27.2 mont hs; p = 0.4293). OS \nwas significantly longer in patients with aBSI < 0. 021 than those with aBSI ≥ \n0.021 (40.8 vs 24.4 months; p < 0.0001; HR 2.18, 95 %CI 1.66–2.85). \nConclusion: WBDWI-derived tBDV is a strong independent prognost ic marker \nfor OS in mCRPC patients. The tBDV measurement is c omparable if not \nsuperior to aBSI as a predictor of disease survival . \nLimitations: Retrospective design and longitudinal imaging data were not \nconsidered. \nFunding for this study: This study represents independent research funded \nby the National Institute for Health and Care Resea rch (NIHR) Biomedical \nResearch Centre at The Royal Marsden NHS Foundation  Trust and The \nInstitute of Cancer Research, London, and by the Ro yal Marsden Cancer \nCharity, and Cancer Research UK (CRUK) National Can cer Imaging Trials \nAccelerator (NCITA) and Prostate Cancer UK. The vie ws expressed are those \nof the author(s) and not necessarily those of the N IHR or the Department of \nHealth and Social Care. This work uses data provide d by patients and \ncollected by the NHS as part of their care and supp ort. \nEthics committee - additional information: The study was approved by the \nInstitutional Ethics Committee (no. 21/LO/0605). \nAuthor Disclosures:  \nGiacomo Avesani: Nothing to disclose \nDow-Mu Koh: Nothing to disclose \nDavide Meo: Nothing to disclose \nNina Tunariu: Nothing to disclose \nRobby Emsley: Nothing to disclose \nSilvia Bottazzi: Nothing to disclose \nAntonio Candito: Nothing to disclose \nLuca D'Erme: Nothing to disclose \nJuliet Carmichael: Nothing to disclose \n \n \nWhy we shouldn’t trust CT in the evaluation of bone  metastases in \npatients with metastatic prostate cancer: a compari son between pattern \nof changes on CT and bone metastases MET-RADS-P cla ssification \n*S. Bottazzi*¹, L. Russo¹, G. Avesani¹, L. D'Erme¹,  C. Messiou², D-M. Koh²,  \nE. Sala¹, N. Tunariu²; ¹Rome/IT, ²Sutton/UK \n(silvia.bottazzi02@gmail.com) \n \nPurpose or Learning Objective: To compare changes on computed \ntomography (CT) with the MET-RADS-P response assess ment categories \n(RAC) of bone metastases in advanced prostate cance r patients (APCb) during \ntreatment. \nMethods or Background: 102 patients (median age 68years, range 51-83) \nwith APCb who underwent both CT and whole-body magn etic resonance \nimaging (WBMRI) within 30 days at baseline and duri ng treatment were \nincluded. Up to five focal lesions > 10 mm per pati ent were selected based on \none or more of the following: [1] Sclerotic, lytic or mixed lesion on CT; [2] active \nbone marrow lesion on WBMRI [3] newly developed CT or MR lesion. Each \nlesion was assigned a CT pattern of change - based on changes in size and \nHounsfield Unit (HU) – and a RAC according to METRA DS-P criteria. The CT \npatterns were corroborated with the RAC, grouped as  response (RAC1-2), \nstable (RAC3), and progression (RAC4-5). \nResults or Findings: 358 lesions were identified. Of these, 70% (252/358 ) \nwere sclerotic (SL), 6% (21/358) lytic (LL) and 4% (13/258) mixed lesions (ML). \n20% (72/358) showing MR characteristics of active b one metastases were \nundetectable on CT. The most frequent CT patterns o f change on treatment \nwere: stable SL (no changes in density or size) in 35.8% (128/358), increasing \nin size SL (>5 mm) in 11.4% (41/358), new SL (appea red during treatment) in \n10.6% (38/358). Stable SL corresponded to the RAC c lassifications as follows: \n22.3% (29/128) stable treated disease, 25.8% (33/12 8) responding, 23.4% \nprogressing (30/128) and 28.1% (36/128) stable acti ve disease. New SLs \ncorresponded to responding disease in 28.9% (11/38) . \nConclusion: 20% of bone metastases are occult on CT. A stable S L on CT is \na poor predictor of disease status. Hence, CT appea rs unreliable in the \nassessment of bone disease response inAPCb \nLimitations: Na \n\n \n \nSaturday \nAbstract-based Programme \n \n 250  \nFunding for this study: This study represents independent research funded \nby the National Institute for Health and Care Resea rch (NIHR) Biomedical \nResearch Centre at The Royal Marsden NHS Foundation  Trust and The \nInstitute of Cancer Research, London, and by the Ro yal Marsden Cancer \nCharity, and Cancer Research UK (CRUK) National Can cer Imaging Trials \nAccelerator (NCITA) and Prostate Cancer UK. The vie ws expressed are those \nof the author(s) and not necessarily those of the N IHR or the Department of \nHealth and Social Care. This work uses data provide d by patients and \ncollected by the NHS as part of their care and supp ort. \nEthics committee - additional information: Reference no. 21/LO/0605 \nAuthor Disclosures:  \nGiacomo Avesani: Nothing to disclose \nChristina Messiou: Nothing to disclose \nDow-Mu Koh: Nothing to disclose \nNina Tunariu: Nothing to disclose \nSilvia Bottazzi: Nothing to disclose \nEvis Sala: Nothing to disclose \nLuca D'Erme: Nothing to disclose \nLuca Russo: Nothing to disclose \n \n \nCancer Detection Rates in DWI-upgraded Transition Z one lesions align \nwith risk assessment categories in PI-RADS v2.1: a Systematic Review \nand Meta-analysis \n*G. Agrotis*, E. H. P. Pooch, R. G. H. Beets-Tan, I . G. Schoots; Amsterdam/NL \n(g.agrotis@hotmail.com) \n \nPurpose or Learning Objective: To assess and compare cancer detection \nrates (CDRs) of transition zone (TZ) lesions that w ere upgraded from PI-\nRADSv2.1 score 2 to 3(\"2+1\") or from score 3 to 4(\" 3+1\") using diffusion-\nweighted imaging (DWI) and evaluate their clinical impact. \nMethods or Background: A systematic literature search was conducted in \nEmbase, Medline and Web of Science for studies eval uating TZ lesions with \nthe use of DWI, with histology proven Grade Group ≥2 cancer (GG≥2) as \nprimary outcome. Pooled estimates for sensitivity, specificity, CDRs, and Odds \nRatio (OR) were derived from extracted data at lesi on level and quantitatively \npooled using a bivariate binomial and random effect s model. \nResults or Findings: A total of 7 studies included 1,437 TZ lesions. GG ≥2 \nCDRs for PI-RADSv2.1 scores of 1, 2, 2+1, 3, 3+1, 4 , and 5 were respectively \n2%[95% CI:0%-12%], 7%[4%-11%], 12%[6%-24%], 21%[18% -25%], \n37%[23%-53%], 53%[33%-72%], and 86%[40%-98%]. GG ≥2 CDRs of TZ \nscores '2+1' and '2' were statistically different, with OR 3.13[1.31-7.48],p=0.01, \nwhile '2+1' and '3' scores were not, with an OR of 0.76[0.42-1.33],p=0.34. \nGG≥2 CDRs of TZ score '3+1' and '3' were statistically  different, with an OR of \n2.3[1.07-4.95],p=0.03, while scores '3+1' and '4' w ere not with an OR of \n0.63[0.28-1.38],p=0.25. Still, false positive rates  were substantial in both \nsubcategories ('2+1': 76%[73.8%-78.2%] and '3+1': 4 5%[42.4%-47.6%]). \nConclusion: The risk of having significant prostate cancer in ‘ 2+1’ and ‘3+1’ \nTransition Zone lesions, with an upgrading based on  DWI images, is \nappropriately categorized within the PI-RADS v2.1 s coring system, as shown \nby this meta-analysis. Especially TZ lesions with s core ‘3+1’ may impact \nindividualized biopsy-decisions, as 2-in-5 harbor s ignificant disease, similar to \nscore ‘4’ lesions. Still, the high false positive r ate in this sub-category \nemphasizes the need for strategies to minimize over diagnosis. \nLimitations: Data availability and population differences \nFunding for this study: None \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nGeorgios Agrotis: Nothing to disclose \nEduardo H. P. Pooch: Nothing to disclose \nRegina G. H. Beets-Tan: Nothing to disclose \nIvo Gerardus Schoots: Nothing to disclose \n \n \nEvaluation of Arterial Enhancement Fraction and Ext racellular Volume \nFraction from Dual-Layer Spectral CT for Typing and  Grading Renal Cell \nCarcinoma \n*X. Zhang*¹, G. Zhang¹, H. Sun¹, Z. Jin¹, X. Lu², S -H. Yu¹, L. Xu³, J. Zhang¹,  \nX. Bai¹; ¹Beijing/CN, ²Shenyang/CN, ³Hangzhou/CN \n(zhang_xiaoxiao23@163.com) \n \nPurpose or Learning Objective: To explore the value of arterial enhancement \nfraction (AEF) and extracellular volume fraction (E CV) obtained from dual-layer \nspectral CT in the typing and grading of renal cell  carcinoma (RCC). \nMethods or Background: In this retrospective study, patients with \npathologically confirmed RCC who has undergone dual -layer spectral CT were \nincluded. RCC was classified into non-clear cell (n on-ccRCC) and clear cell \n(ccRCC). The ccRCC cases were further categorized a s high-grade or low-\ngrade based on the WHO/ISUP grading system. AEF and  ECV parameter \nmaps were generated from both contrast-enhanced and  iodine concentration \n(IC) images, producing quantitative parameters AEFH U, ECVHU, AEFIC, and \nECVIC. Receiver operating characteristic curves wer e used to evaluate the \nability of these parameters in RCC typing and gradi ng. \nResults or Findings: The study included 68 patients, comprising 13 with non-\nccRCC and 55 with ccRCC. CcRCC showed higher values  of AEFHU, ECVHU, \nAEFIC, and ECVIC compared with non-ccRCC. The multi variate model \ncomprising AEFIC, and ECVIC demonstrated the highes t diagnostic accuracy \nfor ccRCC, with an area under curve (AUC) of 0.822,  sensitivity of 83.6%, and \nspecificity of 76.9%. Among the ccRCC cases, 34 wer e low-grade and 21 were \nhigh-grade. High-grade ccRCCs exhibited significant ly higher ECVHU and \nECVIC than low-grade tumors. The multivariate model  with tumor diameter, \nand ECVIC achieved the highest diagnostic accuracy in identifying high-grade \nccRCC, with an AUC of 0.909, sensitivity of 90.5%, and specificity of 76.5%. \nConclusion: AEF and ECV derived from dual-layer spectral CT can  help \ndistinguish ccRCC from non-ccRCC. Additionally, ECV  can accurately identify \nhigh-grade ccRCC, offering valuable insights for RC C. \nLimitations: Given the retrospective design and relatively small  sample size of \nthe present study, further studies should aim to in clude larger cohorts and \nconsider prospective data collection to validate th ese findings. \nFunding for this study: This study has received funding by the National Hig h \nLevel Hospital Clinical Research Funding [2022-PUMC H-A-033]; the Natural \nScience Foundation of Beijing Municipality [L232133 ]; the Chinese Academy of \nMedical Sciences Initiative for Innovative Medicine  [2022-I2M-C&T-B-019]; \nNational High Level Hospital Clinical Research Fund ing [2022-PUMCH-A-035]; \nNational High Level Hospital Clinical Research Fund ing [2022-PUMCH-B-069]. \nEthics committee - additional information: The study was conducted in \naccordance with the principles of the Declaration o f Helsinki and approved by \nthe institutional Research Ethics Committee \nAuthor Disclosures:  \nJiahui Zhang: Nothing to disclose \nGumuyang Zhang: Nothing to disclose \nXiaomei Lu: Nothing to disclose \nHao Sun: Nothing to disclose \nLili Xu: Nothing to disclose \nXin Bai: Nothing to disclose \nZhengyu Jin: Nothing to disclose \nXiaoxiao Zhang: Nothing to disclose \nSheng-Hui Yu: Nothing to disclose \n \n \nPrognostic role of Whole-body MRI (WB-MRI) in patie nts with metastatic \nprostate cancer receiving systemic anti-cancer ther apy \n*C. Sattin*¹, C. Pizzi¹, F. Arnone¹, P. Hoxha¹, D. Berloco¹, F. Zugni¹,  \nP. Summers¹, A. R. R. Padhani², G. Petralia¹; ¹Mila n/IT, ²Northwood/UK \n(caterina.sattin@unimi.it) \n \nPurpose or Learning Objective: To investigate the potential of the response \nassessment category (RAC) from MET-RADS-P guideline s as prognostic \nbiomarker in metastatic castrate resistant prostate  cancer (mCRPC) patients. \nMethods or Background: We enrolled mCRPC patients who underwent \nwhole-body MRI at baseline and at each time point ( every 12 weeks disease \nuntil progression) after systemic anti-cancer thera py (SACT). We correlated the \nmaximum RAC at time point 1 (TP1) with overall surv ival (OS). Patients were \ndivided in two groups: those with a maximum RAC 1-2  (highly likely or likely to \nbe responding, respectively) and those with a maxim um RAC 3-4-5 (stable \ndisease, likely or highly likely to be progressing)  at TP1. Survival curves were \ndepicted in Kaplan-Meier plots and compared via a l og-rank test and hazard \nratio (HR) using Cox regression model, with point c omparisons of three-year \nsurvival and median survival duration, using R. \nResults or Findings: Out of 31 mCRPC patients enrolled, a higher OS was \nobserved in patients with a maximum RAC 1-2 (N=11) than in those with a \nmaximum RAC 3-4-5 (N=21) at TP1 (log-rank test p=0. 005): median 34 months \n(lower bound 95%CI = 27 months) vs median 12 months  (95%CI 11-28 \nmonths). The HR for the RAC 3-4-5 patients was 1.34  (95%CI 0.83 – 1.85, p= \n0.009). Three-year OS was 30.3% for RAC1-2 vs 5.3% for RAC 3-4-5, for a \ndifference of 25.0% (95%CI -11.1% - 61.2%, p=0.175) . \nConclusion: Our observations support the potential of RAC after  TP1 as a \nprognostic biomarker in mCRPC undergoing SACT. \nLimitations: Retrospective and monocentric study. \nFunding for this study: No fundings \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nFabio Zugni: Nothing to disclose \nFrancesca Arnone: Nothing to disclose \nGiuseppe Petralia: Nothing to disclose \nCaterina Pizzi: Nothing to disclose  \nPaolo Hoxha: Nothing to disclose \nPaul Summers: Nothing to disclose \nProfessor Anwar R. R Padhani: Nothing to disclose \nCaterina Sattin: Nothing to disclose \nDonatello Berloco: Nothing to disclose \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 251  \nPrognostic value of WB-MRI derived Bone Marrow Adip ose Tissue \n(BMAT) in bone metastatic prostate cancer patients treated with \nandrogen deprivation + enzalutamide +/- zoledronic acid \n*N. Di Meo*, C. Buizza, P. Rondi, A. Dalla Volta, A . Borghesi, M. Ravanelli,  \nA. Berruti, D. Farina; Brescia, BS/IT \n(nunzia.dimeo@hotmail.it) \n \nPurpose or Learning Objective: To assess the prognostic significance of \nbone marrow adipose tissue (BMAT) in prostate cance r patients with hormone-\nsensitive bone metastases undergoing whole-body MRI  (WB-MRI) and \nreceiving enzalutamide treatment. \nMethods or Background: Imaging was conducted on a 1.5T MRI scanner \nusing a MET-RADS-P-compliant protocol. Manual singl e-slice segmentation of \nfat fraction (FF%) sequences was performed by one o perator (R1) at the L3 \nvertebral level and across three contiguous slices at the femoral head. WB-\nMRI was performed at baseline and at 6 and 12 month s following the initiation \nof therapy. Absolute BMAT values and temporal chang es were recorded and \ncorrelated with survival outcomes. \nResults or Findings: Of the 126 patients enrolled in this prospective ph ase 2 \nclinical trial, 100 were available for analysis. No  correlation was found between \nBMAT measurements at the L3 vertebra and the femora l head, with the latter \nshowing significantly higher values (90.7% vs. 63.9 %, respectively). A \nsignificant positive correlation was identified bet ween baseline L3 BMAT and \nboth progression-free survival (PFS) and overall su rvival (OS), with hazard \nratios (HR) of 0.37 and 0.33, respectively, after a  median split. Additionally, \nearly changes in L3 BMAT were inversely associated with PFS and OS, with \nHRs of 1.89 and 2.96, respectively. BMAT at the fem oral head was not \nassociated with survival outcomes. \nConclusion: L3 BMAT is a valuable prognostic and predictive bio marker that \ncan be easily derived from WB-MRI. It may contribut e to more personalized \ntreatment strategies for patients with metastatic p rostate cancer. \nLimitations: No external validation. \nFunding for this study: No Funding \nEthics committee - additional information: No \nAuthor Disclosures:  \nAndrea Borghesi: Nothing to disclose \nChiara Buizza: Nothing to disclose \nDavide Farina: Nothing to disclose \nPaolo Rondi: Nothing to disclose \nMarco Ravanelli: Nothing to disclose \nAlberto Dalla Volta: Nothing to disclose \nNunzia Di Meo: Nothing to disclose \nAlfredo Berruti: Nothing to disclose \n \n \nConventional Parameters of Periprostatic Fat on 18F -PSMA-1007 PET/CT: \nA Novel Biomarker for Predicting High ISUP Grade an d Short-Term \nPrognosis in Prostate Cancer \n*L. Chen*, Y. Yang, F. Yao; Wenzhou/CN \n(1004491518@qq.com) \n \nPurpose or Learning Objective: This study aimed to investigate the value of \nperiprostatic fat area and 18F-PSMA-1007 uptake in predicting high ISUP \ngrade and postoperative PSA persistence in prostate  cancer patients using \n18F-PSMA-1007 PET/CT. \nMethods or Background: A retrospective analysis was conducted on clinical \ndata and 18F-PSMA-1007 PET/CT data of 350 prostate cancer patients. 3D-\nSlicer and Lifex software were utilized for delinea ting the region of interest for \nperiprostatic fat and measuring periprostatic fat a rea and 18F-PSMA-1007 \nuptake. The primary outcome of this study was the I SUP grade greater than 3 \nbased on surgical pathological results of radical p rostatectomy. The secondary \noutcome was postoperative PSA persistence, defined as routine follow-up \ntPSA > 0.1 ng/ml. Logistic regression analyses were  performed to assess the \nassociation between characteristics and outcomes an d construct predictive \nmodels. Receiver operating characteristic curves we re utilized to determine \noptimal cutoff values and evaluate model performanc e. \nResults or Findings: Larger periprostatic fat area emerged as an indepen dent \nrisk factor for higher ISUP grade (p < 0.001) and p ostoperative PSA \npersistence (p = 0.009) in prostate cancer patients . Higher 18F-PSMA-1007 \nuptake was also closely associated with higher ISUP  grade (p < 0.001) and \npostoperative PSA persistence (p < 0.001). Models r espectively established to \npredict higher ISUP grade and postoperative PSA per sistence showed good \npredictive performance, with AUC values of 0.736 an d 0.745. \nConclusion: Larger periprostatic fat area and higher 18F-PSMA-1 007 uptake \nare independent risk factors for high ISUP grade an d postoperative PSA \npersistence, which can be used to predict high ISUP  grade and the persistence \nof PSA. \nLimitations: This is a small sample study. The generalizability of the results \nrequires further consideration. \n \n \nFunding for this study: This study was supported by the Wenzhou Major \nProgram of Science and Technology Innovation (Grant  No. ZY2020012). \nEthics committee - additional information: This retrospective study has \nbeen reviewed and approved by the the first affilia ted hospital of Wenzhou \nMedical University ethics committee. \nAuthor Disclosures:  \nYunjun Yang: Nothing to disclose \nFei Yao: Nothing to disclose  \nLixuan Chen: Nothing to disclose \n \n \nDecrease in kidney volume predicts loss of renal fu nction in prostate \ncancer patients receiving LuPSMA treatment \n*F. Jungmann*, L. Steinhelfer, M. R. Makowski, M. E iber, R. Braren; \nMunich/DE \n(friederike.jungmann@tum.de) \n \nPurpose or Learning Objective: Lutetium-177 (177Lu) prostate-specific \nmembrane antigen (PSMA) radioligand therapy (RLT) i s a novel treatment \noption for metastatic, castration-resistant prostat e cancer (mCRPC). Evidence \nis increasing that nephrotoxicity is a delayed side  effect in a considerable \nfraction of patients. The purpose of this study was  to identify prognostic \nmarkers for clinically significant deterioration of  kidney function in patients \nundergoing 177Lu-PSMA RLT. \nMethods or Background: Total kidney volume (TKV) at 3 and 6 months \nfollowing 177Lu-PSMA RLT was extracted from routine  clinical CT scans using \ndeep learning. A cut-off at ≥30% eGFR decline was defined as clinically \nsignificant deterioration of kidney function, given  its indication as a substantial \nrisk of end-stage renal disease. Differences betwee n patients developing an \neGFR decline of ≥30% after 12 months and those who did not consideri ng \nbaseline renal parameters, their relative changes ( ∆%), nephrotoxic risk \nfactors, and the number of 177Lu-PSMA cycles were a nalyzed. Furthermore, \ndistinct threshold values of significant features t o differentiate between the two \npatient groups were identified based on ROC analysi s using the Youden-Index. \nResults or Findings: A ≥10% decrease in TKV at six months predicted a \nsevere eGFR decline of ≥30% at 12 months with high diagnostic accuracy \n(ROC-AUC of 0.90), surpassing all other parameters.  Baseline risk factors, the \nnumber of prior treatment regimens and 177Lu-PSMA c ycles did not correlate \nwith a higher eGFR decrease at 12 months. \nConclusion: Our retrospective analysis demonstrates the feasibi lity of fully \nautomated kidney volume assessment from routine cli nical imaging data to \npredicting significant deterioration of kidney func tion at 12-month after 177Lu-\nPSMA RLT in mCRPC. It is more accurate than early r elative eGFR change \nand might contribute as a non-invasive biomarker wh en treatment decisions \nare pending including determining whether to contin ue/discontinue or adapt \n177LuPSMA treatment. \nLimitations: Retrospective, single-center \nFunding for this study: None \nEthics committee - additional information: Ethical approval for this \nretrospective, HIPAA-compliant analysis was obtaine d from the local \ninstitutional review boards. The requirement for in formed consent was waived \nbecause of its retrospective design. \nAuthor Disclosures:  \nRickmer Braren: Nothing to disclose \nMarcus R. Makowski: Nothing to disclose \nMatthias Eiber: Nothing to disclose \nFriederike Jungmann: Nothing to disclose \nLisa Steinhelfer: Nothing to disclose \n \n \nDeep learning-accelerated MRI Imaging in Patients w ith Prostate Cancer \nand Benign Prostatic Hyperplasia \n*V. Koch*, T. Vogl, R. Strecker, C. Booz, S. Mahmou di, L. D. Grünewald; \nFrankfurt/DE \n(vitali-koch@gmx.de) \n \nPurpose or Learning Objective: The purpose of this study was to investigate \nthe impact of deep learning-accelerated T2-weighted  MRI imaging of prostate \ncancer and benign prostatic hyperplasia (BPH). \nMethods or Background: In this prospective study, adults who underwent 3-\nTesla MRI of the prostate due to suspicion of prost ate cancer or benign \nprostatic hyperplasia were included. Standard seque nces were acquired \naccording to a dedicated protocol compromising T1-,  T2-, and diffusion-\nweighted imaging sequences. Additionally, T2-weight ed imaging sequences \nusing the deep learning algorithm (T2DL) were acqui red in axial, coronal, and \nsagittal planes. Quantitative analysis encompassed time efficiency and \nobjective imaging parameters, including signal-to-n oise ratio (SNR) and \ncontrast-to-noise ratio (CNR). Qualitative evaluati on was independently \nperformed by three blinded radiologists to assess d iagnostic confidence, image \nquality, and lesion sharpness subjectively. Interre ader agreement was \ncalculated using Fleiss κ. \n\n \n \nSaturday \nAbstract-based Programme \n \n 252  \nResults or Findings: A total of 46 male patients (mean age, 70 ± 9 years ) \nwere included. The study cohort encompassed 22 pati ents (48%) with prostatic \ncancer and 24 patients (52%) with BPH. Subjective e valuation of T2DL-\nsequences among all three readers revealed slightly  superior diagnostic \nconfidence, image quality, and lesion sharpness whe n compared to standard \nT2w sequences. Especially regarding focal lesions, T2DL-sequences allowed \nfor significantly sharper demarcation with higher d iagnostic confidence in \ncancer diagnosis. Objective image analysis of T2DL revealed significantly \nhigher SNR and CNR values when compared to conventi onal T2w-sequences. \nAcquisition times of T2w-sequences (axial, coronal,  and sagittal plane) could \nbe reduced by an average of 50 % using T2DL-sequenc es. \nConclusion: Our findings suggest that deep learning-accelerated  T2w-\nsequences in MRI imaging of the prostate allow a re levant reduction in \nacquisition time while maintaining both subjective and objective image quality. \nLimitations: Single-center study. \nFunding for this study: No funding. \nEthics committee - additional information: Approval obtained. \nAuthor Disclosures:  \nChristian Booz: Nothing to disclose \nThomas Vogl: Nothing to disclose \nVitali Koch: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nRalph Strecker: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \n \n \nFeasibility of arterial spin labelling in MRI evalu ation of adnexal lesions: \nA comparative study with dynamic contrast enhanceme nt imaging \n*C. Meinzer*¹, K. Zhang¹, R. Gnirs¹, O. Zivanovic¹,  H-U. Kauczor¹,  \nH-P. Schlemmer¹, F. Kurz², T. Mokry¹; ¹Heidelberg/D E, ²Geneva/CH \n(clara.meinzer@dkfz-heidelberg.de) \n \nPurpose or Learning Objective: To evaluate feasibility of arterial spin labeling \n(ASL) as a non-contrast MRI technique for assessing  solid tissue of adnexal \nlesions and to compare its performance with dynamic  contrast enhanced \n(DCE) MRI. \nMethods or Background: We prospectively included 11 adnexal lesions with \nsolid tissue in nine females. Regions of interest ( ROIs) were annotated on DCE \nimages, and then transferred anatomically to corres ponding sites on ASL \nimages. From these ROIs, we extracted semi-quantita tive DCE parameters: \narea under the curve (AUC), relative area under the  curve (relAUC), peak \nenhancement, time to peak, mean residence time, are a under the first moment \ncurve, and wash-in-rate (WiR). From ASL perfusion m aps, we obtained mean \nadnexal blood flow (ABF). Correlation between ABF a nd DCE parameters was \nassessed using Pearson’s correlation coefficient. F or those parameters \nshowing significant correlation, Bland-Altman plots  were generated to evaluate \nagreement. \nResults or Findings: Pearson's correlation revealed significant correlat ions \nbetween ABF and two DCE parameters: relAUC (r=-0.75 , p=0.008), WiR \n(r=0.65, p=0.031). Other analysed DCE parameters sh owed no statistically \nsignificant correlations with ABF (p>0.05). Bland-A ltman analysis was \nperformed for the significantly correlating paramet ers. For relAUC, mean \ndifference was 135.85 (SD=18.04). No data points fe ll outside the limits of \nagreement, indicating good agreement between ASL an d DCE. Similarly, WiR \nshowed a mean difference of -17.02 (SD=11.24), and no points outside the \nlimits of agreement. \nConclusion: ABF demonstrated significant correlations with relA UC and WiR, \nindicating that ASL can provide comparable perfusio n information to DCE for \nthese metrics. The Bland-Altman analysis further su ggests reasonable \nagreement between ASL-derived perfusion and DCE par ameters for relAUC \nand WiR. \nLimitations: The small sample size limits the generalisability o f the findings. \nASL MRI is susceptible to lower signal-to-noise rat ios and variability in \nperfusion measurements, which can impact accuracy a nd reproducibility. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study received institutional \nreview board approval and written informed consent was obtained from all \nparticipants (S-337/2016). \nAuthor Disclosures:  \nOliver Zivanovic: Nothing to disclose \nClara Meinzer: Nothing to disclose \nKe Zhang: Nothing to disclose \nTheresa Mokry: Nothing to disclose \nRegula Gnirs: Nothing to disclose \nHans-Ulrich Kauczor: Nothing to disclose \nFelix Kurz: Nothing to disclose \nHeinz-Peter Schlemmer: Nothing to disclose \n \n \n \n16:00-17:30 Research Stage 3 \nResearch Presentation Session: Breast \nRPS 2102 \nHow to optimise and use breast MRI \n \nModerator \nN. Sharma; Leeds/UK  \n(Nisha.sharma2@nhs.net) \n \n \nDiagnostic accuracy of abbreviated magnetic resonan ce imaging for \nbreast cancer screening: a multi-reader study \n*S. V. Grinsven*¹, R. Mann², K. M. Duvivier³, M. De  Jong⁴,  \nP. K. De Koekkoek-Doll³, C. Loo³, J. Veltman ⁵, W. B. Veldhuis¹,  \nFor The Dense Trial Study Group¹; ¹Utrecht/NL, ²Nij megen/NL, \n³Amsterdam/NL, ⁴Den Bosch/NL, ⁵Almelo/NL \n(s.e.l.vangrinsven-2@umcutrecht.nl) \n \nPurpose or Learning Objective: Costs and time of a full multi-parametric MRI \nprotocol may be reduced by using an abbreviated MRI  (AB-MRI) protocol. The \nDENSE trial’s multiparametric protocol provided the  unique opportunity to \nstudy the accuracy of various AB-MRI protocols, to identify the minimal \nprotocol necessary to maintain high diagnostic accu racy. \nMethods or Background: Seven radiologists performed incremental reads of \na subset of 518 MRI examinations from the DENSE tri al (women with \nextremely dense breasts and negative mammography). Different sequences \nwere added in four incremental steps, starting with : 1) both high resolution (hi-\nres) and ultra-fast T1-weighted images (T1WI), up t o 120 seconds after \ncontrast-injection only, 2) complemented by diffusi on-weighted images (DWI), \n3) T2-weighted images (T2WI), and 4) finally adding  all remaining full protocol \nsequences: non-fatsat-T1-weighted pre-contrast imag es, all remaining dynamic \nphases, and curve-kinetics. Each radiologist assess ed the same MRI \nexaminations and provided BI-RADS scores for all fo ur steps. We calculated \nthe pooled sensitivity and specificity per incremen tal step by using a \ngeneralized estimating equation model, and the pool ed reading time per \nincremental step by using a linear mixed model. \nResults or Findings: The sensitivity and specificity of the most abbrevi ated \nMRI protocol (step 1) were not significantly differ ent from that of the full \nmultiparametric MRI protocol (step 4) (p=0.68, p=0. 39). The pooled reading \ntime of step 1 was almost 50% shorter than that of the full multiparametric MRI \nprotocol (p<0.01), and the MR acquisition time was 70-80% shorter, depending \non the hospital and scanner vendor. \nConclusion: In a screening setting, a full multiparametric MRI protocol, \nincluding pre-contrast DWI and T2WI, and delayed po st-contrast T1WI, did not \nprovide significant additional diagnostic informati on for making a recall/no-\nrecall decision compared to an ultrafast bi-dynamic  T1WI-only protocol. \nLimitations: A prospective screening study should confirm these results. \nFunding for this study: The DENSE trial is financially supported by the \nUniversity Medical Center Utrecht (UMC Utrecht, Pro ject number: UMCU \nDENSE), the Netherlands Organization for Health Res earch and Development \n(ZonMw, Project numbers: ZONMW-200320002-UMCU and Z onMW Preventie \n50-53125-98-014), the Dutch Cancer Society (KWF Kan kerbestrijding, Project \nnumbers: DCS-UU-2009-4348, UU-2014-6859 and UU-2014 -7151), the Dutch \nPink Ribbon / a Sister’s hope (Project number: Pink  Ribbon-10074), Bayer AG \nPharmaceuticals, Radiology (Project number: BSP-DEN SE), and Stichting \nKankerpreventie Midden-West. For research purposes,  Volpara Health \nTechnologies (Wellington, New Zealand) has provided  Volpara Imaging \nSoftware 1.5 for installation on servers in the scr eening units of the Dutch \nscreening program. \nEthics committee - additional information: On November 11, 2011, the trial \nwas approved by the Dutch Minister of Health, Welfa re, and Sport, under \nadvisement from the Health Council of the Netherlan ds. \nAuthor Disclosures:  \nMathijn De Jong: Nothing to disclose \nPetra Katharina De Koekkoek-Doll: Nothing to disclo se \nWouter B. Veldhuis: Nothing to disclose \nKatya M. Duvivier: Nothing to disclose \nJeroen Veltman: Nothing to disclose \nFor The Dense Trial Study Group: Nothing to disclos e \nSophie Van Grinsven: Nothing to disclose \nClaudette Loo: Nothing to disclose \nRitse Mann: Nothing to disclose \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 253  \nBreast MRI protocol strategies to reduce energy con sumption and carbon \nemissions: phantom and patient tests \n*J. T. Lee*¹, B. K. Seo¹, M. S. Bae¹, H. Choi², K. R. Cho², O. Woo², S. E. Song², \nS-Y. Kim², S. Cheon¹; ¹Ansan/KR, ²Seoul/KR \n(jeoungtaek@gmail.com) \n \nPurpose or Learning Objective: Environmental sustainability in healthcare is \ncrucial, and MRI is a major energy-intensive device  in radiology. We aimed to \nidentify optimal energy-saving breast MRI protocols  by comparing energy \nconsumption and carbon emissions of abbreviated MRI , ultrafast dynamic \ncontrast-enhanced MRI (Ultrafast-DCE), and artifici al intelligence (AI)-assisted \nprotocols against multiparametric MRI, conventional  dynamic contrast-\nenhanced MRI (Conventional-DCE), and non-AI-assiste d protocols using \npatient and phantom tests. \nMethods or Background: A 3-T MRI scanner equipped with a dedicated \nbreast coil and a power meter providing a 1-Hz samp ling rate was used. We \ncompared scan time (seconds), total energy (kW), en ergy consumption (kWh), \nand carbon emissions per scan (kgCO2e) between abbr eviated (n=74) and \nmultiparametric (n=81) protocols, Ultrafast-DCE (n= 81) and Conventional-DCE \n(n=81), and AI-assisted (n=76) and non-AI-assisted (n=76) protocols in 307 \npatients. Additionally, the signal-to-noise ratio w as compared between AI-\nassisted and non-AI-assisted protocols using a brea st MRI phantom. The \nabbreviated protocol included T2-weighted imaging ( T2), diffusion-weighted \nimaging (DWI) (b values: 0 and 800 s/mm²), four-pha se DCE T1-weighted \nimaging (T1), and axillary T1. The multiparametric protocol consisted of T2, \nDWI (b values: 0, 800, and 1400 s/mm²), Ultrafast-D CE, five-phase DCE T1, \nand axillary T1. \nResults or Findings: Abbreviated MRI reduced scan time by 40%, total \nenergy by 36%, and energy/carbon emissions by 62% c ompared to \nmultiparametric MRI. Ultrafast-DCE reduced scan tim e by 81%, total energy by \n83%, and energy/carbon emissions by 97% compared to  Conventional-DCE. \nAI-assisted MRI reduced scan time by 29%, total ene rgy by 30%, and \nenergy/carbon emissions by 52% compared to non-AI-a ssisted MRI, while \nincreasing signal-to-noise ratio by 16% (all p < 0. 001). \nConclusion: Abbreviated, ultrafast, and AI-assisted MRI protoco ls significantly \nreduce energy consumption and carbon emissions, sup porting eco-friendly \nMRI practices. \nLimitations: Not applicable. \nFunding for this study: National Research Foundation of Korea funded by the  \nKorea government (RS-2024-00347290). \nEthics committee - additional information: No \nAuthor Disclosures:  \nMin Sun Bae: Nothing to disclose \nOkhee Woo: Nothing to disclose \nSoo-Yeon Kim: Nothing to disclose \nBo Kyoung Seo: Nothing to disclose \nSung Eun Song: Nothing to disclose \nJeong Taek Lee: Nothing to disclose \nSewon Cheon: Nothing to disclose \nKyu Ran Cho: Nothing to disclose \nHangseok Choi: Nothing to disclose \n \n \nUse of Diffusion-Weighted MRI in Screening High-Ris k Women Under 40 \nfor Breast Cancer \n*C. C. Arıkan*, M. A. Arıkan, M. A. Nazli; Istanbul /TR \n(ceydaceren95@gmail.com) \n \nPurpose or Learning Objective: Diffusion-weighted imaging(DWI) presents a \nrapid, cost-effective, and non-contrast alternative  to contrast-enhanced \nMRI(CE-MRI).This study aims to evaluate the diagnos tic performance of DWI \ncompared to CE-MRI in breast cancer screening for h igh-risk women under 40 \nwhile assessing inter-rater agreement and correlati ng findings with biopsy \noutcomes and ultrasound BI-RADS categories. \nMethods or Background: We retrospectively analyzed the DWI and CE-MR \nimages of 112 women under the age of 40 who had bre ast MRI screening due \nto high risk. Two radiologists independently review ed the images without \nknowledge of ultrasound or pathological results. Pa tients were classified as \nhaving \"diffusion restriction present or absent\" an d \"pathological contrast \nenhancement present or absent.\" Inter-rater agreeme nt was assessed using \nCohen’s kappa coefficient.MRI findings were compare d to ultrasound-based \nBI-RADS reports and biopsy results to determine sen sitivity and specificity. \nResults or Findings: Lesion detection inter-rater agreement was high for  \nDWI(kappa=0.83), and moderate for CE-MRI(kappa=0.57 ). Of the 112 \npatients, 43 underwent biopsy, with 37 benign and 6  malignant diagnoses. DWI \ndetected 5 out of 6 malignant lesions (sensitivity 83.3%).DWI exhibited 10 false \npositives (specificity 73%).CE-MRI detected all mal ignant lesions (sensitivity \n100%), but showed 30 false positives (specificity 1 8.9%). Among the 79 \npatients classified by BI-RADS with no biopsy, 57 w ere BI-RADS1-2, with \n5(%8.7) false positives on DWI and 14(%24,5) on CE- MRI. Of the remaining 12 \nBI-RADS3 cases were recommended for follow-up, DWI was positive in \n2(%16.6) and CE-MRI in 7(%58.3). \nConclusion: DWI shows high sensitivity and specificity, particu larly in breast \ncancer screening for women at high risk below 40 ye ars.The high inter-rater \nagreement shows its reliability, and the reduced re call rate suggests that DWI \nhas the potential to reduce unnecessary treatments compared to CE-MRI. \nLimitations: The study is limited by loss to follow-up patients and those with \nincomplete diagnostic tests \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Başakşehir Çam and Sakura City \nHospital Ethics Committee \nAuthor Disclosures:  \nMehmet Ali Nazli: Nothing to disclose \nCeyda Ceren Arıkan: Nothing to disclose \nMehmet Ali Arıkan: Nothing to disclose \n \n \nPatient comfort in supine breast MRI using a wearab le coil - preliminary \nquestionnaire results \n*L. Nohava*, R. Czerny, M. Tik, E. Laistler, R. Fra ss-Kriegl; Vienna/AT \n \nPurpose or Learning Objective: Supine positioning during breast MRI using a \nwearable coil at 3 T has the potential to improve p atient comfort and to extend \nthe inclusion criteria for breast MR examinations ( obesity, pregnancy). While \nthe technical performance of the wearable coil for supine breast MRI is being \nassessed in ongoing studies, the aim of this study was the evaluation of patient \nperception with the goal of ensuring comfortable pa tient-oriented breast MR \nexaminations. \nMethods or Background: A questionnaire evaluating the impact of \nradiofrequency coils on patient comfort in MRI as a n add-on to clinical studies \ncomparing the performance of flexible coils with st andard rigid coils was \ndeveloped. In an IRB-approved breast MRI study, pat ients underwent one \nexam in supine using a wearable flexible coil and o ne prone reference exam. \nAfter each exam, patients filled in the questionnai re with 18 items using a 7-\npoint Likert scale. So far, 10 questionnaire sets w ere collected. The study \npopulation had a range of different bra sizes (70B- 95D), ages (20-64 years), \nand BMIs (19-30 kg/m2). \nResults or Findings: Significant improvement in patient comfort during s upine \nbreast MRI was found for the following items: “I fe lt comfortable before the \nexam.” (p<0.009); “I found it cumbersome or physica lly demanding to take the \nlying position.” (p<0.026); ”I felt comfortable dur ing the exam.” (p<0.043); “I \nfound the lying position comfortable.” (p<0.047). 9  patients commented on pain \nor discomfort in prone whereas only 1 patient comme nted on discomfort due to \nperipheral nerve stimulation in supine breast MRI. \nConclusion: Patients perceived supine BraCoil MRI as significan tly more \ncomfortable than prone MRI, in anticipation of and during the exam. \nLimitations: The limitations of the study are the small sample s ize, and mono-\ncenter character. \nFunding for this study: Funding was provided by by the Austrian Science \nFund (FWF)/Agence Nationale de Recherche (ANR) gran t FWF I-3618/ANR-\n17-CE19-0022 “BraCoil” and FWF grant P37189 “OPTIMA L”, the Horizon \nEurope Grants No. 101078393 “MRITwins” and No. 1010 71008 “CITRUS”, and \nthe Austrian Society for Senology (ÖGS) support gra nt. \nEthics committee - additional information: The study was approved by the \nEthics Committee of the Medical University of Vienn a (EK No. 2137/2021). \nAuthor Disclosures:  \nLena Nohava: Nothing to disclose \nRoberta Frass-Kriegl: Nothing to disclose \nElmar Laistler: Shareholder: ALSIX GmbH \nMartin Tik: Nothing to disclose \nRaphaela Czerny: Nothing to disclose \n \n \nA breast MRI image quality score (BreastMRI-QUAL): preliminary results \nS. Marziali, *L. Corradini*, M. Zanardo, C. Deprett o, G. Della Pepa, G. Irmici, \nG. P. Scaperrotta, F. Sardanelli; Milan/IT \n(corradinilisa@gmail.com) \n \nPurpose or Learning Objective: Breast MRI is an established technique for \ndiagnosing breast cancer using a multiparametric pr otocol, including \nsequences before/after contrast administration. The  diagnostic performance \ndepends on image quality, limited by misregistratio n artefacts due to patient \nmovement. We propose a standardized image quality s core (BreastMRI-\nQUAL). \nMethods or Background: Two independent readers with 3 years of \nexperience assigned a 4-level score to each sequenc e of 50 consecutive 1.5-T \nexaminations at a tertiary cancer centre, as follow s: 0 = not diagnostic for any \ncause; 1 = relevant artefacts/malpositioning with c onserved diagnostic value \nfor the specific case; 2 = slight artefacts/malposi tioning with conserved \ndiagnostic value; 3 = excellent image quality with full diagnostic value. The \nscore per sequence was summed as follows: (T2-weigh ted*1) + (DWI-b=0*0.5) \n+ (ADCmaps*0.5) + (T1-weighted-precontrast*1) + (T1 -weighted-\npostcontrast*2) + (T1-weighted-subtracted*3). To ob tain a global score (GS) \n\n \n \nSaturday \nAbstract-based Programme \n \n 254  \nfrom 0 to 10, the sum was divided by 2.4, with scor es below 6 considered as \ninsufficient. \nResults or Findings: The average GS between the two readers was 8.0 ± 1. 0 \n(mean ± standard deviation), with 42 cases (84%%) r eceiving a score ≥ 7. Only \n1 case (2%) was scored <6 by both readers. The diff erence between the GSs \nassigned by the readers was ≤ 1 for 32 cases (64%), >1 but ≤ 2 for 17 cases \n(34%), and > 2 for 1 case (2%). The Bland-Altman an alysis showed a mean \ndifference (bias) of 0.28, with the limits of agree ment ranging from -1.82 to \n2.38, indicating the level of agreement between rea ders. The average \nassessment time/examination was 3 min. \nConclusion: BreastMRI-QUAL is a reproducible quality score syst em. Breast \nMRI image quality at a tertiary cancer center was g ood-to-excellent in over \n80% of cases. Multicenter-multivendor validation st udies are needed. \nLimitations: Monocentric study, limited sample size. \nFunding for this study: No funding. \nEthics committee - additional information: Use of anonymized datasets \noutside clinical workflow. \nAuthor Disclosures:  \nGianmarco Della Pepa: Nothing to disclose \nSara Marziali: Nothing to disclose \nFrancesco Sardanelli: Research/Grant Support: Bayer  AG, Bracco imaging, \nGE healthcare Speaker: Bayer AG, Siemens Healthinee rs, Esaote Advisory \nBoard: Bayer AG, Bracco imaging, GE healthcare \nGianfranco Paride Scaperrotta: Nothing to disclose \nMoreno Zanardo: Nothing to disclose \nCatherine Depretto: Nothing to disclose \nLisa Corradini: Nothing to disclose \nGiovanni Irmici: Nothing to disclose \n \n \nDetection of residual fibroglandular tissue on brea st MRI in women \ntreated with mastectomy and DIEP flap breast recons truction \n*N. Smeins*, J. Rooij, Van, E. Heuts, J. B. Houwers , S. Tuinder,  \nT. Van Nijnatten; Maastricht/NL \n(nieke.smeins@mumc.nl) \n \nPurpose or Learning Objective: After breast amputation, and especially skin \nsparing mastectomy, there can be residual fibroglan dular tissue (RFGT). RFGT \ncan influence the risk of breast cancer recurrence.  However, women are not \nscreened for RFGT after mastectomy. This study exam ines the frequency in \nwhich RFGT can be detected on breast MRI after mast ectomy and DIEP flap \nreconstruction and the influence of RFGT on breast cancer recurrence risk. \nMethods or Background: This retrospective, single-centre study included \nfemale patients who underwent mastectomy and DIEP f lap reconstruction. \nPost-reconstruction breast MRI exams from 2007-2022  were reassessed by a \nbreast radiologist to detect potential presence of RFGT. The presence of \nRFGT was rated according to a confidence scale (1-5 ), with a score of 1 \nindicating ‘definitely no breast tissue’ and 5 indi cating ‘definitely breast tissue’. \nLocations suspected of RFGT rated with a score of 4  or more were considered \nRFGT. RFGT prevalence was correlated with disease r ecurrence. \nResults or Findings: A total of 73 patients (85 breasts) were included. RFGT \nwas found in 15 (20.5%) patients and 16 (18.8%) bre asts. Ten (13.7%) local \nrecurrences had occurred after a mean follow-up per iod of 164.3 months \n(range: 27.0-381.0 months). Presence of RFGT result ed in a relative risk of \n2.58 (95% CI 0.83-7.98) for recurring disease. Most  breast MRI exams were \nassessed with a score of 3 on the confidence scale (39.2%). \nConclusion: RFGT is frequently detected on breast MRI after mas tectomy and \nDIEP flap breast reconstruction and might have an a ssociation with disease \nrecurrence. Future studies should focus on the clin ical consequences of \nvisualization of RFGT on breast MRI and whether the re is a role for breast MRI \nin post-mastectomy patients. \nLimitations: No limitations were identified. \nFunding for this study: N. Smeins received a salary from \nKankeronderzoekfonds Limburg. \nEthics committee - additional information: The study was approved by \nMETC azM/UM (reference number METC 2022-3122). \nAuthor Disclosures:  \nJoep Rooij, Van: Nothing to disclose \nThiemo Van Nijnatten: Nothing to disclose \nNieke Smeins: Nothing to disclose \nEstherm. Heuts: Nothing to disclose \nStefania Tuinder: Nothing to disclose \nJanneke B. Houwers: Nothing to disclose \n \n \n \n \n \n \n \n \n \nBreast cancer on post bilateral mastectomy surveill ance MRI \n*T. Arazi Kleinman*¹, J. Lvovski¹, D. Walchok¹, G. Michal², T. Sella³;  \n¹Beer Yakov/IL, ²Tel Aviv/IL, ³Jerusalem/IL \n(t_arazikleinman@yahoo.com) \n \nPurpose or Learning Objective: Current guidelines regarding post bilateral \nmastectomy (BMx) follow-up indicate no need for ima ging. Regardless many \npatients are referred for breast MRI, though this p ractice is not evidence \nbased. The aim of this study to evaluate the role o f MRI in detection of cancer \nin post BMx women. \nMethods or Background: Retrospective analysis of surveillance breast MRI i n \nwomen s/p BMx between the years 2017-2020, at a sin gle institution. Data \ncollected included demographic information, persona l and family history of \nbreast cancer, indication for mastectomy (prophylac tic vs. therapeutic) and \nreconstruction type. Suspicious MRI findings underw ent biopsy and were \ncorrelated with pathology. Malignancy or benignity were determined by either \npathology or stability on imaging for at least 12 m onths. Descriptive statistics \napplied with p<0.05 considered significant. \nResults or Findings: 229 asymptomatic women s/p BMx aged 29-76±8.7 \nyears underwent 709 surveillance studies for a tota l of 1418 breasts examined. \nReconstructions included 1324 (93.3%) silicone, 47 (3.3%) autologous flaps \nand 47 (3.3%) with no reconstruction. 158 (69%) wom en underwent risk-\nreducing prophylactic Mx (rr-Mx) on one side and th erapeutic Mx (t-Mx) for \ncancer on the other, 45 (20%) underwent bilateral r r-Mx and 26 (11%) \nunderwent bilateral t-Mx for bilateral breast cance r. Overall, 782/1418 breasts \nunderwent rr-Mx and 184/1418 breasts t-Mx. Cancer w as detected in six \nbreasts, five post t-Mx and one post rr-Mx. Overall  cancer detection rate (CDR) \nwas 0.4 (6/1418), higher in post t-Mx (CDR= 0.78, 5 /636) than post rr-Mx \n(CDR= 0.12, 1/782), p<0.05. No cancers were detecte d in women post rr-BMx. \nConclusion: Cancer risk in women undergoing bilateral rr-Mx is negligible and \nlikely does not warrant MRI surveillance. In contra st, CDR in post t-Mx women \nwas 0.78 on the side of prior cancer and surveillan ce MRI may be considered. \nLimitations: Single-institution, retrospective \nFunding for this study: No Funding for this study \nEthics committee - additional information: Retrospective study \nAuthor Disclosures:  \nGuindy Michal: Nothing to disclose \nTal Arazi Kleinman: Nothing to disclose \nJoana Lvovski: Nothing to disclose  \nDaria Walchok: Nothing to disclose \nTamar Sella: Nothing to disclose \n \n \nPercentage functional tumor volume on pre-treatment  MRI within HER2+ \nbreast cancer predicts pathologic complete response  to combination \nneoadjuvant immunotherapy and chemotherapy \n*R. J. Weinfurtner*, S. Falcon, D. Ataya, M. Abdala h, O. Stringfield,  \nN. Raghunand, B. Czerniecki, H. Soliman, H. Han; Ta mpa, FL/US \n(Tigerphage@yahoo.com) \n \nPurpose or Learning Objective: To determine if functional tumor volume \n(FTV) analysis of human epidermal growth factor 2 p ositive (HER2+) breast \ncancer on pre-treatment MRI can help predict pathol ogic complete response \n(pCR) in patients treated with dendritic cell vacci ne (DC1) neoadjuvant \nimmunotherapy (NAI) followed by neoadjuvant chemoth erapy (NAC). \nMethods or Background: Patients with HER2+ breast cancer in this pilot tri al \nunderwent pre-treatment MRI, followed by ultrasound -guided intratumoral and \nintranodal DC1 injections, and then NAC prior to po st-treatment MRI and \nsurgery. FTV was calculated on pre-treatment post-c ontrast T1-weighted MRI \nimages using a percent enhancement threshold of 70%  and signal \nenhancement ratio set to 0. The %FTV was calculated  as %FTV = FTV / \nsegmented tumor volume. These were correlated with pathologic response at \nsurgery using unpaired t-tests where p<0.05 was con sidered significant. FTV \nanalysis was also compared to post-NAI/NAC pre-surg ical MRI reports for \ndiagnostic test accuracy comparison. \nResults or Findings: Nineteen patients aged 29-74 (average 54) were \nincluded in the study, and 11 (57%) achieved pCR. M RI complete response \n(mCR) was seen in 11/19 (57%). However, accuracy fo r mCR predicting pCR \nwas only 38%. For %FTV, median was 71%, and patient s achieving pCR had \nhigher %FTV (78% vs 57%, respectively, p=0.007). As  a diagnostic test, %FTV \nabove median accurately predicted pCR in 79% (95% c onfidence interval of \n54-94%) with sensitivity 75%, specificity 86%, PPV 90%, and NPV 67%. \nConclusion: In this pilot study of combination NAI/NAC treatmen t for HER2+ \nbreast cancer, tumors with above median %FTV on pre -treatment MRI \ndemonstrated more favorable response to treatment, achieving pCR in 78%. \nGiven that post-treatment MRI evaluation demonstrat ed low accuracy in \npredicting pCR, pre-treatment FTV analysis may prov e a more accurate \npredictor in future studies. \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 255  \nLimitations: This study was a pilot study with limited sample si ze. \nFunding for this study: Internal institution grant \nEthics committee - additional information: Institutional Review Board (IRB) \nAuthor Disclosures:  \nRobert Jared Weinfurtner: Nothing to disclose \nMahmoud Abdalah: Nothing to disclose \nHatem Soliman: Nothing to disclose \nShannon Falcon: Nothing to disclose \nDana Ataya: Nothing to disclose \nBrian Czerniecki: Nothing to disclose \nHyo Han: Nothing to disclose \nOlya Stringfield: Nothing to disclose \nNatarajan Raghunand: Nothing to disclose \n \n \nBreast MRI: assessment of the Kaiser score in diffe rentiation of non-\nmass lesions \n*M. Vukojevic*, M. M. Nadrljanski, I. B. Krušac, D.  Dimitrijevic,  \nL. J. Raspopović, A. Djajic, M. Mihajlović; Belgrade/RS \n(milosvukojevic999@gmail.com) \n \nPurpose or Learning Objective: To examine the performance of the Kaiser \nscore in the diagnosis of nonmass breast lesions on  MRI. \nMethods or Background: There were 39 female patients with pathologically \nconfirmed nonmass lesions on breast MRI retrospecti vely analyzed. For each \npatient, the Kaiser score was determined and the BI -RADS category assigned. \nAll patients were examined with full diagnostic pro tocol (T2W-STIR, T2W-TSE, \nT1W-TSE, DWI-ADC, 3D-FLASH) on 1.5T and 3T. Specifi city and sensitivity of \nthe Kaiser score were computed and the correlation between the Kaiser score \nand the BI-RADS classification was determined. \nResults or Findings: In the group of patients with nonmass lesions (N=39 ), \nthere were 10 patients (n1) with benign lesions (25 .64%) and 29 patients with \nmalignant lesions (n2). The mean Kaiser score value  in n1=3 and in n2=7. \nThere was significant correlation between the Kaise r score and BI-RADS \ncategory: n1=0.92; n2=0.66; N=0.79. Sensitivity of the Kaiser score was 89.7% \nand specificity equaled 70.0%. ROC curve value reac hed 0.89. \nConclusion: Kaiser score represents a reproducible, sensitive a nd specific \ndiagnostic tool for assessment of nonmass lesions o n breast MRI and may \ncontribute to the adequate BI-RADS categorization a nd appropriate further \nsteps in diagnostic algorithm of the patients with nonmass lesions. \nLimitations: A single center retrospective analysis with the lim ited number of \npatients. \nFunding for this study: None \nEthics committee - additional information: No decision was required for the \nretrospective analysis without the patient interven tion. \nAuthor Disclosures:  \nIva B. Krušac: Nothing to disclose \nDejan Dimitrijevic: Nothing to disclose \nLuka Josif Raspopović: Nothing to disclose \nMarko Mihajlović: Nothing to disclose \nAndjela Djajic: Nothing to disclose \nMilos Vukojevic: Nothing to disclose \nMirjan M. Nadrljanski: Nothing to disclose \n \n \nDistinguish HER2-low expression level in breast can cer: insights from \nqualitative and quantitative MRI analysis \n*Y. Shen*, C. You, Y. Gu; Shanghai/CN \n \nPurpose or Learning Objective: To investigate whether qualitative and \nquantitative MRI features can reflect HER2-low expr ession breast cancer. \nMethods or Background: The benefit of novel antibody-drug conjugates in \nHER2-low expression breast cancer suggests that the  conventional binary \nclassification HER2 status is insufficient to meet the needs of clinical diagnosis \nand treatment.The pre-treatment breast MRI images o f 232 patients with \npathologically confirmed breast cancer were retrosp ectively analyzed. \nClinicopathologic features and MRI features were re corded. The qualitative \nMRI features included BI-RADS descriptors in DCE-MR I, and intratumoral T2 \nhyperintensity and peritumoral edema in T2WI. The q uantitative features were \ngenerated by multi-b-value DKI, including mean, med ian, 5th, 95th percentile, \nskewness, kurtosis and entropy of ADC, Dapp and Kap p histogram from the \nmono-b and multi-b value models. \n \n \n \n \n \n \n \n \n \nResults or Findings: HER2 status was categorized into HER2-zero (n=60), \nHER2-low (n=91) and HER2-over expression (n=81). Fo r MRI features, the \nproportion of intratumoral T2 hyperintensity was hi gher in HER2-low than in \nother groups (p=0.009, p=0.008). For the lesion typ e, the mass lesions were \nmore common in HER2-zero group than in HER2-low gro up (p=0.038). For \nmass lesions, mass shape (p<0.001) and margin(p<0.0 01) were significantly \ndifferent between HER2-low and other groups, and ma ss shape is the \nindependent predictive factor (HER2-low vs. HER2-ze ro: p=0.010, HER2-low \nvs. HER2-over: p=0.012). The area under the ROC cur ve (AUC) of qualitative \nfeatures to distinguish HER2-low and -zero was 0.76 3 (95% CI: 0.667-0.859). \nQuantitative features differed between HER2-low and  -overexpression groups, \nespecially in NME-related lesions. All combined var iables (Combinedall) had \nthe best performance in predicting HER2-low, with a n AUC of 0.802 (95% CI: \n0.701 - 0.903). \nConclusion: Qualitative and quantitative MRI features are valua ble for \nnoninvasively distinguishing HER2-low expression br east cancer, and have \ntheir advantages in mass and NME lesions, respectiv ely. \nLimitations: Single-center retrospective study with limited samp les. \nFunding for this study: Not applicable. \nEthics committee - additional information: Fudan University, Shanghai \nCancer Center \nAuthor Disclosures:  \nYajia Gu: Nothing to disclose \nChao You: Nothing to disclose \nYiyuan Shen: Nothing to disclose \n \n \nPre- and post-contrast assessment of apparent diffu sion coefficient in \nearly tumor response assessment in patients on neoa djuvant \nchemotherapy \n*M. M. Nadrljanski*, I. B. Krušac, D. Dimitrijevic,  L. J. Raspopović, A. Djajic,  \nM. Mihajlović; Belgrade/RS \n(dr.m.nadrljanski@gmail.com) \n \nPurpose or Learning Objective: To assess the difference in pre- and post-\ncontrast assessment of apparent diffusion coefficie nt (ADC) in early tumor \nresponse to neoadjuvant chemotherapy (NACT) after t he 2nd cycle. \nMethods or Background: There were 43 patients (N=43) included in \nretrospective analysis of ADC (b50, b850) in assess ment of early tumor \nresponse in responders (R, n1=19) and non-responder s (NR, n2=24), defined \npathologically . In all patients, diffusion-weighte d imaging (EPI sequence) was \nperformed before and after application of contrast medium (gadobutrol, 1 \nmmol/L; 0.1 mL/kg). All patients were examined on e ither 1.5T or 3T unit in \nsame institution with full diagnostic protocol (T2W -STIR, T2W-TSE, T1W-TSE, \nDWI b50, b850, 3D-FLASH, DWI b50, b850). \nResults or Findings: In R, mean pre-contrast ADC: 1.17+/-0.07 mm2/s x 10 -\n3, significantly differs from pre-contrast ADC in N R: 0.98+/-0.09 mm2/s x 10-3; \np<0.0001. In R, mean post-contrast ADC value: 1.09+ /-0.08 mm2/s x 10-3, \nsignificantly differs from post-contrast ADC in NR:  0.91+/-0.09 mm2/s x 10-3; \np<0.0001. Significant difference was obtained for R  between mean value of \nADC on pre- and post-contrast DWI: 1.17+/-0.07 mm2/ s x 10-3 vs. 1.09+/-0.08 \nmm2/s x 10-3; p<0.001. Significant difference was o btained for NR between \nmean value of ADC on pre- and post-contrast DWI: 0. 98+/-0.97 mm2/s x 10-3 \nvs. 0.91+/-0.09 mm2/s x 10-3; p<0.001. \nConclusion: DWI before and after the application of contrast me dia did not \nclinically influence the differentiation between R and NR, although ADC was \ngenerally lower after the application of contrast m edium. Although significantly \ndifferent, ADC values for R before and after the ap plication of the contrast \nmedium did not show clinical relevance and did not interfere with the \ninterpretation of the findings. The same applied fo r the ADC values for NR. \nLimitations: Small number of patients in a single center retrosp ective study. \nFunding for this study: None, \nEthics committee - additional information: Institutional Review Board \nwaived the need for decision for the collection and  analysis of medical record \ninformation, with no participant interaction. \nAuthor Disclosures:  \nIva B. Krušac: Nothing to disclose \nDejan Dimitrijevic: Nothing to disclose \nLuka Josif Raspopović: Nothing to disclose \nMarko Mihajlović: Nothing to disclose \nAndjela Djajic: Nothing to disclose \nMirjan M. Nadrljanski: Nothing to disclose \n \n \n \n \n \n \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 256  \n16:00-17:30 Research Stage 4 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 2105 \nArtificial intelligence in chest imaging \n \nModerator \nC. M. Schaefer-Prokop; Amersfoort/NL  \n(cornelia.schaeferprokop@gmail.com) \n \n \nStandardized platform to evaluate, compare, and ana lyze AI-based \nsoftware for detection and classification of lung n odules for the purpose \nof CT lung cancer screening implementation \n*X. Ouyang*¹, K. Togka², D. Han², H. L. Lancaster²,  I. Schuldink²,  \nA. N. Walstra², C. Van Der Aalst¹, H. J. De Koning¹ , M. Oudkerk²; \n¹Rotterdam/NL, ²Groningen/NL \n(x.ouyang@erasmusmc.nl) \n \nPurpose or Learning Objective: Low-dose CT detects lung nodules and \nconsequently lung cancer (LC) at an early stage, pr oven to reduce LC \nmortality. To aid radiologists, commercially availa ble AI-based software have \nbeen developed to analyze lung nodules. Self-report ed performance metrics \nappear promising, however, there remains no indepen dent, standardized \nplatform for external validation. We aimed to devel op a standardized, \nindependent, trustworthy platform to assess and com pare the performance of \ncommercially available AI-lung nodule analysis soft ware. \nMethods or Background: We developed a platform using a sequestered \ndataset of 560 scans from the EU-funded 4-IN-THE-LU NG-RUN (4ITLR) lung \ncancer screening implementation trial. The platform  is based on systematic \nStructured Query Language (SQL) database architectu re. Output of AI \nsoftware in different data formats was reformatted and stored to standardized \nSQL records, eliminating manual errors, and allowin g AI software results to be \ncompared using uniformed data analysis algorithms t o the final consensus \nresult of an expert radiologist panel. Performance is evaluated on two levels: \nnodule level and participant level. Nodule level co mpares the detection/ \nclassification of the reference nodule per particip ant and participant level was \nbased on the largest-detected solid nodule. \nResults or Findings: Performance of AI software at nodule level was repo rted \nusing frequencies of agreement and discrepancies wi th the 4ITLR consensus \nresult on reference nodule. At the participant leve l, Cohen's kappa coefficient is \nused to measure the agreement level with reference.  \nConclusion: The standardized platform developed provides an ind ependent \nassessment of AI software performance. Clinical use rs benefit from reliable \ncomparison of outcomes for lung nodule analysis and  transparency of \ncommercial AI in radiology. \nLimitations: No limitations have been identified yet. \nFunding for this study: The 4-IN-THE-LUNG-RUN trial is funded by the \nEuropean Union (grant number:848294) \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMatthijs Oudkerk: Nothing to disclose \nAnna N.H. Walstra: Nothing to disclose \nHarriet Louise Lancaster: Nothing to disclose \nHarry J. De Koning: Nothing to disclose \nKaterina Togka: Nothing to disclose \nDaiwei Han: Nothing to disclose \nXiaotong Ouyang: Nothing to disclose \nIlona Schuldink: Nothing to disclose \nCarlijn Van Der Aalst: Nothing to disclose \n \n \nBenchmarking of Artificial Intelligence and Radiolo gists for Lung Cancer \nScreening in CT: The LUNA25 Challenge \n*D. Peeters*¹, B. Obreja¹, N. Antonissen¹, R. Dinne ssen¹, Z. Saghir²,  \nE. Scholten¹, R. Vliegenthart³, M. Prokop¹, C. Jaco bs¹; ¹Nijmegen/NL, \n²Hellerup/DK, ³Groningen/NL \n(dre.peeters@radboudumc.nl) \n \nPurpose or Learning Objective: The imminent implementation of lung cancer \nscreening and growing workload for radiologists dem onstrates the need for \nsafe and validated AI algorithms. At present, it is  challenging to adequately \nvalidate and benchmark the increasing amount of AI algorithms being \ndeveloped. In this study, we present the LUNA25 cha llenge, a public \ncompetition aiming to evaluate the diagnostic perfo rmance of AI algorithms and \nradiologists in lung nodule malignancy risk estimat ion at screening CT. \nMethods or Background: The LUNA25 dataset will include 5051 screening \nCT scans from the National Lung Cancer Screening Tr ial (NLST), with 624 \nmalignant and 7414 benign nodules. Participating te ams can access this \ndataset to develop AI algorithms. For algorithm val idation, a separate set of 65 \nmalignant and 818 benign nodules from the Danish Lu ng Cancer Screening \nTrial (DLCST) will serve as a hidden test set. Addi tionally, a subset from \nDLCST with indeterminate nodules measuring 5-15mm i n diameter will be \nassessed by a panel of radiologists with varying ex perience levels to \nbenchmark radiologists’ performance against AI algo rithms. Performance will \nbe measured using area under the ROC curve (AUC) an d at different operating \npoints in terms of sensitivity and specificity. \nResults or Findings: With the NLST and DLCST cohorts collected, the \nchallenge is ready to be introduced to the ECR audi ence. Preliminary results \nwith an in-house developed AI algorithm demonstrate d a mean AUC of 0.91 \n[0.87, 0.95] on DLCST. \nConclusion: The LUNA25 challenge expects to establish a worldwi de \nbenchmark for AI algorithms in estimating lung nodu le malignancy risk at \nscreening CTs and offer insights into how AI compar es to radiologists across \ndifferent experience levels and operating points. \nLimitations: LUNA25 only benchmarks AI’s stand-alone performance , and \ndoes not address workflow integration or radiologis t-AI interaction, which are \nimportant for clinical adoption. \nFunding for this study: Funding was provided by the Dutch Cancer Society \nEthics committee - additional information: The institutional review board \nwaived the need for informed consent because of the  retrospective design and \ndata pseudonymization. \nAuthor Disclosures:  \nRenate Dinnessen: Nothing to disclose  \nDre Peeters: Nothing to disclose \nMathias Prokop: Nothing to disclose \nZaigham Saghir: Nothing to disclose \nErnst Scholten: Nothing to disclose \nRozemarijn Vliegenthart: Nothing to disclose \nNoa Antonissen: Nothing to disclose \nBogdan Obreja: Nothing to disclose \nColin Jacobs: Nothing to disclose \n \n \nSystematic prioritisation of ai-detected chest x-ra y abnormalities for \noptimised lung cancer detection: a multicentre stud y \n*R. Bramley*¹, A. Sharman¹, R. Duerden², S. Lyon¹, M. Ryan³, E. Weber⁴,  \nL. Brown¹, M. Evison¹; ¹Manchester/UK, ²Stockport/U K, ³Sydney/AU, \n⁴Linköping/SE \n(rhidian.bramley@nhs.net) \n \nPurpose or Learning Objective: This multicentre study aimed to establish a \nreproducible and data-driven method for selecting A I-detected chest X-ray \n(CXR) abnormalities to be prioritised for urgent re porting, supporting faster \nlung cancer diagnosis. By analysing cancer prevalen ce and clinical \nsignificance across two distinct cohorts from seven  acute trusts, the study \nsought to maximise lung cancer detection while main taining a high negative \npredictive value (NPV). \nMethods or Background: The study involved two cohorts: a retrospective \ncohort of 1,282 CXR from primary care with detectab le lung cancer (Cohort 1), \nand a prospective cohort of 13,802 consecutive prim ary care adult CXR \n(Cohort 2), with AI deployed in shadow mode. The An nalise-AI platform \nidentified 124 distinct findings. An interactive to ol was developed to assess \nprioritisation strategies based on the cancer preva lence ratio of each AI finding \nindividually and in combination, combined with clin ical judgement. \nResults or Findings: The final prioritisation strategy flagged 41 AI fin dings \nwhich included 95.9% of cancers in Cohort 1 and 21. 6% of CXR in Cohort 2 \n(sensitivity 95.87%, specificity 79.11%, PPV 4.43%,  NPV 99.95%). A further 15 \nAI findings were prioritised based on clinical judg ement as findings not \nassociated with cancer, but requiring prioritisatio n as potentially needing \nprompt intervention. \nConclusion: This study demonstrates a reproducible and data-dri ven method \nfor prioritising AI-detected CXR abnormalities, bal ancing the need for high \nsensitivity and NPV while reducing unnecessary prio ritisation of low-risk cases. \nThe shadow mode approach ensured clinical safety be fore deployment, and \nthe interactive tool provided a systematic means to  assess prioritisation \nstrategies, offering a practical alternative to tra ditional judgement-based \nmethods and supporting more efficient lung cancer d iagnosis. \nLimitations: The tool is designed to support assessment of AI pe rformance in \nshadow mode in the referral population. Performance  metrics should be \nvalidated before deployment in other populations. \nFunding for this study: Funding was provided by the NHS England National \nAI Diagnostics fund (AIDF). \nEthics committee - additional information: The study was performed in \nshadow mode and did not impact on patient care. \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 257  \nAuthor Disclosures:  \nRebecca Duerden: Nothing to disclose \nMatthew Evison: Nothing to disclose \nRhidian Bramley: Nothing to disclose \nSarah Lyon: Nothing to disclose \nElodie Weber: Employee: Sectra \nMelissa Ryan: Employee: Annalise.ai \nLouise Brown: Nothing to disclose \nAnna Sharman: Nothing to disclose \n \n \nBeyond Nodules: A Deep Learning Approach for Compre hensive Lung \nTumour Segmentation on CT \n*L. Petrychenko*, V. Pugliese, R. G. H. Beets-Tan, L. Topff, K. Groot Lipman; \nAmsterdam/NL \n(l.petrychenko@nki.nl) \n \nPurpose or Learning Objective: Several commercially available AI \napplications for lung nodule analysis on chest CT a re limited to the detection \nand segmentation of nodules up to 30 mm. There is c linical potential for AI-\nassisted volumetric analysis and treatment monitori ng of lung tumours of any \nsize, including masses. We aim to develop a Deep Le arning model to detect \nand segment lung lesions, including primary cancers  of all T-stages. \nMethods or Background: In this retrospective study, we collected 1001 ches t \nCT scans from 504 patients (mean age 66.4±10.3 years; 52% female) with \nhistopathologically confirmed primary lung cancer, treated at the Netherlands \nCancer Institute. Both the baseline and first follo w-up scans after treatment \nwere included. Patients were randomly assigned to 9 0% training and 10% \ntesting sets. Two radiologists (4-7 years of experi ence) performed manual \nsegmentation of all lung nodules ≥ 3mm and masses. The deep learning model \ndevelopment utilized a Residual Encoder nnUNet back bone. SGD was \nselected as the optimizer, 10⁻² was set as the initial learning rate, 2 was set a s \nBatch Size, and a nnUNet ResEnc XL architecture was  selected. \nResults or Findings: The dataset represented all T-stages (Tis/T1/T2/T3/ T4: \n1.6/34/21/16/28%) and major histopathological types , with lesion sizes ranging \nfrom 3 to 135 mm. DL model achieved a median Dice S imilarity Coefficient \n(DSC) of 90.0% across all lung lesions, with a medi an of 1 false positive \ndetection per scan. For primary lung tumors, detect ion sensitivity was 77.2%, \nand median DSC was 90.4%. \nConclusion: The DL model demonstrated very good segmentation \nperformance for primary lung tumors of all sizes, i ncluding masses. The model \nhas the potential to assist physicians in treatment  monitoring and planning, \nthough further improvements in detection sensitivit y could enhance its clinical \nutility. \nLimitations: The model requires both external and clinical valid ation. \nFunding for this study: No additional funding was received; the study was \nconducted entirely at the Netherlands Cancer Instit ute. \nEthics committee - additional information: The study received Institutional \nReview Board (IRB) approval. \nAuthor Disclosures:  \nValerio Pugliese: Nothing to disclose \nKevin Groot Lipman: Nothing to disclose \nRegina G. H. Beets-Tan: Nothing to disclose \nLiliana Petrychenko: Nothing to disclose \nLaurens Topff: Nothing to disclose \n \n \nFoundation Model-based Unsupervised CT Kernel Conve rsion for \nStandardizing Emphysema Quantification \n*D. Park*, J-H. Kang, J. Jeong; Seoul, Republic of Korea/KR \n(dhpark.ee@gmail.com) \n \nPurpose or Learning Objective: Emphysema quantification is crucial for \nevaluation and management of chronic obstructive pu lmonary disease \n(COPD). Typically, emphysema is identified in compu ted tomography (CT) \nimages reconstructed with smooth kernels. However, CT reconstruction \nkernels vary, and raw data are often deleted after reconstruction, making it \nhard to adjust the kernel retrospectively. Therefor e, this study aims to develop \nand validate a method for kernel conversion to stan dardize emphysema \nquantification using a foundational deep learning m odel. \nMethods or Background: Paired CT images from nine cases reconstructed \nwith different kernels were used. Automated lung se gmentation was performed \nusing TotalSegmentator, a foundational deep learnin g model. An unsupervised \nkernel conversion method was then applied to transf orm the images to a pre-\ndefined kernel. The kernel conversion was evaluated  by comparing the \nemphysema score (ES), defined as the ratio of regio ns with HU below -950 \nwithin the lung area, before and after the conversi on. \n \n \n \n \nResults or Findings: Before kernel conversion, the mean ES difference \nbetween images reconstructed with smoother kernels (ex: B30f and \nSTANDARD) and those with sharper kernels (ex: B60f and LUNG) was \n11.00±6.85%. After conversion to the target smooth kernel, the mean ES \ndifference was reduced to 2.30±2.65%. Although the sample size was small, \nthis reduction was statistically significant based on a paired t-test (p=0.011). \nConclusion: The foundational model enables the conversion of CT  images \nreconstructed with different kernels to a target sm ooth kernel, allowing for \nstandardized emphysema quantification without the n eed for additional \ndatasets for model development. This result suggest s that the approach can be \neasily used by anyone with the appropriate software . \nLimitations: For more rigorous validation, it is necessary to no t only compare \nthe difference of ES before and after kernel conver sion but also comparative \nevaluation on ground-truth emphysema masks. \nFunding for this study: Not applicable. \nEthics committee - additional information: We used a dataset from the \nKorea Testing Laboratory (KTL) challenge. \nAuthor Disclosures:  \nDoohyun Park: Employee: VUNO Inc. \nJung-Hyun Kang: Employee: VUNO Inc. \nJonghun Jeong: Employee: VUNO Inc. \n \n \nScientific Evidence of AI in Lung Nodule Evaluation  on CT-examinations: \nA Systematic Review \n*J. Paramasamy*, J-W. Groen, A. Leliveld, B. Willem s, J. Aerts, A. Odink,  \nJ. J. Visser; Rotterdam/NL \n(j.paramasamy@erasmusmc.nl) \n \nPurpose or Learning Objective: The purpose of this study was to \nsystematically review the scientific evidence demon strating the efficacy of CE-\nmarked and/or FDA-cleared AI-applications for pulmo nary nodule evaluation \non CT examinations. \nMethods or Background: Following the PRISMA guidelines, Medline, \nEmbase, Web of Science, Cochrane, and Google Schola r databases were \nsearched (Jan 1, 2012–Sep 30, 2024) for studies on AI-based evaluation of \npulmonary nodules on CT-scans. Included articles we re classified according to \na hierarchical model of AI-efficacy: Radiology AI D eployment and Assessment \nRubric (RADAR) framework. Additionally, the evoluti on of evidence over time \nwas examined. \nResults or Findings: A total of 98 articles encompassing AI-applications  for \nlung nodule evaluation from 16 vendors were include d, with approximately \n90% of clinical questions addressed through cross-s ectional studies. These \npublications primarily focused on automatic lung no dule detection, accounting \nfor 61.8% of the studies. All included articles wer e classified based on their \nhighest level of efficacy using RADAR, with the maj ority (41/98) at level 2 \n(diagnostic accuracy). Standalone nodule detection sensitivities in these \nstudies ranged from 50% to 99%. No studies were ide ntified at efficacy levels 5 \n(patient outcomes), 6 (cost-effectiveness), or 7 (l ocal impact). The number of \narticles at levels 3 (diagnostic thinking) and 4 (t herapeutic impact) was 1 \nbetween 2012-2016, and increased to 40 between 2020 -2024. \nConclusion: Current scientific evidence for AI-applications in lung nodule \nevaluation primarily emphasizes diagnostic accuracy . However, there is a \nnoticeable shift in research towards exploring the potential clinical impact of \nthis technology. \nLimitations: No meta-analysis was conducted due to significant h eterogeneity \nin methods and reporting. Moreover, vendor involvem ent in most studies could \npotentially influence outcomes and introduce bias. Furthermore, as AI for lung \nnodule evaluation rapidly evolves, the included art icles since 2012 may reflect \nvariations in AI-application performance over time.  \nFunding for this study: Unrestricted institutional grant \nEthics committee - additional information: Systematic reviews of existing \npublished literature \nAuthor Disclosures:  \nJasika Paramasamy: Nothing to disclose \nArlette Odink: Nothing to disclose \nJoachim Aerts: Nothing to disclose \nAsabi Leliveld: Nothing to disclose \nBo Willems: Nothing to disclose \nJacob Johannes Visser: Grant Recipient: Grant to in stitution from Qure.ai; \nconsulting fees from Tegus; payment to institution for lectures from Roche; \ntravel grant from Qure.ai; participation on a data safety monitoring board or \nadvisory board from Quibim, Contextflow, Noaber Fou ndation, and NLC \nVentures; leadership or fiduciary role on the steer ing committee of the \nPINPOINT Project (payment to institution from Astra Zeneca) and RSNA \nCommon Data Elements Steering Committee (unpaid); p hantom shares in \nContextflow and Quibim. \nJan-Willem Groen: Nothing to disclose \n \n \n \n\n \n \nSaturday \nAbstract-based Programme \n \n 258  \nOn the effect of lesion number on the FROC performa nce in AI-based \nlung nodule detection \n*T. Escobar*, E. Oubel; Montpellier/FR \n(thibault.escobar@intrasense.fr) \n \nPurpose or Learning Objective: In the lung nodule detection context, we \naimed to determine weather the number of lesions pe r patient acts as a \nconfounding variable in performance evaluation, pot entially affecting metrics \nlike sensitivity and FROC, and if this factor shoul d thus be rigorously controlled \nduring testing. \nMethods or Background: Two experiments were conducted using the LIDC-\nIDRI dataset. In both experiments, a trained model was evaluated on sub-\ngroups of patients sorted by deacreasing lesion num ber. The first experiment \nformed cumulative sub-groups by adding and discardi ng 10 patients at a time, \ncreating groups with varying patient numbers. To en sure no effect nor spurious \ncorrelation related to patient number, the second e mployed a sliding window of \n100 patients with a step of 10. For each sub-group,  FROC were computed \nbased on 5-fold cross-validation predictions for th e whole dataset. Additionally, \nfalse positives (FP/s), false negatives (FN/s), tru e positives per scan (TP/s), \nand sensitivity (Se) were evaluated to identify whi ch parts of the FROC were \naffected. \nResults or Findings: A clear inverse relationship between the number of \nlesions and FROC scores was observed. Pearson and S pearman correlation \ncoefficients were significant and equal to -0.9 for  both experiments. As lesion \nnumber increased, TP/s, FP/s, and FN/s increased, w hile Se decreased (i.e. \nTP/s increase did not compensate for FP/s and FN/s ones.). \nConclusion: The number of lesions per patient inversly affects the FROC in \nlung nodule detection models. The number of lesions  per patient should thus \nbe controlled and documented during model evaluatio n to ensure accurate \nperformance assessments and to clarify under which conditions they are \nguaranteed. Further studies are required to rigorou sly examine these effects \nand validate the hypotheses. \nLimitations: Limitations include no specific investigation of th e sources of FP/s \nand FN/s increase with number of lesions. \nFunding for this study: This study was totally funded by the compagny \nIntrasense SA as part of its research and developme nt activity. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nThibault Escobar: Employee: Intrasense SA \nEstanislao Oubel: Employee: Intrasense SA \n \n \nOptimizing Healthcare Sustainability through AI-Ass isted Lung Cancer \nDetection at the time of initial CXR \n*J. Packer*, M. Storey, A. Chung, S. J. Rickaby, G.  Dean, S. C. Shelmerdine, \nC. Malamateniou; London/UK \n(jack.packer1@nhs.net) \n \nPurpose or Learning Objective: Lung cancer is the leading cause of cancer \nmortality in the UK. Early diagnosis is essential b ut hindered by workforce \nshortages and limited CT access. This study evaluat es whether the 'Artificial \nIntelligence triage to same-day CT' (AI-CT) pathway , using the Annalise CXR \nv2.3 model, can enhance healthcare sustainability b y reducing patient visits, \ntravel emissions, and administrative workload while  improving CT access for \nsuspected cancer. \nMethods or Background: Sustainability indices for 26,660 patients (January  \n2022–October 2023) were assessed across five NHS ce ntres in London. Key \nmetrics included time from chest radiograph to CT r eport (Time to CT), AI \naccuracy, and cancer suspicion on CXR and CT, pre- and post-AI-CT \nimplementation. Time to CT was measured using survi val analysis, and \ndiagnostic performance (AUC-ROC, F1 scores) was cal culated based on CT-\nconfirmed cancer. Time saved for patients and admin  teams was estimated, \nand carbon reduction was calculated using the Carbo n Trust online calculator. \nResults or Findings: From 26,660 chest radiographs and 573 CT scans, 75 of \n10,833 patients received same-day CT post-AI, compa red to 13 of 8,434 pre-\nAI, eliminating 150 appointments. Each appointment saved ~1.5 hours per \npatient, totalling 225 hours, with ~37.5 additional  hours saved for admin teams. \nWith travel emissions estimated at 1 kgCO2e per pat ient, this potentially \nresulted in a 150 kgCO2e reduction. The AI model sh owed high sensitivity \n(91%) but low specificity (22%, F1 score 0.56). The re was a significant \nincrease in CT within 1 and 3 days post-suspicious CXR (HR 1.93, 1.34; p < \n0.001). \nConclusion: The AI-CT pathway improved same-day CT access and r educed \npatient visits and emissions. However, the model's low specificity suggests a \nneed for supervised triage to optimize performance.  \nLimitations: The AI’s low specificity and dependence on co-locat ed facilities, \nlimit generalizability. \nFunding for this study: None \nEthics committee - additional information: Local trust clinical audit and QI \nregistration forms approved. \n \n \nAuthor Disclosures:  \nMathew Storey: Nothing to disclose \nSimon Joseph Rickaby: Nothing to disclose \nSusan Cheng Shelmerdine: Nothing to disclose \nJack Packer: Nothing to disclose \nAnthony Chung: Nothing to disclose \nGeraldine Dean: Nothing to disclose \nChristina Malamateniou: Nothing to disclose \n \n \nAn artificial intelligence software for the detecti on of benign and non-\ntypically benign pulmonary nodules on chest CT scan s \n*S. Bennani*¹, N-E. Regnard², M. Durteste¹, V. Mart y¹, R. Quilliet¹,  \nA. Pourchot¹, L. Clovis¹, J. Ventre¹, G. Chassagnon ¹; ¹Paris/FR, ²Lieusaint/FR \n \nPurpose or Learning Objective: Detecting lung nodules on chest computed \ntomography (CT) is an important task that extends b eyond the realm of lung \ncancer screening. This study aimed to compare the p erformance of radiologists \nto an AI software in identifying both non-typically  benign and benign nodules \non CT scans. \nMethods or Background: We retrospectively collected thin-section chest CT \nscans from private practices across France focusing  on patients aged 15 or \nolder. The dataset included patients with non-typic ally benign (solid and sub-\nsolid nodules), benign (granulomas and intrapulmona ry lymph nodes), or no \nnodules. An expert thoracic radiologist defined the  ground truth using past and \nfollow-up scans as well as radiologist reports. We compared the performance \nof four radiologists who had access to limited clin ical information with that of an \nAI software, LungCT (Gleamer, Paris, France). We co nducted patient-wise \nROC and lesion-wise FROC analyses. \nResults or Findings: The final dataset included 250 chest CT scans (age = 66 \n± 23 y, 117 women, 133 men). Among these, 128 scans  contained at least one \nnon-typically benign nodule, 40 displayed only beni gn nodules and 82 were \nnodule-free. The analysis focused on nodules with a  diameter >6mm. The \npatient-wise AUC of the AI was 0.97 [0.93,1.00] and  that of radiologists was \n0.88 [0.83,0.92]. On a lesion-wise basis, the AI ac hieved a sensitivity of 79% \n[75%,83%] for 0.30 false positive (FP) per scan. Ra diologists exhibited an \naverage sensitivity of 72% [67%,76%] with a mean FP  rate of 0.34 [0.25,0.45]. \nConclusion: The AI solution demonstrated robust patient-wise pe rformance \nand comparable lesion-wise detection of non-typical ly benign and benign \nnodules on CT scans to radiologists. \nLimitations: The study’s retrospective design and limited sample  size could \naffect the generalisability of results. Future rese arch should evaluate the \nperformance of AI-assisted radiologists. \nFunding for this study: Gleamer (Paris, France) funded this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nNor-Eddine Regnard: Founder: Gleamer \nGuillaume Chassagnon: Nothing to disclose \nVincent Marty: Employee: Gleamer \nLauryane Clovis: Employee: Gleamer \nRémi Quilliet: Employee: Gleamer \nSouhail Bennani: Employee: Gleamer \nMarion Durteste: Employee: Gleamer \nJeanne Ventre: Employee: Gleamer \nAloïs Pourchot: Employee: Gleamer \n \n \nImproving the generalisation of radiographic AI usi ng automated data \ncuration to mitigate shortcut learning \n*I. A. Selby*¹, E. González Solares¹, A. Breger², M . Roberts¹, J. Babar¹,  \nF. J. Gilbert¹, N. Walton¹, C-B. Schönlieb¹, J. R. Weir-Mccall³; ¹Cambridge/UK, \n²Vienna/AT, ³London/UK \n(iselby@doctors.org.uk) \n \nPurpose or Learning Objective: To investigate whether automated data \ncuration pipelines for chest radiographs can improv e deep-learning model \nperformance on unseen data. \nMethods or Background: Two public datasets, MIDRC-1A and MIDRC-R1, \nwere used to develop diagnostic COVID-19 models usi ng four architectures \n(DenseNet121/ResNet152V2/VGG16/EfficientNetB3). Eac h was trained four \ntimes using a different data curation workflow: WF1 . Raw pixel data with \npartitioning stratified on dataset and COVID-19 sta tus; WF2. DICOM-cleaned \ndata with look-up tables applied, lateral projectio ns and non-chest radiographs \nexcluded, classes balanced on Manufacturer and Proj ection tags, and \npartitioning additionally stratified on the same me tadata; WF3. Cases excluded \nusing an open-source data-cleaning pipeline (AutoQC , \nhttps://gitlab.developers.cam.ac.uk/maths/cia/covid -19-projects/autoqc). \nPartitioning was stratified on projection and the p resence of a pacemaker using \nAutoQC annotations; and WF4. The previous two workf lows combined. \nCOVID-19 diagnosis was inferred from laboratory tes ts, and model \nperformance was assessed using five other public da tasets. Generalisation \nfrom internal-to-external data was quantified using  ΔAUCs. \n\n \n \nSaturday \nAbstract-based Programme \n \n 259  \nResults or Findings: 43,176 radiographs were included in WF1, with 33.2%  \n(14,328) being COVID-19-positive. The development s ets of the other \nworkflows were up to 60% smaller. Similarly, the ex ternal test sets ranged from \n24,563-to-38,417 patients, depending on workflow. T he WF1 models \nexperienced the largest fall in generalisation (mea n ΔAUC = -0.15 [95%CI:-\n0.17,-0.14]), while models trained utilising AutoQC  (WF3-4) demonstrated the \nmost consistent performance with mean ΔAUCs = -0.04 [95%CI:-0.06,-0.02] \nand -0.02 [95%CI:-0.04,0.00] for WF3 and WF4 (p<0.0 5). The WF2 models \nhad a mean ΔAUC = -0.07 [95%CI:-0.09,-0.05]. \nConclusion: Automated data curation can improve the generalisat ion of deep \nlearning models for chest radiographs, facilitating  more consistent performance \non data from new locations and equipment. \nLimitations: Future work should evaluate the tool in multiclassi fication tasks \nand non-COVID-19 datasets. In addition to the curre nt pacemaker detection, \ntools for a broader range of support apparatus are necessary. \nFunding for this study: The authors wish to acknowledge support from the \nEU/EFPIA Innovative Medicines Initiative 2 Joint Un dertaking - DRAGON \n(101005122) (I.S., A.B., M.R., L.E.S., J.B., C.-B.S ., E.S., J.W.M., AIX-\nCOVNET); the National Institute for Health and Care  Research (NIHR) \nCambridge Biomedical Research Centre (BRC-1215-2001 4) (I.S., L.E.S., \nJ.H.F.R., E.S., J.W.M.); Wellcome Trust (J.H.F.R.),  British Heart Foundation \n(J.H.F.R.); the EPSRC Cambridge Mathematics of Info rmation in Healthcare \nHub EP/T017961/1 (M.R., J.H.F.R., C.-B.S.); Cancer Research UK (CRUK) \nNational Cancer Imaging Translational Accelerator ( NCITA) [C42780/A27066] \n(L.E.S.); Cambridge Mathematics of Information in H ealthcare (CMIH) Hub \nEP/T017961/1; Austrian Science Fund (FWF, project T -1307) (A.B.); and the \nTrinity Challenge BloodCounts! project (M.R., C.-B. S.). The AIX-COVNET \ncollaboration is also grateful to Intel for financi al support. \nC.B.S. additionally acknowledges support from the P hilip Leverhulme Prize, \nthe Royal Society Wolfson Fellowship, the EPSRC adv anced career fellowship \nEP/V029428/1, EPSRC grants EP/ S026045/1 and EP/T00 3553/1, \nEP/N014588/1, EP/T017961/1, the Wellcome Innovator Awards 215733/Z/ \n19/Z and 221633/Z/20/Z, the European Union Horizon 2020 research and \ninnovation program under the Marie Skodowska-Curie grant agreement No. \n777826 NoMADS, the Cantab Capital Institute for the  Mathematics of \nInformation and the Alan Turing Institute. \nPlease note that the content of this publication re flects the authors’ views and \nthat neither IMI nor the European Union, EFPIA, or the DRAGON consortium \nare responsible for any use that may be made of the  information contained \ntherein. \nEthics committee - additional information: The Brent Research Ethics \nCommittee, the Health Research Authority (HRA), and  Health and Care \nResearch Wales (HCRW) provided ethical approval for  our retrospective study \n(IRAS ID: 282705, REC No.: 20/HRA/2504, R&D No.: A0 95585). Informed \nconsent was not required as data was pseudonymised.  \nAuthor Disclosures:  \nCarola-Bibiane Schönlieb: Nothing to disclose \nEduardo González Solares: Nothing to disclose \nMichael Roberts: Nothing to disclose \nFiona J. Gilbert: Nothing to disclose \nJonathan R. Weir-Mccall: Nothing to disclose \nAnna Breger: Nothing to disclose \nIan Andrew Selby: Nothing to disclose \nJudith Babar: Nothing to disclose \nNicholas Walton: Nothing to disclose \n \n \n \n \n \n\n \n \n 260  \n \n  \nSunday, March 2 \n\n \n \nSunday \nAbstract-based Programme \n \n 261  \n \n08:00-09:00 Research Stage 1 \nResearch Presentation Session: Breast \nRPS 2202 \nPersonalised risk prediction of breast \ncancer \n \nModerator \nP. Clauser; Vienna/AT  \n(clauser.p@hotmail.it) \nAuthor Disclosures:  \nPaola Clauser: Board Member: EUSOBI; Grant Recipien t: Siemens \n \n \nAssociations of automatically measured breast densi ty with breast \ncancer risk and duration of the pre-clinical detect able phase in a Dutch \nscreening cohort \n*J. Peters*¹, D. Van Der Waal¹, M. Smid-Geirnaerdt¹ , C. Van Gils²,  \nM. Broeders¹; ¹Nijmegen/NL, ²Utrecht/NL \n(Jim.Peters@radboudumc.nl) \n \nPurpose or Learning Objective: Breast density could impact screening \nstrategies. Women with dense breasts face higher br east cancer risk. \nFurthermore, lesion masking may shorten the pre-cli nical detectable phase \n(PCDP), increasing interval cancer rates. This stud y examines the associations \nof automated breast density measures with breast ca ncer risk and PCDP \nduration. \nMethods or Background: Digital mammograms were used from 60,739 \nparticipants in a prospective Dutch screening cohor t (PRISMA study, 2014-\n2019). Dense volume (DV,cm3), volumetric breast den sity (VBD,%) and \nVolpara Density Grade (VDG1-4) were assessed using Volpara version 1.5.0. \nBreast cancer diagnoses were ascertained through li nkage with the \nNetherlands Cancer Registry. Participants with prio r breast cancer (N=73) or \nscreen-detected breast cancer at study entry (N=401 ), were excluded. \nInformation on time to screen-detected and interval  cancers was used to fit a \nthree-state (1:cancer-free, 2:pre-clinical detectab le cancer, 3:clinical cancer) \nMarkov regression model. Hazard ratios (HRs) were c alculated for the effects \nof breast density on state transition intensities 1 ->2 (=breast cancer risk) and \n2->3 (=1/PCDP duration). \nResults or Findings: After a median 4.2 years (IQR 3.9–4.6) we observed 430 \nscreen-detected and 316 interval cancers. Log-trans formed VBD and DV were \npositively associated with increased breast cancer risk (HR 1.12 [95%CI 1.05-\n1.21] and HR 1.32 [95%CI 1.05-1.35] per one-standar d-deviation increase, \nrespectively). Both measures were associated with s horter PCDP (HR 1.48 \n[95% CI 1.30-1.69] and HR 1.19 [95%CI 1.05-1.35]). Mean PCDP duration was \n1.63 years [95%CI 1.11–2.41] for women with highest  density (VDG4), \ncompared to 3.41 years [95%CI 3.31-3.52] for VDG1-3 . \nConclusion: Breast density is associated with breast cancer ris k and PCDP \nduration. VBD has the strongest association with PC DP, indicating a reduced \nsensitivity of biennial mammography, while DV conta ins more information on \nbreast cancer risk. \nLimitations: Other confounders than age, e.g. BMI, were not yet included in \nthe model. \nFunding for this study: The PRISMA study is funded by ZonMw and KWF. \nEthics committee - additional information: CMO Arnhem-Nijmegen \nAuthor Disclosures:  \nJim Peters: Nothing to disclose  \nCarla Van Gils: Nothing to disclose \nMaartje Smid-Geirnaerdt: Nothing to disclose \nMireille Broeders: Nothing to disclose \nDanielle Van Der Waal: Nothing to disclose \n \n \nMammographic biomarkers of cardiovascular risk: the  BAKER study \n*D. Capra*, O. Hoda, C. B. Monti, M. Zanardo, F. Sa rdanelli; Milan/IT \n(davide.capra@unimi.it) \n \nPurpose or Learning Objective: Mammography could offer two sex-specific \nbiomarkers to spotlight cardiovascular risk in wome n: breast arterial \ncalcifications (BAC) and breast density. We conduct ed a prospective case-\ncontrol study evaluating the association between BA C and gynaecological and \ncardiovascular risk factors. \nMethods or Background: Consecutive women showing BAC and age- and \nbreast density-matched controls referred for annual  mammography were \nprospectively enrolled. We recorded anthropometric variables, traditional \ncardiovascular risk factors and gynaecological risk  factors. Breast density was \nclassified as low breast density (BI-RADS categorie s A and B) or high breast \ndensity (BI-RADS C and D). \nResults or Findings: 72 BAC patients and 72 controls were enrolled (medi an \nage 70.0 years, IQR 62.5 to 78 years). Women with B AC had a younger age at \nmenopause (50 vs 52 years, p=0.008), and showed ass ociations with \nbreastfeeding (p=0.041) and parity (p=0.038). Women  with BAC show a \nborderline significant trend towards the use of ant i-ipertensive medications \n(p=0.092). No other differences between cases and c ontrols were observed \n(p>0.101). Higher breast density was significantly associated with younger age \n(p<0.001), lower body weight (p<0.001), lower systo lic blood pressure \n(p=0.003), and higher HDL cholesterol (p=0.017), wh ereas lower breast \ndensity was associated with longer time since menop ause (p=0.003) and use \nof anti-ipertensive medications (p<0.001). \nConclusion: Women with BAC have a younger menopausal age, which  \nrepresents a precocious shift towards a less favour able cardiometabolic \nhormonal balance. Similarly, women with low breast density show an \nunfavourable cardiovascular risk profile, using mor e often anti-ipertensive \nmedications, having a higher systolic blood pressur e and lower levels of HDL \ncholesterol. \nLimitations: Age and breast density matching reduces the statist ical power to \nobserve associations between breast density and car diovascular risk factors \namong women of similar age. Absence of follow up to  record cardiovascular \nevents. \nFunding for this study: General Electric Healthcare supported this study. \nEthics committee - additional information: Ethics committe approval \nnumber 90/INT/2020, 08/09/2020. All participants si gned informed consent. \nAuthor Disclosures:  \nFrancesco Sardanelli: Advisory Board: Bayer AG Advi sory Board: Siemens \nHealthineers Advisory Board: Bracco Imaging Advisor y Board: Esaote Advisory \nBoard: GE Healthcare \nOmar Hoda: Nothing to disclose \nCaterina Beatrice Monti: Nothing to disclose \nMoreno Zanardo: Nothing to disclose \nDavide Capra: Nothing to disclose \n \n \nShort-term risk prediction of breast cancer compari ng risk tools for \ndigital mammography and digital breast tomosynthesi s in U.S. screening \npopulations \nE. F. Conant¹, C. Parghi², P. Hall³, *M. Eriksson*³ ; ¹Philadelphia, PA/US, \n²Addison, TX/US, ³Stockholm/SE \n(mikael.eriksson@ki.se) \n \nPurpose or Learning Objective: Image-derived AI-based risk models \ndemonstrate ability to predict breast cancer risk u sing digital mammography \n(DM) and digital breast tomosynthesis (DBT) imaging  data. However, a direct \ncomparison of performances within the same screenin g population has yet to \nbe conducted. \nMethods or Background: We conducted a nested case-control study \nincluding women aged 35-98 from Solis and UPenn scr eening cohorts, \nbetween 2014 and 2021. Participants were followed f or two screens, with \ncancer diagnosed before August 2022. Two image-base d ProFound AI Risk \nmodels, one for DM and one for DBT, estimated absol ute 1-year breast cancer \nrisks at study-entry. We assessed models’ discrimin atory performance (AUC) \ncontrolling for screening site and classified risks  according to U.S. Preventive \nServices Task Force (USPSTF) thresholds. \nResults or Findings: Study included 780 women with incident breast cance r \n(mean age 63.4±11.2) and 7,481 controls (mean age 5 7.0±10.6). Cancers \nwere diagnosed on average 1.3±0.5 years (range 4 mo nths to 4 years) after \nstudy-entry. At study entry, AUCs of DM and DBT mod els were 0.71 (95% CI: \n0.69-0.73) and 0.75 (95% CI: 0.73-0.77), respective ly (p<0.01). Comparing \nUPenn and Solis data, similar estimates were observ ed for respective models. \nBased on USPSTF guidelines, 14% of women were class ified as high-risk. \nAmong this group, DM model predicted 40% (95% CI: 3 6-43%) of future breast \ncancers, compared to 48% (95% CI: 44-52%) by DBT mo del (p<0.01). Non-\nsignificant differences in proportions of future br east cancers were observed \ncomparing sites. \nConclusion: The image-derived DM and DBT AI-risk models predict ed 40-\n48% of future breast cancers at study-entry in two U.S. screening populations. \nThe DBT model predicted a significantly higher prop ortion of future cancers \ncompared to the DM model emphasizing the need for s ome women to obtain \nsupplemental screening. \nLimitations: Study is limited to only 2 sites in U.S. \nFunding for this study: iCAD, Inc. Swedish Research Council Swedish \nBreast Cancer Association \nEthics committee - additional information: The study has been reviewed \nand approved. \nAuthor Disclosures:  \nMikael Eriksson: Patent Holder: iCAD, Inc. \nChirag Parghi: Nothing to disclose \nPer Hall: Patent Holder: iCAD, Inc. \nEmily F. Conant: Advisory Board: iCAD, Inc. \n\n \n \nSunday \nAbstract-based Programme \n \n 262  \nImpact of Breast Density Metrics on Personalized Br east Cancer \nScreening Protocols \n*G. Gennaro*¹, L. Bucchi², A. Ravaioli², F. Caumo¹;  ¹Padova/IT, ²Meldola/IT \n(gisella.gennaro@iov.veneto.it) \n \nPurpose or Learning Objective: To evaluate the impact of different breast \ndensity metrics on the personalization of breast ca ncer screening. \nMethods or Background: The RIBBS study (ClinicalTrials.gov NCT05675085) \nis a personalized breast screening study targeting young women. The protocol \nused digital breast tomosynthesis (DBT) and double reading to stratify \nparticipants based on individual breast cancer risk  and breast density. In this \nprotocol, breast density determined the need for su pplemental ultrasound (US), \nwhile breast cancer risk guided the frequency of sc reening. A quantitative \nsoftware tool provided volumetric breast density, w ith a 25% threshold used to \nidentify women who needed supplemental US. This stu dy compares \nstratification based on this volumetric approach wi th stratification using \ncategorical breast density metrics, both objective and human. \nResults or Findings: A total of 10,269 women, all aged 45 years, were \nenrolled in the RIBBS study. Of these, 1,904 women (18.5%) had ≥25% breast \nvolumetric density and underwent additional US. Usi ng the categorical breast \ndensity provided by the same software, 41.1% of par ticipants would have been \ncategorized as having a BIRADS “d” category density , 2.2 times higher than \nthat identified through the volumetric metric. If B IRADS categorization had \nbeen performed by human readers, considering every DBT classified as “d” by \nat least one reader, the percentage would have drop ped to 32.3%, still 1.7 \ntimes higher than the current protocol. \nConclusion: This study shows that the choice of breast density metrics can \nsignificantly influence the stratification process in personalized breast cancer \nscreening. Quantitative metrics allow more precise stratification than \ncategorical approaches, improving the feasibility o f supplemental imaging in \nclinical practice. \nLimitations: Changes in screening performance due to the use of categorical \nbreast density for supplemental US remain unassesse d, as the current results \nare based solely on the applied protocol. \nFunding for this study: This specific subanalysis had no funding \nEthics committee - additional information: The RIBBS study was approved \nby the Ethics Committee with the following code \"RI BBS 2019/37\" \nAuthor Disclosures:  \nFrancesca Caumo: Nothing to disclose \nLauro Bucchi: Nothing to disclose \nAlessandra Ravaioli: Nothing to disclose \nGisella Gennaro: Nothing to disclose \n \n \nChanges in Mammographic Density for Breasts Develop ing and not \nDeveloping Breast Cancer \n*J. Gjesvik*, N. Moshina, S. Sagstad, M. Larsen, Å.  S. Holen, M. B. Bergan,  \nS. Hofvind; Oslo/NO \n(jogj@kreftregisteret.no) \n \nPurpose or Learning Objective: The evidence on longitudinal changes in \nmammographic density in breasts developing cancer i s limited. We aimed to \nanalyse mammographic density among women developing  and not developing \nbreast cancer over three consecutive screening roun ds in BreastScreen \nNorway. \nMethods or Background: In this retrospective cohort study, 66,696 women \naged 50-69 years with three consecutive screening e xaminations performed in \nRogaland and Hordaland counties, 2007-2020, were in cluded. A total of 909 \nwomen were diagnosed with screen-detected and 287 w ith interval cancer. \nMammographic density data was obtained from an auto mated software \n(Volpara 1.5.0 and 1.5.4.0) and included absolute ( cm3) and percent (%) \ndense volume per woman and breast. A linear mixed-e ffects model with a fixed \neffect for each woman was applied on a breast level  to define the changes in \nabsolute and percent dense volume. The model was ad justed for age at entry, \nbreast volume, and history of benign breast disease . \nResults or Findings: Mean age for women not developing breast cancer was  \n62.5 years (standard deviation, SD: 5.1), while it was 62.3 (SD: 4.4) for women \nwith screen-detected cancer and 61.9 (SD: 4.8) for interval cancer. Absolute \nand percent dense volume decreased over time in all  women. In breasts \ndeveloping cancer the rate of decrease was lower fo r absolute dense volume, \nestimate=0.004 (95% CI: 0.002-0.007, p=0.041), comp ared to breasts not \ndeveloping cancer. The rate of decrease was also lo wer for percent dense \nvolume, estimate=0.003 (95%CI 0.000-0.007, p=0.053) , in breasts developing \nversus not developing cancer. \nConclusion: Absolute dense volume decreased to a lower degree i n breasts \ndeveloping versus not developing cancer. Longitudin al changes in absolute \ndense volume could be used for more precise breast cancer risk prediction and \nscreening personalization. \nLimitations: The study is retrospective and the population is fa irly \nhomogenous. \nFunding for this study: No funding \nEthics committee - additional information: Regional Commitee for Medical \nand Health Research Ethics \nAuthor Disclosures:  \nNataliia Moshina: Nothing to disclose \nJonas Gjesvik: Nothing to disclose \nMarie Burns Bergan: Nothing to disclose \nÅsne Sørlien Holen: Nothing to disclose \nSilje Sagstad: Nothing to disclose \nSolveig Hofvind: Nothing to disclose \nMarthe Larsen: Nothing to disclose \n \n \nUsing Artificial Intelligence to Detect Subclinical  Breast Cancer \n*J. Gjesvik*¹, N. Moshina¹, C. Lee², D. L. Migliore tti³, S. Hofvind¹; ¹Oslo/NO, \n²Seattle, WA/US, ³Davis, CA/US \n(jogj@kreftregisteret.no) \n \nPurpose or Learning Objective: Investigate whether an artificial intelligence \nalgorithm (AI) trained for detecting breast cancer scored the breast developing \nbreast cancer and the breast not developing breast cancer differently years \nbefore diagnosis. \nMethods or Background: In this retrospective cohort study, we included \nwomen aged 50-69 who attended three consecutive bie nnial screening rounds \nbetween 2004 and 2018, as part of BreastScreen Norw ay. A total of 116 495 \nwomen were included in the final study population, 1265 with screen-detected \nbreast cancer detected at, and 342 with interval ca ncer diagnosed within two \nyears after, the third screening round. We used a c ommercial AI algorithm to \nscore each breast with a risk score between 0 and 1 00. \nResults or Findings: For women developing screening-detected breast \ncancer the mean AI-score at the first screening rou nd for the breast developing \nbreast cancer was 19.2 (SD: 28.6), and 82.7 (SD: 26 .7) after the third \nscreening round. The score was 9.5 (SD: 19.0) in th e first and 5.0 (SD: 15.7) in \nthe third screening round for the breast not develo ping breast cancer. For \ninterval cancer, the mean scores for breasts develo ping cancer were 17.8 \n(SD:26.3) and 33.1 (SD: 33.8), respectively, and me an scores for breasts not \ndeveloping cancer were 10.5 (SD: 19.9) and 8.4 (SD:  18.7), respectively. For \nwomen not developing breast cancer, the mean AI sco re was 7.1 (SD: 15.2) in \nthe first and 6.4 (SD: 14.5) in the third screening  rounds, respectively. \nConclusion: AI-scores were higher in breasts developing cancer up to 6 years \nbefore it was diagnosed. The findings suggests that  commercial AI algorithms \nfor breast cancer detection might be considered for  identifying women at higher \nrisk of developing breast cancer. \nLimitations: This is a retrospective study, and the population i s mostly \nhomogenous. \nFunding for this study: Funded by the Norwegian Cancer Society (Pink \nRibbon) \nEthics committee - additional information: Regional Committee for Medical \nand Health Research Ethics, Norway \nAuthor Disclosures:  \nChristopher Lee: Nothing to disclose \nNataliia Moshina: Nothing to disclose \nJonas Gjesvik: Nothing to disclose \nDiana L Miglioretti: Nothing to disclose \nSolveig Hofvind: Nothing to disclose \n \n \nExternal validation of a mammographic masking predi ction model in the \nDutch Breast Cancer Screening Program \n*S. D. Verboom*¹, J. G. Mainprize², J. Peters¹, M. Broeders¹, M. Yaffe²,  \nI. Sechopoulos¹; ¹Nijmegen/NL, ²Toronto, ON/CA \n(sarah.verboom@radboudumc.nl) \n \nPurpose or Learning Objective: To externally validate a lesion masking \nprediction model for mammograms, Mammatus, develope d on a North \nAmerican cohort, in a larger retrospective breast c ancer screening cohort from \none screening center in The Netherlands. \nMethods or Background: A total of 935 digital mammograms from the Dutch \nBreast Cancer Screening Program with a unilateral i nvasive breast cancer that \nwas either screen detected or diagnosed within 24 m onths after a negative \nscreening (interval cancer) were included. All mamm ograms were \nretrospectively evaluated for the visibility of mal ignant masses using all \navailable diagnostic imaging and clinical informati on. Mammatus was applied \non the contralateral mammogram to eliminate the inf luence of the lesion. The \narea under the receiver operator characteristics (R OC) curve (AUC) when \nusing Mammatus to distinguish examinations with scr een-detected cancers \n(assumed low masking risk) from interval cancers (a ssumed high masking risk) \nwas computed. The AUC was compared to that of the o riginal cohort and to \nthat obtained using volumetric breast density (VBD)  as a predictor. A second \nthree-category ROC analysis was performed, with int erval cancers that were \nretrospectively visible classified as intermediate lesion masking. \n \n\n \n \nSunday \nAbstract-based Programme \n \n 263  \nResults or Findings: Mammatus achieved an AUC of 0.70 (95%CI 0.67-0.74) \nfor distinguishing between screen-detected- (n=632)  and interval-cancer \nexams (n=303). This performance did not differ from  the original study \n(AUC=0.75 (0.68-0.82), p=0.20), and outperformed VB D (AUC=0.66 (0.62-\n0.70, p<0.002). The three-category ROC analysis sho wed that Mammatus \noutperformed VBD at identifying low risk of lesion masking (AUC=0.74 (0.70-\n0.77)), however, not for identifying high risk (AUC =0.69 (0.65-0.74)). \nConclusion: Mammatus performed well in predicting breast cancer -masking \nrisk in a Dutch screening cohort. This suggests tha t adding information other \nthan density improves prediction of lesion masking.  \nLimitations: There is no ground truth of lesion masking risk, th erefore the best \npossible approximation is used. \nFunding for this study: aiREAD financed by the Dutch Research Council \n(NWO), Dutch Cancer Society (KWF), and Health Holla nd (HH) \nEthics committee - additional information: The Radboudumc ethics \ncommittee declared that this study falls outside th e scope of the Dutch Medical \nResearch involving Human Subjects Act and could be carried out without \napproval of an Institutional Review Board. \nAuthor Disclosures:  \nJim Peters: Nothing to disclose \nMireille Broeders: Research/Grant Support: Hologic,  Screenpoint Medical, \nSectra Benelux, Volpara Healthcare, Lunit, and iCAD  Speaker: Hologic and \nSiemens Healthcare \nSarah Delaja Verboom: Nothing to disclose \nJames G. Mainprize: Founder: Calavera Surgical Desi gn, Inc. Research/Grant \nSupport: GE Healthcare \nMartin Yaffe: Research/Grant Support: GE Healthcare  \nIoannis Sechopoulos: Speaker: Canon, Siemens Health care Research/Grant \nSupport: Siemens Healthcare, Canon Medical Systems,  ScreenPoint Medical, \nSectra Benelux, Volpara Healthcare, Lunit Advisory Board: Koning Corp. \n \n \n08:00-09:00 Research Stage 2 \nResearch Presentation Session: Neuro \nRPS 2211 \nContrast agents and energy sustainability \nin neuroimaging \n \nModerator \nY. Özsunar; Aydın/TR  \n(yeldaozsunar@gmail.com) \n \n \nEfficacy of gadoquatrane, a novel low-dose high rel axivity macrocyclic \ngadolinium-based contrast agent, at 5-, 10- and 15- min post injection in \ncomparison to gadobutrol for CNS CE-MRI \n*C. Deuschl*¹, K. Kudo², A. Liu³, M. A. Klemens³, B . M. Hofmann³,  \nP. Palkowitsch³, B. P. Liu⁴; ¹Essen/DE, ²Sapporo/JP, ³Berlin/DE,  \n⁴Chicago, IL/US \n(cornelius.deuschl@uk-essen.de) \n \nPurpose or Learning Objective: This dose-finding Phase 2 study for the \nnovel tetrameric macrocyclic GBCA gadoquatrane inve stigated the efficacy of \n0.04 mmol Gd/kg body weight (bw) gadoquatrane in co mparison to 0.1 mmol \nGd/kg bw gadobutrol for CE-MRI in patients with CNS  lesions at 5-, 10- and \n15-min post injection (pi). \nMethods or Background: Adult patients with CNS lesions were included in \nthe study and underwent two CE-MRIs within an inter val of 3 to 14 days, first \ngadobutrol, then gadoquatrane. Images were acquired  at 5- (3D IR-GRE), 10- \n(2D-SE) and 15-min pi (3D IR-GRE). The 5-, 10- and 15-min CE-MRIs were \nassessed in the Blinded Independent Central Review using visualization \nparameters (contrast enhancement, border delineatio n, internal morphology) to \ninvestigate the time course of enhancement. Quantit ative signal intensity \nmeasurements were also performed. Safety parameters  were assessed and \npharmacokinetic (PK) samples were taken to determin e plasma concentrations \nfor both GBCAs. \nResults or Findings: The sum of the lesion visualization parameters at e ach \npoint for the average reader was very similar for g abobutrol (9; 8.86; 9.09) and \ngadoquatrane (8.94; 8.8; 8.97), with differences cl ose to zero (-0.06; -0.06; -\n0.12). These results were representative for the th ree individual readers and \neach visualization parameter. The results of the qu antitative measurements \nsupported the qualitative evaluations. The safety a nd concentration-time profile \nof gadoquatrane were very similar to gadobutrol, bu t dose-proportionally lower. \nConclusion: Gadoquatrane at a dose of 0.04 mmol Gd/kg bw shows similar \nefficacy as gadobutrol at 0.1 mmol Gd/kg bw on 5-, 10- and 15-min pi CE-MRI \nimages. \nLimitations: The limitation of the study is the small sample siz e of 50 \nevaluable patients. \nFunding for this study: This Phase 2 study was sponsored by Bayer AG. \nEthics committee - additional information: Ethics vote from all sites \navailable. \nAuthor Disclosures:  \nAlex Liu: Employee: Bayer \nBenjamin P. Liu: Nothing to disclose \nCornelius Deuschl: Nothing to disclose \nPetra Palkowitsch: Employee: Bayer \nKohsuke Kudo: Nothing to disclose \nBirte Maria Hofmann: Employee: Bayer \nMark Alexander Klemens: Employee: Bayer \n \n \nClinical experience with gadopiclenol in patients wi th brain metastases: \nSingle and double doses, comparison of standard and  delayed scanning \n*J. Vymazal*, Z. Ryznarova, R. Liščák, A. Rulseh; Prague/CZ \n(josef.vymazal@homolka.cz) \n \nPurpose or Learning Objective: To demonstrate the role of gadopiclenol in \nMR imaging of brain metastases with at standard (0. 05 mmol/kg) and double \ndose (0.1 mmol/kg), including delayed examination ( ca 15 minutes after \ncontrast administration). \nMethods or Background: Ninety-one subjects with known brain metastases \nunderwent MRI prior to radiosurgery (44 female, mea n age 64.2 ±12 years). A \nstandard T1W 3D SPACE sequence was acquired in all subjects (3 subjects at \n1.5T, 88 at 3T) post-contrast, followed by a delaye d acquisition. In roughly half \nof subjects (n=47), the delayed acquisition was pre ceded by additional contrast \nadministration (cumulative double dose). After rand omization, 3 blinded \nreaders evaluated the number of lesions for each pa tient in 2 sessions, 1 \nmonth apart. A paired t-test was used to compare ea rly and late evaluations in \nthe two groups (subjects with cumulative single dos e or double dose), the \nWelch t-test was used to compare differences in (la te-early) single versus (late-\nearly) double doses. \nResults or Findings: Agreement between all three blinded readers was use d \nfor evaluation. The late single dose exam revealed 170 metastatic lesions \ncompared to 161 metastatic lesions on the early exa m. Late double-dose exam \nrevealed 138 metastatic lesions compared to 121 met astatic lesions on the \nsingle dose early exam. The number of lesions on ea rly versus late \n(single/double dose) exams did not differ significa ntly (p=0.4 and p=0.11, \nrespectively). No significant difference was detect ed between double dose and \ndelayed single dose. \nConclusion: Although relatively more lesions were detected on d elayed exam \n(single/double dose), the difference did not reach significance, likely reflecting \nsmall effect size. Furthermore, no significant diff erence was found between late \nsingle and double dose. \nLimitations: The study was performed at a single institution. On ly the SPACE \nsequence was used for blinded reading. \nFunding for this study: Supported by MH CZ - DRO (NNH, 00023884) No. \n244302 \nEthics committee - additional information: Approved on June, 5, 2024 \nAuthor Disclosures:  \nRoman Liščák: Author: co-author \nZuzana Ryznarova: Author: co-author \nJosef Vymazal: Consultant: Bracco Consultant: GE He althcare \nAaron Rulseh: Author: corresponding author \n \n \nGlobal body exposure to gadolinium after administra tion of the Human \nEquivalent Dose (HED) of gadopiclenol, in compariso n to the HED of \ngadobutrol and gadoterate in Healthy Rats \n*M. Rasschaert*, E. Couloumy, C. Hollenbeck, E. Ren ard, I. Janot, N. Decout, \nM. Lefebvre, C. Factor, P. Robert; Roissy CdG Cedex /FR \n \nPurpose or Learning Objective: Gadopiclenol is a high relaxivity macrocyclic \nGadolinium-Based Contrast Agent (mGBCA) approved in  2022 in the USA and \nDecember 2023 in Europe. The aim of this study was to evaluate the time-\ndependent gadolinium Gd exposure in main body organ s over 5 months in rats \nafter a single injection of gadopiclenol or two mGB CAs. \nMethods or Background: Healthy female rats were allocated to 3 groups: \ngadopiclenol (Elucirem®), gadobutrol (Gadovist®) an d gadoterate (Dotarem®). \nOne single administration of GBCA at the Human Equi valent Dose was \nrealized, blinded. Selected tissues (including cent ral (CNS) and peripheric \nnervous system (PNS) organs, excretion organs, bone ) were collected for total \nGd determination by Inductive Coupled Plasma-Mass S pectrometry at 1 day, 1 \nweek, 1 or 5 months after injection (n=10/group and  timepoint). The global \nexposure to Gd in all investigated organs was estim ated by calculating the \narea under the curve (AUC). \n\n \n \nSunday \nAbstract-based Programme \n \n 264  \nResults or Findings: After gadopiclenol administration, the Gd exposure over \nthe 5 months compared to gadoterate and gadobutrol was respectively -43% \nand -41% in CNS, -28% and -32% in PNS (sciatic nerv e, spinal nodes, \nfootpads, spinal cord), -25% and -68% in skin, -1% and -26% in liver, -19% to -\n44% in kidney and -31% to -39% in spleen. In bone, the Gd exposure after \ngadopiclenol administration was +80% and +16% highe r in femur diaphysis, -\n1% and +48% in epiphysis. In bone marrow, the AUC w as -22% and -35% \nlower compared to gadobutrol and gadoterate respect ively. Globally, the \noverall Gd exposition was -25% and -40% after gadop iclenol compared to \ngadoterate and gadobutrol, respectively. \nConclusion: In our experimental conditions, the measured exposu re to Gd \nafter gadopiclenol injection is significantly reduc ed compared to the other \nmarketed GBCAs in most of the investigated organs. \nLimitations: Animal Study \nFunding for this study: None \nEthics committee - additional information: Research project \n2015080615264829 approved by French MESR Ministry \nAuthor Disclosures:  \nClaire Hollenbeck: Employee: Guerbet \nPhilippe Robert: Employee: Guerbet \nCecile Factor: Employee: Guerbet \nIlona Janot: Employee: Guerbet \nNathalie Decout: Employee: Guerbet \nMylène Lefebvre: Employee: Guerbet \nEmilie Couloumy: Employee: Guerbet \nElisabeth Renard: Employee: Guerbet \nMarlène Rasschaert: Employee: Guerbet \n \n \nUse of gadolinium-based contrast agents (GBCA) over  the years: impact \non monitoring multiple sclerosis disease activity \nS. M. Sceppacuercia¹, C. Tur¹, *A. H. M. E. Hammam* ², W. H. E. Hamed²,  \nO. Sarwani², D. Deborah¹, C. Auger Acosta¹, T. A. Y ousry²,  \nA. Rovira Cañellas¹; ¹Barcelona/ES, ²London/UK \n(zekus83@gmail.com) \n \nPurpose or Learning Objective: To describe the evolution of GBCA use, the \ndetection of Gd-enhancing, and new T2 lesions over time in two independent \ncohorts of MS patients. We also examined the associ ation between new T2 \nlesions and clinical features, focusing on the role  of GBCA use. \nMethods or Background: We included all MS patients clinical MRIs at the \nMultiple Sclerosis Centre of Catalonia, Spain, and the Queen Square MS \nCentre, UK, during May from 2015 to 2022. Clinical and demographic data \nincluded age, sex, disease duration, clinical pheno type, progression, relapses \nsince the last MRI, and DMT variables. Brain MRI da ta included GBCA use, the \nnumber of Gd+ lesions, and new T2 lesions. Statisti cal analysis used logistic \nregression models adjusted for confounders. \nResults or Findings: 479 patients(cohort1), and 794 (cohort2). Both coho rts \nshowed a significant decrease in GBCA use over time , an increase in DMT \nexposure (p=0.032), and a reduction in relapses (p< 0.001) and progression \nevents (p=0.006). A significant reduction in Gd-enh ancing lesions was \nobserved in Barcelona. The detection of new lesions  also decreased over time, \nlikely due to an older population (p<0.001) and inc reased use of high-efficacy \nDMTs (p<0.001). GBCA administration was independent ly associated with a \nhigher likelihood of detecting new lesions (p=0.030 ). \nConclusion: Over time, there was a decrease in GBCA use, Gd-enh ancing \nlesion detection, and new/enlarging T2 lesions, sug gesting a trend towards a \nless aggressive disease course, possibly due to mor e effective DMTs and an \naging population. GBCA use was independently associ ated with higher \ndetection of new T2 lesions, possibly reflecting he ightened neuroradiologist \nvigilance when new inflammatory activity is suspect ed and the facilitation of T2 \nlesion detection by Gd-enhancing lesions. \nLimitations: This includes the retrospective nature, a single mo nth/year \nrepresentation and unequal MRIs' frequencies among years and centres. \nFunding for this study: A retrospective study with no fund needed. \nEthics committee - additional information: A retrospective study. \nAuthor Disclosures:  \nSofia Maria Sceppacuercia: Nothing to disclose \nTarek A. Yousry: Nothing to disclose \nWeaam Hamed Elsayed Hamed: Nothing to disclose \n Alejandro Rovira Cañellas: Other: Alex Rovira serv es/ed on scientific advisory \nboards for Novartis, Sanofi, Synthetic MR, BMS, Roc he, and Biogen, and has \nreceived speaker honoraria from Bayer, Sanofi, Merc k-Serono, Teva \nPharmaceutical Industries Ltd, Novartis, Roche, BMS  and Biogen, and is CMO \nand co-founder of TensorMedical. \nAhmed Hassan Mohamed Elzeki Hammam: Nothing to disc lose \nCarmen Tur: Other: Carmen Tur has received honorari a from Roche and \nNovartis and is a steering committee member of the O’HAND trial and of the \nConsensus group on Follow-on DMTs. \nCristina Auger Acosta: Nothing to disclose \nOmran Sarwani: Nothing to disclose \nDeborah Deborah: Nothing to disclose \nDifferent rates and patterns of symptoms associated  with gadolinium \nexposure (SAGE) between linear and macrocyclic gado linium based \ncontrast agents \n*I. Shahid*, E. Lancelot; Paris/FR \n(imran.shahid@guerbet.com) \n \nPurpose or Learning Objective: Some patients who received multiple \nadministrations of gadolinium-based contrast agents  (GBCAs) have been \nreported to develop “symptoms associated with gadol inium exposure” (SAGE). \nThe aim of this study was to analyze pharmacovigila nce data and to explore \nthe various SAGE patterns of linear and macrocyclic  GBCAs among patients \nexhibiting three or more SAGE symptoms. \nMethods or Background: SAGE symptoms were searched by preferred terms \n(PTs) in different system organ class (SOC) categor ies in the FDA Adverse \nEvent Reporting System (FAERS) database, over a 6-y ear period ranging from \n2014 to 2019, for 3 linear and 3 macrocyclic GBCAs.  \nResults or Findings: The analysis of FAERS data revealed a significantly  \nhigher SAGE weight for the linear GBCAs (20-24%) th an for the macrocyclic \nGBCAs (5-9%). For the linear GBCAs, the most preval ent combinations of 3 \nSAGE symptoms were reported in 152-164 occurrences,  while for the \nmacrocyclic GBCAs this range was significantly lowe r (1-13 occurrences). \nMoreover, the patterns of SAGE combinations differe d significantly between \nboth categories of GBCAs. \nConclusion: Linear GBCAs are associated with higher SAGE report ing rates \nthan macrocyclic GBCAs. They also present different  patterns of SAGE \ncombinations. \nLimitations: This study did not provide any demonstration of a c ausal \nrelationship between the reported events and GBCA a dministration to the \npatients. \nFunding for this study: None \nEthics committee - additional information: Not Applicable (No patients were \ninvolved) \nAuthor Disclosures:  \nEric Lancelot: Employee: Guerbet \nImran Shahid: Employee: Guerbet \n \n \nHow Compressed-Sense acceleration technique impacts  the Central \nNervous System Magnetic Resonance Imaging energy co nsumption \n*S. Kalari*¹, I. Seimenis¹, E. Psatha², D. Apostolo u¹, I. Loulakas¹,  \nE. Karavasilis²; ¹Athens/GR, ²Alexandroupolis/GR \n(raykalari@gmail.com) \n \nPurpose or Learning Objective: Compressed-sensing is a new acquisition \ntechnique that enables efficient signal acquisition  and reconstruction reducing \nthe scanning time. The aim of our study was to esti mate the impact of CS on \nenergy consumption during brain imaging. \nMethods or Background: Thirty individuals underwent brain MRI using the \nsame exam protocol including 3DT1, 3DT2 and 3D Τ2 Flair with and without CS \non a 3.0TPhilips MRI system. Energy consumption was  recorded using \nkilowatt-hour energy measurement sensors (0.017 Hz sampling rate) for all the \nabove sequences. The 3DT1, 3DT2 and 3DT2 Flair with  and without CS image \nquality was assessed quantitatively using VolBrain free available online \nsoftware. Images were also reviewed from an experie nced neuroradiologist \nwho evaluated their total image quality and 3DT2 Fl air lesion detection \nsensitivity. Sequences’ signal to noise ratio (SNR) , image quality indices \nderived from VolBrain, qualitative indices and thei r energy consumption were \nintroduced to SPSS to perform paired t-test statist ical analysis (p<0.05). \nResults or Findings: The mean acquisition time of 3DT2 FLAIR, 3DT2 and \n3DT1 with and without CS were 210sec±0.8sec vs 390s ec±1.1sec, \n200sec±0.7sec vs 370sec±1sec and 230sec±0.1sec vs 400sec±1.2sec \nrespectively. We didn’t find any statistical differ ences in both image qualitative \nand quantitative quality metrics. There was no subs tantial difference of the \nSNR. Concerning the energy consumption/min CS incre ased by 42% the \nenergy consumption rate in 3DT2 FLAIR (0.57kWh/min versus 0.40kWh/min) \ntherefore the total energy consumption was decrease d by 10% (1,99 versus \n2.2 kWh). 3DT2 and 3DT1 energy with and without CS consumption rate were \n0.60kWh/min, 0.56kWh/min and 0.58kwh/min and 0.53kw h/min. Thus, the \n3DT2 and 3DT1 total energy consumption were decreas ed 45,9% and 42,5%, \nrespectively. \nConclusion: CS image acceleration technique has the potential t o reduce MRI \nscanning time and energy required while maintaining  high image quality. \nLimitations: Not applicable \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The study received institutional \nreview board approval and written consent was obtai ned from all participants. \n \n \n \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 265  \nAuthor Disclosures:  \nEvlampia Psatha: Nothing to disclose  \nSelmina Kalari: Nothing to disclose  \nEfstratios Karavasilis: Nothing to disclose  \nIoannis Seimenis: Nothing to disclose  \nIoannis Loulakas: Nothing to disclose  \nDimitrios Apostolou: Nothing to disclose \n \n \nA methodology for reducing the environmental impact  of the energy \nconsumption of a neuroradiology MRI system using da ily electricity \nemission data \n*A. Roletto*¹, M. Verga², G. L. Viganò², S. Zanoni² ; ¹Milan/IT, ²Brescia/IT \n(rolettoandrea@yahoo.it) \n \nPurpose or Learning Objective: Magnetic resonance imaging (MRI) scanners \ncontribute significantly to the carbon footprint of  healthcare sector, and efforts \nto reduce environmental impact are based on reducin g energy consumption. \nThe aim of this work was to develop a methodology t o reduce carbon \nemissions by reorganising an MRI neuroradiology sys tem's diagnostic activities \nbased on daily electricity emission data. \nMethods or Background: In July 2024, at a large public hospital of Brescia  \n(Italy), the energy profiles of two weeks of diagno stic activity of a \nneuroradiology MRI system were recorded using HT GS C60 Power \nMeter/Analyzer (HT-Italia, Italy), calculating the total energy consumption and \nits distribution over days. Finally, the amount of carbon dioxide equivalent \n(CO2e) emitted by electricity consumption was calcu lated using open-source \ndaily emission data (www.electricitymaps.com) and t he amount that can be \nsaved by rescheduling diagnostic activities to lowe r emission day periods. \nResults or Findings: The total weekly energy consumption was 2305.1 kWh,  \nspread over 5 working days, from 7am to 7pm. The mo st performed \nexaminations were brain MRI (25 kWh/scan) and lumba r spine MRI (18 \nkWh/scan). Total emissions were 373.5 kgCO2e, equiv alent to emissions from \none homes' energy use for six months. Examinations performed between 4pm \nand 7pm have the highest carbon emission index (213 .4-252.3 gCO₂eq/kWh). \nRescheduling them earlier in the working day or at the weekend would reduce \nemissions by 18%, without any decrease of energy co nsumed. \nConclusion: This study shows how clinicians and managers can re duce \ncarbon emissions from MRI diagnostic activity, not only by reducing energy \nconsumption through cutting unnecessary examination s or optimising the \nacquisition technique, but also by rescheduling wor k activities according to \ndaily electricity emission data, scheduling examina tions during low-carbon \nemission times. \nLimitations: Short period of data collection limited the strengt h of the \nconclusions of this study. \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nMatteo Verga: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nGian Luca Viganò: Nothing to disclose \nSimone Zanoni: Nothing to disclose \n \n \n08:00-09:00 Research Stage 3 \nResearch Presentation Session: Oncologic \nImaging \nRPS 2216 \nStructured reporting, radiomics and deep \nlearning \n \nModerator \nM. E. Mayerhöfer; Vienna/AT  \n \n \nEvaluating the Impact of Structured Radiology Repor ting on Clinical \nPractice and Decision-Making: A Survey in a Large T ertiary University \nHospital – STAR Study \n*M. Mancino*, G. Avesani, A. Infante, S. Gaudino, B . Merlino, L. Natale,  \nE. Sala; Rome/IT \n(matteomancino@gmail.com) \n \nPurpose or Learning Objective: To evaluate the impact of structured \nradiology reports (SRRs) on clinical decision-makin g and patient management, \nspecifically focusing on improvements in diagnostic  accuracy, treatment \nplanning, patient outcomes, and data standardizatio n. Additionally, we aim to \nmeasure clinicians' satisfaction with the clarity, comprehensiveness, and utility \nof SRRs in comparison to standard narrative reports . \nMethods or Background: SRRs were introduced one and a half years ago at \nFondazione Policlinico Universitario A.Gemelli, a l arge tertiary university italian \nhospital , and have since been implemented in nearl y all radiological \nprocedures . This transition was the result of a co mprehensive collaboration \namong radiologists, surgeons, and clinicians to ens ure that the structured \ntemplates met clinical needs while adhering to guid elines from major \nradiological societies. An extensive and anonymous survey was conducted \namong non-radiologist clinicians, surgeons, and res idents from various \nspecialties to gather feedback on SRRs' clarity, cl inical impact, adaptability, \nresearch value, and efficiency compared to traditio nal reports. \nResults or Findings: Survey responses indicate increased clinician \nsatisfaction, improved communication, and greater w orkflow efficiency with \nSRRs. Preliminary findings suggest better data inte rpretation, ease of retrieval, \nand enhanced multidisciplinary discussions, particu larly in oncology. SRRs \nwere perceived as more effective than traditional r eports in supporting clinical \ndecisions and improving collaboration. \nConclusion: SRRs significantly improve clinical practice by del ivering clearer, \nmore consistent interpretations that directly enhan ce patient outcomes and \nsupport effective decision-making. The increased cl arity and standardization \nfoster better collaboration among clinicians, ultim ately benefiting patient care. \nContinuous feedback from users is essential to refi ne SRRs and ensure they \nremain adaptable and impactful. Furthermore, SRRs p rovide a foundation for \nconsistent data collection, crucial for advancing r esearch and supporting \nevidence-based practices. \nLimitations: The single-center survey limits the generalizabilit y of the results. \nFuture studies should include multiple centers to v alidate these findings. \nFunding for this study: No specific funding was obtained for this study. \nEthics committee - additional information: This study has been notified to \nthe Ethics Committee of Fondazione Policlinico Univ ersitario A. Gemelli \nIRCCS. \nAuthor Disclosures:  \nGiacomo Avesani: Nothing to disclose \nAmato Infante: Nothing to disclose \nSimona Gaudino: Nothing to disclose \nLuigi Natale: Nothing to disclose \nEvis Sala: Nothing to disclose \nMatteo Mancino: Nothing to disclose \nBiagio Merlino: Nothing to disclose \n \n \nUnraveling tumour heterogeneity with radiogenomics:  comparing single \ninstance and multiple instances learning AI approac hes \n*D. I. Rodríguez Sánchez *, R. Spaans, S. Rostami, O. Maxouri, Z. Bodalal,  \nR. G. H. Beets-Tan; Amsterdam/NL \n(i.rodriguez@nki.nl) \n \nPurpose or Learning Objective: Tumour genetic heterogeneity is a fact in \ncancer research. While biopsying every lesion is cl inically infeasible, imaging-\nbased methods (such as radiogenomics) promise clini cians non-invasive \ninsights into the underlying tumour biology. This s tudy assesses the impact of \naccounting for tumour heterogeneity by comparing Si ngle Instance Learning \n(SIL) and Multiple Instance Learning (MIL) AI appro aches. \nMethods or Background: A cohort of 1666 routine contrast-enhanced CT \nscans, including over 11.000 segmented lesions, was  retrospectively collected \nalong with matched next-generation sequencing data.  The morphological \nphenotype was quantified from each lesion by radiom ic features, with \nsubsequent feature selection using orthogonal princ ipal feature selection \n(OPFA) and five-fold cross-validation. SIL and MIL- based machine learning \nmethods were compared to predict the mutational sta tus of TP53, KRAS, and \nEGFR. Interlesional morphological heterogeneity was  measured using spatial \ndistance metrics. \nResults or Findings: For genes with established low biological variabili ty \nbetween lesions (TP53 and KRAS), accounting for tum our heterogeneity did \nnot improve the radiogenomic predictive performance . However, for genes with \nhigh discordance (EGFR), MIL-based machine learning  methods (AUC \nrange=0.72-0.78) performed significantly better tha n SIL (AUC=0.56). These \nfindings are also supported by interlesional hetero geneity scores, which did not \ndiffer between wild-type and mutated TP53 or KRAS c ases. Conversely, \nEGFR-mutated patients demonstrated significantly hi gher interlesional \nmorphological heterogeneity than their wild-type co unterparts (p<0.0001). \nConclusion: MIL models may better reflect tumour heterogeneity,  particularly \nin cases with high biological variability. Incorpor ating MIL into radiogenomic \nmodels may enhance their predictive accuracy by acc ounting for real-world \ntumour heterogeneity. \nLimitations: The study's limitations include external validation . A large-scale \nmulticenter radiogenomics dataset is currently bein g finalised to validate these \nresults. \n \n\n \n \nSunday \nAbstract-based Programme \n \n 266  \nFunding for this study: Funding was provided by the European Union’s \nHorizon 2020 research and innovation program under the Marie Skłodowska-\nCurie grant agreement number 101034290 (EMERALD Int ernational PhD \nProgram for Medical Doctors). \nEthics committee - additional information: IRB approval: IRBd19-147 \nAuthor Disclosures:  \nZuhir Bodalal: Nothing to disclose \nRobert Spaans: Nothing to disclose \nDiana Ivonne Rodríguez Sánchez : Nothing to disclos e  \nSajjad Rostami: Nothing to disclose  \nRegina G. H. Beets-Tan: Nothing to disclose \nOlga Maxouri: Nothing to disclose \n \n \nEnhancing speed and precision of lesion tracking in  follow-up lung CT \nusing deep-learning-based registration \n*S. Kuckertz*¹, S. Heldmann¹, F. Peisen², J. H. Mol tz³; ¹Lübeck/DE, \n²Tübingen/DE, ³Bremen/DE \n \nPurpose or Learning Objective: Continuous lesion assessment in cancer \npatients is integral to radiologists’ work. Part of  this process is the tedious and \ntime-consuming (re-)localisation and measurement of  lesions. Fast and precise \nimage registration facilitates the workflow by auto matically matching previous \nand current observations. In this study we evaluate  our deep-learning-based \nlung registration using a longitudinal dataset incl uding expert lesion \nsegmentations, comparing it to a state-of-the-art c onventional non-learning \napproach. \nMethods or Background: We follow a 3-level coarse-to-fine deep-learning \nregistration approach. At each level, we input the baseline and follow-up CT \nscan at a different resolution into a U-Net, result ing in a deformation field that \nmaps corresponding locations from baseline to follo w-up. Combining the \ndeformations from all levels allows accurate tracki ng of anatomical points. Our \nmethod is trained on 681 follow-up image pairs and evaluated on a distinct \ndataset consisting of 90 image pairs with 307 manua lly segmented lung \nlesions. We compare our approach to a non-learning GPU-accelerated \nregistration. Each lesion centre in the baseline sc an is propagated onto the \nfollow-up, where we check whether it maps inside th e given corresponding \nlesion. \nResults or Findings: With our learning-based approach, 81.1% of the \nbaseline lesion centres were correctly mapped to th e corresponding lesion in \nthe follow-up (conventional approach: 73.6%). The m edian distance between \nthe propagated and the given lesion centre was 1.9 mm (conventional \napproach: 3.2 mm) and the average calculation time was 0.92 s (conventional \napproach: 14.56 s). \nConclusion: Our learning-based registration approach enhances b oth speed \nand accuracy, enabling precise relocation of all lu ng lesions in follow-up scans \nin less than a second. This facilitates radiologist s’ workflow, also enabling \ncursor synchronisation and change highlighting in v iewers. \nLimitations: CTs were cropped to the thorax area and resampled t o an \nisotropic resolution of 1.5 mm. \nFunding for this study: Funding was provided by the Federal Ministry of \nEducation and Research of Germany (BMBF) as part of  SPIRABENE (project \nnumber 13GW0561B). \nEthics committee - additional information: Only retrospective data was used \nfor this work. The outcome had no effect on patient  treatment. \nAuthor Disclosures:  \nSven Kuckertz: Nothing to disclose \nFelix Peisen: Nothing to disclose \nStefan Heldmann: Nothing to disclose \nJan Hendrik Moltz: Nothing to disclose \n \n \nRadiomic gradient in the peritumoral tissue of live r metastases: A \nbiomarker to drive clinical practice? \n*A. Ammirabile*¹, F. Fiz², E. M. Ragaini¹, S. Sirch ia³, S. Viganò¹, M. Francone¹, \nL. Cavinato¹, E. Lanzarone³, L. Viganò¹; ¹Milan/IT,  ²Genova/IT, ³Bergamo/IT \n(angela.ammirabile@humanitas.it) \n \nPurpose or Learning Objective: To investigate the variation of three textural \nfeatures (mean HU, entropy, and uniformity) in the peritumoral tissue around \ncolorectal liver metastases (CRLM) as distance from  the tumor increases. \nMethods or Background: This retrospective study included all consecutive \npatients with histologically proven CRLM who underw ent locoregional \ntreatment (resection/ablation) between January 2010  and December 2022. \nInclusion criteria were high-quality CT with an ade quate portal phase and \nidentifiable hypodense CRLM (>10 mm). Multiple VOIs  were generated: 1) \nmanual tumor segmentation (Tumor-VOI); 2) multiple automatic concentric rims \nat increasing distance from CRLM (1 to 10 millimete rs); 3) manual \nsegmentation of a virtual biopsy of non-tumoral par enchyma (Liver-VOI). \nRadiomic features were extracted by the LifeX softw are. The percentage \nvariation of index values across different VOIs was  calculated, using Liver-VOI \nas reference. Subgroup analyses were based on tumor  size (10-30 vs. >30 \nmm) and chemotherapy administration (no chemotherap y vs. responders). \nResults or Findings: 63 CRLM in 51 patients (median age 67 years, 14 \nfemales) were analyzed. Median peritumoral HU value s were similar to Liver-\nVOI, except within the 1-mm VOI around CRLM (p=0.00 2). Entropy \nprogressively decreased (from 3.11 of CRLM to 2.54 of Liver-VOI, p<0.001) \nwhile uniformity increased (from 0.135 of CRLM to 0 .199 of Liver-VOI, \np<0.001). At 10 mm from CRLM, entropy was similar t o Liver-VOI in 62% of \ncases and uniformity in 46%. Smaller CRLM and respo nders to chemotherapy \nshowed higher and earlier normalization of entropy and uniformity values. \nConclusion: The radiomic analysis of peritumoral tissue in CRLM  \ndemonstrated a peculiar gradient of decreasing entr opy and increasing \nuniformity despite a normal radiological appearance , representing a potential \nbiomarker for personalized clinical decision-making . \nLimitations: Retrospective analysis; small cohort; heterogeneous  CT data; \nmissing correlation with pathologic/surgical data. \nFunding for this study: AIRC grant #2019−23822 \nEthics committee - additional information: Protocol 988/22 \nAuthor Disclosures:  \nSara Sirchia: Nothing to disclose \nAngela Ammirabile: Nothing to disclose \nEttore Lanzarone: Nothing to disclose \nFrancesco Fiz: Nothing to disclose \nSamuele Viganò: Nothing to disclose \nElisa Maria Ragaini: Nothing to disclose \nMarco Francone: Nothing to disclose \nLara Cavinato: Nothing to disclose \nLuca Viganò: Nothing to disclose \n \n \nIntegrating MRI and PET/CT Radiomics for Enhanced S urvival Prediction \nin Esophageal Cancer \n*C. Noirot*, D. Abler, L. Haefliger, S. Mantziari, M. Schäfer, N. Vietti Violi,  \nA. Depeursinge, C. Dromain, M. Jreige; Lausanne/CH \n(camille.noirot@chuv.ch) \n \nPurpose or Learning Objective: Prognosis evaluation in esophageal cancer \nremains challenging. Accurate survival prediction i s crucial for treatment \nplanning and follow-up strategies. Although MRI and  18F-FDG PET/CT provide \nvaluable information, they have limitations in accu rately predicting patient \noutcomes. This study aimed to develop radiomics mod els based on MRI and \nPET/CT to predict overall survival in esophageal ca ncer patients using \nbaseline and follow-up imaging. \nMethods or Background: Sixty patients (M/F: 50/10, mean age 66±9 years) \nwith newly diagnosed esophageal cancer were prospec tively included (2017-\n2022). Patients underwent staging with 18F-FDG PET/ CT and MRI, with follow-\nup MRI after neoadjuvant treatment. Tumors were man ually segmented using \nMint™ Software, and radiomics features were extract ed via QuantImage v2 \nplatform. The dataset, including 645 features from MRI and PET/CT, was split \ninto training (80%) and test (20%) sets. Various su rvival prediction algorithms \nwere compared. Model performance was assessed with the concordance index \n(C-index) using bootstrapping for confidence interv al (CI) estimation. \nResults or Findings: Radiomics features were analyzed at baseline from b oth \nPET/CT and MRI for 52 patients, and at follow-up MR I for 49 patients. Mean \nsurvival was of 37 months (range: 1.8 to 78.1). The  MRI model (14 features) \nachieved a C-index of 0.733 (95% CI: 0.718–0.756), and the PET/CT model (5 \nfeatures) achieved 0.724 (95% CI: 0.707–0.746) for predicting OS. A combined \nmodel with 19 features improved the C-index to 0.86 8 (95% CI: 0.853–0.881), \nwhile a follow-up MRI model (16 features) reached 0 .807 (95% CI: 0.790–\n0.827). \nConclusion: The radiomics model based on MRI and PET/CT demonst rated \nrobust performance in predicting survival for esoph ageal cancer patients. \nIntegrating multi-modal baseline and follow-up imag ing radiomics features into \nsurvival models could enhance prognostic accuracy, improving personalized \nmanagement strategies in esophageal cancer. \nLimitations: The limitations of the study are the number of pati ents. \nFunding for this study: No funding was received. \nEthics committee - additional information: The study was approved by \n\"Commission Cantonale d'Ethique de la Recherche sur  l'être humain\" à \nLausanne (CER-VD 2017-00388) \nAuthor Disclosures:  \nMarkus Schäfer: Nothing to disclose \nCamille Noirot: Nothing to disclose \nAdrien Depeursinge: Nothing to disclose \nLaura Haefliger: Nothing to disclose \nDaniel Abler: Nothing to disclose \nStyliani Mantziari: Nothing to disclose \nNaïk Vietti Violi: Nothing to disclose \nMario Jreige: Nothing to disclose \nClarisse Dromain: Nothing to disclose \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 267  \nAI-Assisted Annotation for Improving Federated Lear ning in Automated \nRCC Image Segmentation \nK. S. Younis¹, *J. Garrett*², A. Elhanashi³, A. Gen tili⁴, S. Faghani⁵, S. Kuanar⁵, \nY. Singh⁵, Y. Huo⁶, G-M. Conte⁵, J. Yacoub⁷, O. Unal²; ¹Cleveland, OH/US, \n²Madison, WI/US, ³Pisa/IT, ⁴San DIego, CA/US, ⁵Rochester, MN/US, \n⁶Nashville, TN/US, ⁷Washington, DC/US \n(jgarrett@uwhealth.org) \n \nPurpose or Learning Objective: To evaluate the impact of AI-assisted \nannotation tools on improving consistency in datase t labeling for federated \nlearning, focusing on image segmentation tasks for renal cell carcinoma (RCC) \nin CT and MR images. \nMethods or Background: One of the major challenges in distributed learning  \nis variability in data labels across sites, especia lly in image segmentation tasks \nwhere mask generation methods can differ significan tly. This study leverages a \npublicly available dataset from The Cancer Imaging Archive (TCGA-KIRC) for \nRCC, which contains rich imaging and clinical data.  AI-assisted annotation \ntools standardize data labeling before training mod els such as Unet and Swin \nUNETr. The primary goal is to assess if these tools  enhance model \nperformance or reduce the number of cases needed fo r effective training in \nfederated learning environments. Radiologists from various institutions across \nmultiple states in the US manually annotated the im ages and federated \nlearning was conducted using NVIDIA nvFLARE. \nResults or Findings: Preliminary results suggest that AI-assisted annota tion \nimproves model consistency, with segmentation accur acy increasing by \napproximately 12% when compared to non-standardized  data labels. The \nmodel efficiency was also reflected in reduced data  redundancy and higher \nannotation agreement between annotators. This enhan cement allowed for \nmore precise radiomic analysis across datasets. Add itionally, the number of \nsamples required for training decreased by 20%, ind icating the efficiency of AI-\nassisted annotation in generating reliable training  datasets. \nConclusion: AI-assisted annotation holds promise for improving performance \nand efficiency in federated learning environments, particularly for automated \nimage segmentation tasks like RCC detection. The en hanced consistency in \nlabel generation helps to reduce the variability in troduced by multiple sites, \nthereby improving model generalizability. \nLimitations: The study does not include external validation acro ss diverse \nimaging platforms, and annotation tools were not ev aluated for real-time \nperformance during federated learning. \nFunding for this study: Nvidia Education grant \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nYashbir Singh: Nothing to disclose \nAbdussalam Elhanashi: Nothing to disclose \nOrhan Unal: Nothing to disclose \nAmilcare Gentili: Nothing to disclose \nShahriar Faghani: Nothing to disclose \nJoseph Yacoub: Consultant: Microsoft Nuance \nKhaled Salem Younis: Nothing to disclose \nYuankai Huo: Nothing to disclose \nShiba Kuanar: Nothing to disclose \nGian-Marco Conte: Nothing to disclose \nJohn Garrett: Grant Recipient: Research Support fro m the NIH, GE Healthcare, \nand the American Cancer Society Advisory Board: Rad Unity Corp. \nShareholder: NVIDIA, Inc. and RadUnity Corp. \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n08:00-09:00 Research Stage 4 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 2205 \nImaging informatics, quality and new \ntechniques \n \nModerator \nL. Lofino; Milan/IT  \n(Ludovica.Lofino@humanitas.it) \n \n \nUniversal Medical Imaging Encoding Datasets - A new  standard for \ncreating large-scale diagnostic imaging datasets \n*B. O. Klaudel*¹, A. Obuchowski², A. Komor¹, P. Fr ąckowski¹, K. Rogala¹,  \nK. Knitter¹; ¹Gdańsk/PL, ²Banino/PL \n(klaudel.b@gmail.com) \n \nPurpose or Learning Objective: We present a new standard for creating and \nunifying large-scale diagnostic imaging datasets. I t aims to address the \nchallenges in medical AI, particularly the lack of standardized, comprehensive \ndata for training foundation models in medical imag ing. By combining multiple \nopen-source datasets, unifying them to a common ont ology and providing \nstandardized preprocessing pipelines, we seek to ac celerate the development \nof more generalized and robust medical AI models. \nMethods or Background: We combined over 20 open-source datasets, \nresulting in more than one million annotated medica l images, including CT, \nMRI, and X-ray modalities. We created reusable prep rocessing pipelines to \ntransform diverse source datasets into a unified fo rmat. A key innovation is the \nadoption of the RadLex ontology, developed by the R adiological Society of \nNorth America, to standardize labels and masks acro ss all included datasets. It \naddresses issues of inconsistent labeling, regional  variations, and differing \ngranularity in existing medical imaging datasets. \nResults or Findings: The project has created the largest publicly availa ble \ndataset of annotated radiological imaging to date, with over one million images, \n40+ labels, and 15 annotation masks. The datasets a re accompanied by ready-\nto-use preprocessing pipelines that can be easily a dapted to incorporate new \ndata sources. The unified ontology based on RadLex enables consistent \nlabeling, facilitating more effective model trainin g and cross-dataset \ncompatibility. \nConclusion: UMIE datasets represent a significant advancement i n medical \nimaging AI, providing a standardized foundation for  developing more \ngeneralized models. By addressing the challenges of  data scarcity, \ninconsistent formatting, and labeling discrepancies , it paves the way for more \nrobust and widely applicable AI solutions The open- source nature of the project \nencourages collaboration and further expansion of t he datasets. \nLimitations: UMIE datasets inherits potential biases and incompl ete labelling \ncoverage from source datasets, and limitations of t he RadLex ontology. \nFunding for this study: No funding was required. \nEthics committee - additional information: All data used in this study come \nfrom existing opensource datasets. \nAuthor Disclosures:  \nAndrzej Komor: Nothing to disclose \nKacper Rogala: Nothing to disclose \nBarbara Olga Klaudel: Nothing to disclose \nAleksander Obuchowski: Nothing to disclose \nKacper Knitter: Nothing to disclose \nPiotr Frąckowski: Nothing to disclose \n \n \nAnalysis of key principles for improving the effici ency of medical data \nannotation processes for machine learning \n*P. Pilius*¹, N. Smirnov²; ¹Almaty/KZ, ²Haar/DE \n(polishka1903@gmail.com) \n \nPurpose or Learning Objective: The aim of this work is to analyze the \nfundamental principles of project management and ev aluate potential \nopportunities for its implementation for improving efficiency in the preparation \nof medical data for machine learning: for data coll ection, working with \nannotators, and results validation. \nMethods or Background: We analyzed personal experiences from 9 projects \nconducted during 2022-2024 years, aimed at creating  annotated medical \ndatasets with various levels of complexity and acro ss different fields. The \nclients were private companies developing artificia l intelligence technologies, \nand the annotators included certified radiologists,  residents, and radiographers. \n\n \n \nSunday \nAbstract-based Programme \n \n 268  \nTo enhance efficiency, we studied and applied princ iples from an advanced \ntraining course on the foundations of project manag ement. \nResults or Findings: The duration of the projects ranged from 3 weeks to  3 \nmonths. The number of annotated datasets included: 75 MRI, 3720 CT, 55,500 \nX-rays, 6000 angiograms, 1700 mammograms, and 3200 radiological text \nreport annotations. All projects were completed on time and met quality \nrequirements. Based on practical experience and the  implementation of \nfundamental project management concepts, we've form ulated \nrecommendations and standards for conducting simila r projects. \nConclusion: The organization of medical data annotation process es for \nmachine learning presents numerous specific challen ges. To optimize the \nworkflow and achieve desired outcomes, it is essent ial to have not only \nsufficient experience in handling medical images bu t also a solid understanding \nof fundamental project management principles. \nLimitations: The work is based on personal experience and observ ations and \nis primarily of a recommendatory nature. It is not possible to objectively \ncompare the outcomes of projects where basic manage ment principles were \nand were not applied, as the projects were not repe ated under identical \nconditions. However, the variability and uniqueness  of each project allowed for \na comprehensive analysis, leading to the formulatio n of universal standards. \nFunding for this study: No funding \nEthics committee - additional information: Kazakh National Scientific and \nResearch Center of Oncology and Radiology \nAuthor Disclosures:  \nPolina Pilius: Nothing to disclose \nNikolai Smirnov: Nothing to disclose \n \n \nData Interoperability in a Clinical Pathway From Fr ee-Text Reports \n*K. Nairz*, N. Cihoric, F. Dennstädt, M. Schmerder,  H. Bonel,  \nH. Von Tengg-Kobligk; Bern/CH \n(knud.nairz@insel.ch) \n \nPurpose or Learning Objective: Structured reporting (SR) in radiology has \nbeen shown to be the most favorable scheme to provi de imaging information to \nreferrers. Implementation of SR is associated with additional effort for the \nradiologists, but there are recent advances with La rge Language Models \n(LLMs) that are prompted to convert free text into a structured format. We \naimed at killing two birds with a stone by leveragi ng LLMs to generate \nstructured data that enhances interoperability as w ell. As a proof of concept we \nwe selected breast cancer patient pathways for docu mentation. We structured \nthe free-text information from various sources, inc luding health interviews, \nmammography reports, biopsy results, pathology find ings, and tumor board \ndiscussions, to ensure that the data could be effec tively transmitted and \nsupport therapeutic decision-making. \nMethods or Background: Our approach is based on the use of Common Data \nElements (CDEs), which are minimal information unit s or precisely defined \nquestions associated with a set of standardized ans wers, each having explicitly \ndefined values. Building on in-depth analyses of re ports and guidelines such as \nBI-RADS, we defined corresponding sets of Common Da ta Elements and \ncreated templates to establish a structured format.  To ensure data protection \nwe utilized a locally installed LLM (Llama 3), whic h was prompted to answer \nCDE-based questions. This process enabled the mappi ng of free-text content \nto a structured format, which was ultimately stored  in FHIR compatible JSON \nformat. \nResults or Findings: By defining key CDE values, or specific combination s of \nthese values, clinicians can identify critical insi ghts that may suggest a \nparticular therapeutic course or provide predictive  indicators for patient \noutcomes. \nConclusion: Prompting LLMs to answer CDE-based structures prove s to be a \nviable approach to promote data interoperability in  complex medical settings. \nLimitations: The study focuses specifically on breast cancer pat hways. \nFunding for this study: Innosuisse 59228.1 \nEthics committee - additional information: Kantonale Ethikkommission Bern \nAuthor Disclosures:  \nHarald Bonel: Nothing to disclose \nMax Schmerder: Nothing to disclose \nNikola Cihoric: Nothing to disclose \nKnud Nairz: Nothing to disclose \nFabio Dennstädt: Nothing to disclose \nHendrik Von Tengg-Kobligk: Nothing to disclose \n \n \nDeep learning based automated field of view positio ning for prostate \nmagnetic resonance imaging \n*A. S. Quinsten*¹, A. Wetter², M. Raczkowski³, L. T rembecki³, R. Buchkremer⁴, \nD. Matusiewicz¹, K. Nassenstein¹, M. Forsting¹, A. Demircioglu¹; ¹Essen/DE, \n²Hamburg/DE, ³Wrocław/PL, ⁴Düsseldorf/DE \n \nPurpose or Learning Objective: Prostate magnetic resonance imaging (MRI) \nis typically conducted according to manual prescrip tions by radiographers. This \napproach is time-consuming, error-prone, inconsiste nt due to rater variability, \nand has low reproducibility. The aim of the study w as to develop a deep \nlearning-based framework for the automatic planning  of the field of view (FoV) \nin the oblique coronal and axial planes in prostate  MRI according to Prostate \nImaging Reporting and Data System (PI-RADS) guideli nes. \nMethods or Background: The retrospective multicentre study included 2109 \npatients from diagnostic (Sites I and III) and radi otherapy (Site II) centres. The \nvariability within and between raters was evaluated  by three assessors. Three \ndistinct deep neural networks were developed with t he objective of predicting \nthe oblique coronal and axial FoV. The optimal netw ork was evaluated on three \nexternal cohorts using a non-inferiority test, and its clinical utility was assessed. \nResults or Findings: The optimal model demonstrated non-inferior \nperformance, with slice position differences rangin g from 0.21 ± 0.99 and 0.37 \n± 0.48. At Sites I and III, the predictions were predominantly non-inferior, with \nFoV overlaps of 86.6 ± 5.8% and 88.7 ± 6.0% and angle differences of 4.66 ± \n4.89° (Site I) as well as 3.46 ± 2.80° (Site III). In contrast, the predictions for \nSite II demonstrated inferior overlap (67.0 ± 9.7% and 63.6 ± 8.8%) and higher \nangle differences (9.18 ± 9.49°). Consequently, the clinical utility was excellent \nfor Sites I and III (97.9–100%) but lower for Site II (85.3–89.0%). \nConclusion: The utilisation of a deep learning-based framework for the \nautomated positioning of the FoV in oblique coronal  and axial planes for \nprostate MRI is a viable approach, exhibiting high clinical utility. \nLimitations: The present study did not include images acquired w ith the \nendorectal coil. \nFunding for this study: No funding was provided for this study \nEthics committee - additional information: The ethics committee notification \ncan be found under the number 22-10740-BO. \nAuthor Disclosures:  \nMichael Forsting: Nothing to disclose \nKai Nassenstein: Nothing to disclose \nAnton Sheahan Quinsten: Nothing to disclose \nRüdiger Buchkremer: Nothing to disclose \nDavid Matusiewicz: Nothing to disclose \nAxel Wetter: Nothing to disclose \nLukasz Trembecki: Nothing to disclose \nAydin Demircioglu: Nothing to disclose \nMaciej Raczkowski: Nothing to disclose \n \n \nRADAR - real-time automated detection and analysis of radiopaque \ndevices using CT topograms \n*C. S. Schmidt*, M. Walter, J. Haubold, F. Nensa, R . Hosch; Essen/DE \n \nPurpose or Learning Objective: The aim of this study was to develop a deep \nlearning (DL) model for the automatic detection and  localisation of medical \ndevices known to cause metal artefacts in CT images , utilising their \ncorresponding topograms. \nMethods or Background: A dataset of 943 CT topograms with radiopaque \nmedical devices was manually annotated via box labe lling by a radiology \nresident with three years of experience in CT imagi ng. The following classes \nwere defined: cochlear implant, cardiac conduction device (pacemaker, \ndefibrillator, stimulator), implanted port, prosthe tic heart valve, (embolisation) \ncoil, osteosynthesis (nail-, plate-, screw-, and wi re-fixation, spinal \ninstrumentation hardware), sternal wires, external fixation hardware, hip \nprosthesis, shoulder prosthesis, knee prosthesis, d enture (prosthesis, implant). \nAn 80/10/10% split for training, validation and tes ting was performed and the \nYOLO11X model was trained for 100 epochs. The model  was evaluated using \nmAP50 scores, precision (P) and recall (R). \nResults or Findings: The model achieved an average mAP50 score of 0.83, \nPrecision of 0.86 and Recall of 0.79 over all class es and the following \n(mAP50/P/R) scores for the respective classes: coch lear implant \n(0.92/0.93/0.85), cardiac conduction device (0.90/0 .82/0.93), implanted port \n(0.91/0.97/0.79), prosthetic heart valve (0.86/0.82 /0.78), coil (0.99/0.98/1), \nosteosynthesis (0.63/1/0.55), sternal wires (0.51/0 .74/0.57), external fixation \nhardware (0.52/0.51/0.33), hip prosthesis (0.99/0.8 9/1), shoulder prosthesis \n(0.88/0.88/0.86), knee prosthesis (0.99/0.95/1), de nture (0.88/0.81/0.76). \nConclusion: The presented model demonstrates an accurate detect ion of \nmost radiopaque medical devices in CT scout images.  It could thus be utilised \nas an efficient orchestration tool for selecting a cohort of high quality imaging \nstudies without interfering artefacts. \nLimitations: The limitations of the study are its small sample s ize and that \nscout images were annotated by a single observer. A dditionally, certain \nmedical devices can be challenging to identify and localise on topograms, \nwhich could cause relevant features to go undetecte d. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Informed consent was waived by \nthe ethics committee due to the retrospective setti ng. \nAuthor Disclosures:  \nJohannes Haubold: Nothing to disclose \nCynthia Sabrina Schmidt: Nothing to disclose \nMarie Walter: Nothing to disclose \nRené Hosch: Nothing to disclose \nFelix Nensa: Nothing to disclose \n\n \n \nSunday \nAbstract-based Programme \n \n 269  \nEvaluating the Impact of Quantum Technology on Radi omics: A \nComparative Study of Classical and Quantum Random F orest Models \n*F. Mariotti*, A. Agostini, A. Borgheresi, L. Pierp aoli, F. Ricciardiello,  \nA. Zannotti, D. Nicolini, A. B. Galosi, A. Giovagno ni; Ancona/IT \n(Dottfrancescomariotti@gmail.com) \n \nPurpose or Learning Objective: This study aims to evaluate the impact of \nquantum technology on the performance of radiomics random forest (RF) \nmodels for medical imaging. We simulated a semi-qua ntum approach, \ninvolving quantum embedding followed by classical R F, and a fully quantum \napproach using a quantum random forest (QRF) model.  \nMethods or Background: We used three radiomic datasets: 1. Perineural \ninfiltration of peripancreatic fat in pancreatic ad enocarcinoma on CT, 2. \nCharacterization of renal nodules in CT, 3. Predict ion of LI-RADS category on \nabbreviated MRI protocols. For the quantum approach es, we compared the \noriginal random forest (RF) models with simulated q uantum-embedded RF and \nquantum random forest (QRF) algorithms, implemented  in Python using an 8-\nqubit configuration. The comparison involved analyz ing the accuracy and the \nreceiver operating characteristic (ROC) curves usin g statistical significance set \nat p-values < 0,05 \nResults or Findings: The classical RF achieved the highest accuracy for the \npancreas (0.9167) and kidney (0.8571) datasets. For  the liver dataset, both the \nquantum embedding RF and QRF outperformed the class ical approach \n(0.8462 vs. 0.7692), with the ROC curves showing st atistically significant \nimprovement (p < 0.01). In the pancreas dataset, qu antum methods showed \nslightly lower accuracy (0.8333), and for the kidne y, they also performed worse \n(0.7857). This indicates that the benefits of quant um approaches may be data-\ndependent, providing advantages in some cases but n ot yielding consistent \nimprovements across all datasets. \nConclusion: Quantum machine learning is a feasible approach for  radiomic \ndatasets, showing variable results and the potentia l to outperform classical \nmethods. However, the variability in performance su ggests that fine-tuning of \nquantum algorithms may be necessary depending on th e specific \ncharacteristics of each dataset. \nLimitations: Small datasets used and simulation of quantum proce sses with a \n8-qubit setup. Further research should involve larg er datasets and physical \nquantum devices. \nFunding for this study: This study did not receive any specific funding fro m \npublic, commercial, or not-for-profit sectors. The research was conducted \nwithout external financial support. \nEthics committee - additional information: Not Applicable \nAuthor Disclosures:  \nDaniele Nicolini: Nothing to disclose \nAlice Zannotti: Nothing to disclose \nLuca Pierpaoli: Nothing to disclose \nAndrea Agostini: Nothing to disclose \nFrancesco Ricciardiello: Nothing to disclose \nFrancesco Mariotti: Nothing to disclose \nAndrea Benedetto Galosi: Nothing to disclose \nAlessandra Borgheresi: Nothing to disclose \nAndrea Giovagnoni: Nothing to disclose \n \n \nA new framework for 3D data representation in Exten ded Reality (XR) on \niPhone, iPad and Apple Vision Pro \n*A. M. C. Boehner*, A. Jacob, A. Isaak, C. C. Piepe r, J. A. Luetkens,  \nD. Kütting; Bonn/DE \n(boehner.amc@gmail.com) \n \nPurpose or Learning Objective: 3D data is rarely spatially displayed in \nroutine. However, patient-clinician and clinician-c linician interaction may benefit \nfrom such representation in Extended Reality (XR). Additionally, radiologists \nmay aid surgeons during surgery via audiovisual com munication to \ndemonstrate 3D data if needed. \nMethods or Background: We developed and tested a workflow integrating \ndifferent software platforms (e.g.‘Medical Imaging XR’, ‘Fiji’) to display DICOM \nimages on iPhone and iPad (n=35) and Apple Vision P ro (AVP, n=10). The \nsystem enables fused XR visualization of CT, MRI, P ET. Handheld devices \nwere utilized to aid sonographic correlations of he patic lesions (n=11); by \nsurgeons during surgery preparations (n=10); and fo r patient information \n(n=14). Integrated systems were tested in a mock au diovisual call from the \noperating room via the AVP to the other devices loc ated on and off campus. \nResults or Findings: Our framework allowed for fast integration of 3D \ndatasets across devices with low computational burd en. XR during \nsonographic correlation of hepatic lesions signific antly reduced the time \nneeded to identify lesions from 4:50min to 2:45min (P<0.05). Patient reported \nfull acceptance of XR usage. AVP allowed real-time image-data and view \nsharing between the radiologist and surgeon. \nConclusion: Integration of XR across smartphones, tablets and A VP \nenhanced medical imaging communication between all parties, reducing time \nto locate lesions and improving patient-physician i nteractions. AVP further \nfacilitates sterile audiovisual communication betwe en surgeons and \nradiologists during procedures, allowing for remote  and swift consultation \nwithout leaving the sterile field. \nLimitations: Our method was tested exclusively on Apple products , limiting its \ngeneralizability to other platforms. \nFunding for this study: This project was part of the ISMC, funded by the \nMinistry of Economic Affairs, Innovation, Digitaliz ation and Energy of the state \nof North Rhine-Westphalia \nEthics committee - additional information: Ethics University Hospital Bonn, \nGermany (2024-228-BO) \nAuthor Disclosures:  \nJulian Alexander Luetkens: Nothing to disclose \nAlexander Isaak: Nothing to disclose \nAlexander Marc Christian Boehner: Nothing to disclo se \nAlice Jacob: Nothing to disclose \nDaniel Kütting: Nothing to disclose \nClaus Christian Pieper: Nothing to disclose \n \n \n08:00-09:00 Room G1 \nResearch Presentation Session: \nRadiographers \nRPS 2214 \nAdvancing radiography through education \nand research: innovations, challenges, \nand future directions \n \nModerators \nB. Horehledova; Heerlen/NL  \n(barbora.horehledova@gmail.com) \nJ. Santos; Coimbra/PT \n(joanasantos@estesc.ipc.pt) \n \n \nThe College of Radiographers’ Education and Career Framework (fourth \nedition): Exploring the guideline implementation ga p across England \nusing Normalisation Process Theory \n*H. L. Spencer*¹, K. Williamson², A. Robertson², M.  N. K. Anudjo¹, C. Burton³; \n¹Bournemouth/UK, ²London/UK, ³Norwich/UK \n(HSpencer@aecc.ac.uk) \n \nPurpose or Learning Objective: In 2022, the College of Radiographers (CoR) \npublished the fourth edition of their Education and  Career Framework (ECF). \nThis essential document provides a professional blu eprint for the radiography \ncareer trajectory with the overarching aim of impro ving patient outcomes. \nHowever, publication does not guarantee implementat ion; there often exists a \ndissonance between policy intent and policy in-acti on. Therefore, if we are to \naccess the full benefits of the ECF, its implementa tion requires careful \nconsideration. \nMethods or Background: To advance our understanding of the translational \ngap between policy and practice, this observational  mixed-methods study \nemployed Normalisation Process Theory (NPT) as a th eoretical frame. \nFocusing on the diagnostic radiography profession i n England, a national \nconsultation survey was deployed, alongside four co nsultation workshops. The \ndata collection methods were underpinned by NPT. Th e framework approach \nwas adapted for the qualitative data analysis. The quantitative survey data, \nmeanwhile, was analysed using descriptive and infer ential statistics. \nResults or Findings: The data collection took place between April-June 2 023. \nIn total, 142 survey responses were returned. Each workshop was comprised \nof 7-11 participants. The findings were deductively  interpreted through the lens \nof NPT, from which five core themes emerged: making  sense of complexity \n(coherence); bringing people together (cognitive pa rticipation); being strategic \n(collective action); evaluating complexity (reflexi ve monitoring); implementation \nin the ‘real world’ (barriers and enablers). \nConclusion: By furthering our understanding of the guideline im plementation \ngap, it was then possible to propose recommendation s to enhance the ECF’s \nadoption. The recommendations were study-derived, l inked to responsible \nstakeholders, and grouped into four strategic prior ities, aligned with the NPT \ndomains. Through these evidence-based recommendatio ns, it is hoped the \nECF can be translated more fully from page to the ‘ real world’ for the benefit of \nthe profession and its service users. \nLimitations: Nonapplicable. \nFunding for this study: This study was undertaken as part of a Clinical \nEducation Improvement Fellowship secondment, suppor ted by NHS England \n(South East) Workforce, Training, and Education, Ca nterbury Christ Church \n\n \n \nSunday \nAbstract-based Programme \n \n 270  \nUniversity, and the Florence Nightingale Foundation . However, no direct \nfunding was received for this study. \nEthics committee - additional information: Canterbury Christ Church \nUniversity Faculty of Medicine, Health, and Social Care Ethics Panel \n(Reference: ETH2223-0262). \nAuthor Disclosures:  \nChristopher Burton: Nothing to disclose \nMessiah Narh Kwame Anudjo: Nothing to disclose \nHolly Louise Spencer: Nothing to disclose \nKathryn Williamson: Employee: Society and College o f Radiographers \nAmy Robertson: Employee: Society and College of Rad iographers \n \n \nNewly qualified radiographers' perception of the in duction programme in \na radiology department: a survey study \n*L. Bombelli*, G. R. Bonfitto, A. Roletto, E. Scara melli, S. V. Fasulo,  \nD. Catania; Milan/IT \n(bombelli.luca@virgilio.it) \n \nPurpose or Learning Objective: The role of radiographers is rapidly evolving, \nputting them in a key role in a contest of increasi ng complexities in patient \ncare. With the rising demand for specialized skills , it is crucial to implement \nstructured induction programs for newly qualified r adiographers (NQR), also to \navoid poor workplace performance and even the dismi ssal of workers. This \nstudy aims to explore the perceptions of NQR who ha ve already experienced \nan induction programme. \nMethods or Background: A survey was distributed to NQR in a large \nuniversity hospital in Italy. The questionnaire gat hered demographic data and \nby using a 5-point Likert scale assessed 29 sentenc es concerning issues in \ntheir work, including patient management, decision- making, work organization, \nself-confidence development and relationships with other team members. \nResults or Findings: Twenty-two NQR participated in the survey. Among \nthese, 32% (n=7) graduated within the last year and  46% (n=10) reported that \nthey had their first experience in a healthcare set ting. Regarding Likert scale \nevaluation, only 10 out of 29 sentences received sc ores of 4 or higher, \nindicating “Agreement”. Participants with prior wor k experiences reported \ngreater self-confidence in their skills once the in duction program has been \ncompleted. Conversely, radiographers with no work e xperience indicated \nfeeling more supported by management. \nConclusion: In conclusion, NQR involved in this study felt adeq uately \nprepared for clinical practice after completing the  induction program, despite \nsome differences between radiographers with differe nt levels of experiences. A \nsuccessful induction program for NQR is essential t o foster a proactive \nmindset, promote appropriate work methods, enhance collaboration among \nteam members, reduce radiographers’ stress, turnove r and ensure a high \nquality of patient care. \nLimitations: Quantitative study design and limited sample size m ay have \ncaused limitations. \nFunding for this study: No funding for the study. \nEthics committee - additional information: No ethic committe \nAuthor Disclosures:  \nLuca Bombelli: Nothing to disclose \nSimone Vito Fasulo: Nothing to disclose \nDiego Catania: Nothing to disclose \nAndrea Roletto: Nothing to disclose \nElena Scaramelli: Nothing to disclose \nGiuseppe Roberto Bonfitto: Nothing to disclose \n \n \nPatient, Public and Practitioner Partnership within  Imaging and \nRadiotherapy: An exploration of the implementation and use of the \nCollege of Radiographers Guiding Principles \n*R. M. Strudwick*¹, A. Ramlaul², P. Shuttleworth³, C. Fiyebor¹; ¹Ipswich/UK, \n²High Wycombe/UK, ³Leeds/UK \n(r.strudwick@uos.ac.uk) \n \nPurpose or Learning Objective: In 2014 the National Health Service (NHS) \nreleased the Five Year Forward plan1, envisioning a  shift in power from health \nprofessionals to patients and the public. In respon se the Society and College of \nRadiographers (SCoR) produced the “Patient, Public and Practitioner \nPartnership within Imaging and Radiotherapy: Guidin g Principles” (P4) \ndocument which was implemented within four domains of radiography practice; \nservice delivery, service development, education an d research2. This project \nexplored how these guidelines were implemented; and  whether improvement \nto the quality and scope were needed, leading to ma king recommendations for \nupdating the document. \nMethods or Background: A qualitative methodological framework was \nadopted with two phases. Phase 1 – a survey explori ng use of the P4 \ndocument’s guiding principles. There was no maximum  number of participants \nto ensure inclusivity. Phase 2 - six focus groups f rom the four domains3. \nResults or Findings: 626 participants completed the phase 1 survey. 18.8 5% \n(n=118) of participants were aware of the document and used it as a reference \ntool for practice, teaching, and research. 81.15% ( n=508) of participants stated \nthey were unaware of the document. Themes from phas e 2; importance of \nservice user involvement in service delivery and ev aluation, resources to \nensure service user involvement, suggestions to upd ate the P4 document and \nuse of the P4 document in radiographer education. P articipants acknowledged \nguidance in the document was best practice. They re ported more awareness of \npatients’ needs and the effect this has on radiogra phers in supporting their \nneeds. \nConclusion: Participants recommended the document be given grea ter \nvisibility. The voices of patients and the public m ust be heard within \nradiography practice. \nLimitations: Small sample size \nFunding for this study: Feedback from this study can be used for the future  \ndevelopment of the P4 document. \nEthics committee - additional information: University of Suffolk Ethics \ncommittee approval \nAuthor Disclosures:  \nPamela Shuttleworth: Nothing to disclose \nAarthi Ramlaul: Nothing to disclose \nRuth Mary Strudwick: Grant Recipient: CoRIPS funded  project \nChioma Fiyebor: Nothing to disclose \n \n \nPatients’ perception of Radiographers’ communicatio n skills during \nplanar X-ray imaging: a single centre study \n*F. Zarb*, P. Bezzina, D. Ciantar; Msida/MT \n(francis.zarb@um.edu.mt) \n \nPurpose or Learning Objective: To explore patients’ perception of \nradiographers’ communication skills during planar x -ray imaging examinations \nat a general hospital in Malta \nMethods or Background: A random sample of patients attending for planar x-\nray examinations at a medical imaging department at  a general hospital in \nMalta filled out a self-designed questionnaire cons isting of demographic data \nand scored a series of statements on a Likert scale  of 1-4, with 1 being ‘Very \nunsatisfied’, 2 being ‘Unsatisfied’, 3 being ‘Satis fied’ and 4 being ‘Very \nsatisfied’. Data was collected and analysed to iden tify trends and understand \npatients’ perception of radiographers’ communicatio n skills. The Friedman test \nand Kruskal Wallis test were used to analyse the ga thered data. \nResults or Findings: Mean scores were either 3-Satisfied or 4-Very Satis fied. \nThere were no significant (p>0.05) discrepancies wh en comparing the different \ndemographics. Females provided significantly (p<0.0 5) higher mean rating \nscores than males. Patients’ scores varied the most  in the statement \n‘Radiographers informed me of what they had to do n ext’. \nConclusion: This is the first study of its kind performed local ly evaluating \npatients’ perceptions of Radiographers communicatio n skills. The mechanism \nin place locally for training radiographers in pati ent communication appears to \nbe effective, as patients are given instructions cl early and without undue \ndiscomfort. To make it simpler for patients once th ey have completed the x-ray \nand put their minds at ease during their entire sta y at the hospital, a clearer \npathway should be implemented. Having radiographers  with good \ncommunication skills improves the patients’ experie nce when attending for \nimaging services. \nLimitations: Limited sample size, honesty of participants fillin g the \nquestionnaire and a number of radiographers being f oreigners making \ncommunication a challenge. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: Ethical permission for this study \nwas sought and obtained from the Faculty of Health Science Research Ethics \nCommittee (FREC), University of Malta. \nAuthor Disclosures:  \nFrancis Zarb: Nothing to disclose  \nPaul Bezzina: Nothing to disclose \nDean Ciantar: Nothing to disclose \n \n \nA post-graduation expectation analysis of Italian R adiographers. The \nOPEN project \n*A. Masperi*; Abbiategrasso/IT \n(andrea.masperi1992@gmail.com) \n \nPurpose or Learning Objective: The aim of this study was to explore student \nsatisfaction with the OPEN project via a survey and  to identify factors \ninfluencing radiographers' post-graduation decision s. \nMethods or Background: Radiography undergraduates face changing career \npaths, influenced by technology, aspirations, finan cial gain and post-graduate \neducation choices. In January 2024 the Radiographer  bachelor’s degree \ncourse at the University of Milan launched the OPEN  project aimed at creating \na new postgraduate orientation programme. Following  AMEE guidelines, a \nsemi-structured survey was sent to 17 undergraduate  students in the OPEN \nproject to identify factors influencing their post- graduation decisions. University \nof Milan protocols ensured consent, anonymity, and confidentiality. The survey, \n\n \n \nSunday \nAbstract-based Programme \n \n 271  \nin three parts with Likert scale questions, showed good internal consistency via \nCronbach's alpha. \nResults or Findings: Out of 17 participants, 16 consented to participate  \n(94%), with all attending more than 50% of meetings . The internal consistency \nof the survey was excellent (αC = 0.83551) and results were expressed in \nterms of median and interquartile range. Students p rioritise careers in \nresearch-oriented hospitals with opportunities for research and academic \nadvancement. They seek facilities that offer perman ent contracts, incentivise \nclinical activities for salary growth and a balance d working environment. \nConclusion: The Radiography degree programme at the University of Milan \nhas demonstrated the effectiveness of postgraduate orientation programmes \nthat bridge academic and career gaps. \nLimitations: It should be noted that the results of this new pro ject are based \non a limited cohort of subjects from a single unive rsity. Furthermore, gender \npreferences were not explored, which could have pro vided valuable insights to \nthe survey by broadening the field of interview. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nAndrea Masperi: Nothing to disclose \n \n \nRadiation Awareness and Occupational Concerns Among  Radiographers \nand Students \nK. Brage¹, J. Jensen¹, O. Brage¹, *M. W. Kusk*², P.  L. Hansen¹,  \nM. Roland Pedersen³, H. Precht⁴; ¹Odense/DK, ²Esbjerg/DK, ³Vejle/DK, \n⁴Kolding/DK \n(martin.weber.kusk@rsyd.dk) \n \nPurpose or Learning Objective: To assess the knowledge, perceptions, and \nconcerns of professionals working with ionizing rad iation (IR) regarding their \noccupational exposure and its implications on their  health, fertility, and \noffspring. \nMethods or Background: This cross-sectional survey included European \nradiographers and students and was distributed via social media and \nprofessional bodies (EFRS and Danish Society of Rad iographers) from March \nto July 2024. The questionnaire covered demographic s, knowledge of IR, and \nconcerns, using a five-point Likert scale. \nResults or Findings: A total of 629 participants from 32 countries were \nincluded: 414 women (mean age 34.34 ± 12.18), 208 m en (mean age 38.27 ± \n11.81), and 7 non-binary or undisclosed individuals . Of these 28.30% were \nstudents. Mean exposure time was 10.69 years ± 10.6 8. Overall, 29.77% of \nrespondents agreed that their radiation exposure co uld negatively impact their \nhealth, 23.75% expressed concerns about fertility, and 18.59% about their \nchildren's health. No significant differences were observed between sex in \nthese responses. Of the radiographers 0.44% of did not fully understand the \noccupational risks of IR, 5.77% were dissatisfied w ith their radiation protection \neducation, and 4.66% felt unqualified to inform pat ients. Additionally, 1.99% \nwere unsure about self-protection, and 4.00% felt t hey lacked the necessary \nprotection means. Regarding the International Basic  Safety Standard, a total of \n28.95% felt not up to date while the number was 12. 42% for the National \nLegislation. \nConclusion: This study highlights concerns regarding IR and rel ated health \nwith up to 30% expressing concerns. While only a mi nority of respondents felt \nthey needed more education or resources to protect themselves, a third lacked \nknowledge on the International Basic Safety Standar ds. \nLimitations: Selection bias could be present as this topic may h ave attracted \nrespondents with greater concerns. \nFunding for this study: None \nEthics committee - additional information: The University of Southern \nDenmark Research Ethics Committee (23/70920) approv ed this project on 8 \nDecember 2023. Before accessing the questionnaire, participants were \ninformed about the study's purpose and were assured  of the confidentiality of \ntheir responses. Only those who provided their info rmed consent proceeded to \ncomplete the questionnaire. \nAuthor Disclosures:  \nHelle Precht: Nothing to disclose \nKaren Brage: Nothing to disclose \nMartin Weber Kusk: Nothing to disclose \nPernille Lund Hansen: Nothing to disclose \nJanni Jensen: Nothing to disclose \nMalene Roland Pedersen: Nothing to disclose \nOliver Brage: Nothing to disclose \n \n \n \n \n \n \n \nAbdominal Ultrasound Simulation based on CT examina tions as an \neducational tool for enhancing Ultrasound acquisiti on competences of \nradiography students \nR. S. T. Ribeiro, C. Schiesser, *C. Campeanu*, C. S . D. Reis; Lausanne/CH \n(cosmin.campeanu@hesav.ch) \n \nPurpose or Learning Objective: To assess the effectiveness of simulate d CT \nexaminations as an educational intervention for enh ancing radiography \nstudents' competencies in ultrasound(US). \nMethods or Background: A pilot study was conducted with third-year \nstudents enrolled in a four-week US module. Partici pants had prior knowledge \nof physics but no US practical experience. The educ ational intervention utilised \nabdominal CT examinations to enhance competences in  US image acquisition, \nanalysis and transducer positioning. A 20 images se t comprising abdominal \nanatomy was administered before and after the modul e. Assessments were \nmade using a Likert-scale across 4 categories: anat omical identification, \nanatomical topographical correlation, sectional pla nes recognition, transducer \npositioning. Data were analysed using measures of c entral tendency and \ndispersion to assess improvements. \nResults or Findings: The simulated US based on CT examinations improved \nstudents' US competencies. Anatomical Identificatio n scores increased by \n49.8%, mainly in the gallbladder (2.30-point increa se), liver (1.85-points), and \npancreas (1.55-points). Anatomical Topographical Co rrelation improved by \n34.2%, particularly for the pancreas (1.80-point) a nd gallbladder (1.50-points), \nreflecting enhanced spatial understanding. Sectiona l Planes Recognition was \nenhanced by 46.5%, for gallbladder (2.40-point) and  kidneys (1.86 ), \ndemonstrating better comprehension of cross-section al anatomy. Transducer \nPositioning presented a a gain of 66.1%, with impro vements for gallbladder \n(2.60) and kidneys (2.27). Overall, the interventio n effectively addressed initial \nskill gaps, leading to advancements in both specifi c anatomical structures \nanalysis and hands-on practical competencies in US.  \nConclusion: The use of CT examinations proved to be a valuable tool to \nimprove not only anatomical identification but also  students’ spatial \nunderstanding and practical application skills in U S. The greatest \nimprovements were observed in areas that were initi ally the most challenging, \nsuch as pancreas and gallbladder identification and  transducer positioning. \nLimitations: Only abdominal anatomy was tested. Reduced student panel (5 \nstudents) \nFunding for this study: Not applicable \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nCosmin Campeanu: Nothing to disclose \nClaudia Sa Dos Reis: Nothing to disclose  \nRicardo Silva Teresa Ribeiro: Nothing to disclose \nClaire Schiesser: Nothing to disclose \n \n \nEvaluating internship guides: Can student opinion c reate a virtuous \ncircle? \n*A. Devetti*¹, S. Da Dalt¹, F. R. Fabris², L. Ceres er¹, M. G. Belgrano²,  \nR. Girometti¹, C. Zuiani¹; ¹Udine/IT, ²Trieste/IT \n(angie.devetti@uniud.it) \n \nPurpose or Learning Objective: This study aimed to evaluate the efficacy of \na peer-assessment tool designed to enhance the qual ity of clinical tutoring \nexperiences for undergraduate students of a Radiogr apher Bachelor Degree \ninternship program. By collecting student feedback on their tutors' \nperformance, the institution sought to identify str engths, weaknesses, and \nimplement targeted improvement strategies. \nMethods or Background: Over three academic years, 940 student \nevaluations were collected from 180 clinical tutors  across four departments \n(Radiotherapy, Nuclear Medicine, Medical Physics, a nd Diagnostics). Tutors \nwere evaluated on seven dimensions using a 10-point  Likert scale. The \nreliability of the evaluation tool was assessed usi ng Cronbach's alpha, which \nconsistently yielded values between 0.92 and 0.94. \nResults or Findings: The results indicate a high overall level of studen t \nsatisfaction with the clinical tutoring experience.  The mean overall rating was \n8.87 out of 10, with a standard deviation of 0.97. While slight variations were \nobserved across departments, these differences were  not statistically \nsignificant. The high Cronbach's alpha values sugge st that the evaluation tool \nis reliable and internally consistent. \nConclusion: The implementation of a peer-assessment tool has pr oven to be \nan effective method for gathering valuable feedback  on the quality of clinical \ntutoring. By providing tutors with individualized f eedback, the institution has \nfostered a culture of continuous improvement. The a ggregated data has also \nallowed for department-wide analysis, enabling the identification of areas \nwhere additional training or support may be needed.  \nLimitations: The evaluation tool was designed specifically for t he \nRadiographer Bachelor Degree internship program and  may not be \ngeneralizable to other healthcare settings. Additio nally, the study relied solely \non student perceptions of tutor performance. \nFunding for this study: None \n\n \n \nSunday \nAbstract-based Programme \n \n 272  \nEthics committee - additional information: None \nAuthor Disclosures:  \nStefano Da Dalt: Nothing to disclose \nAngie Devetti: Nothing to disclose \nChiara Zuiani: Nothing to disclose \nManuel Gianvalerio Belgrano: Nothing to disclose \nRossano Girometti: Nothing to disclose \nLorenzo Cereser: Nothing to disclose \nFrancesca Romana Fabris: Nothing to disclose \n \n \n09:30-11:00 Research Stage 1 \nResearch Presentation Session: \nMusculoskeletal \nRPS 2310 \nImaging of the various pathologies of the \nspine \n \nModerator \nC. Loupatatzis; Männedorf/CH  \n(c.loupatatzis@spitalmaennedorf.ch) \n \n \n3D Ultrashort Echo Time MRI for Assessing the Carti laginous Endplate of \nthe lumbar intervertebral discs: Correlation with D isc Degeneration and \nModic Changes in Conventional Fast Spin Echo Sequen ces \n*Y. Kim*¹, J. G. Cha², S. Lee¹; ¹Seoul/KR, ²Bucheon /KR \n(kimyeoju@hanyang.ac.kr) \n \nPurpose or Learning Objective: To investigate the association between \ncartilaginous endplate (CEP) abnormalities on 3-dim ensional ultrashort echo \ntime MRI with cone trajectory technique (3D UTE) an d disc degeneration and \nendplate Modic change on conventional MRI. \nMethods or Background: Ninety one patients (44 men, 47 women, mean age: \n55.75 years, range: 19-85 years) underwent MRI of t he lumbar spine with \nconventional sagittal T1, T2-weighted and fat-suppr essed T2 weighted fast \nspin echo sequence and sagittal 3D UTE cone traject ory technique (TR = 16.1 \nms, TE = 0.032 ms and 6.6 ms) in 3T MRI. Two muscul oskeletal radiologists \nassessed CEP abnormalities (irregularity, thickenin g, thinning and defects) of \nthe superior and inferior endplates of the L3-4, L4 -5 and L5-S1 discs on 3D \nUTE and disc degeneration with the Pfirrmann gradin g system, and presence \nof Modic change of the endplate on conventional MRI  by consensus. The \nrelationship of CEP abnormalities with the disc deg eneration and Modic \nchange was tested using Pearson's chi-square test a nd Spearman's correlation \nanalysis. \nResults or Findings: All CEP abnormalities were positively correlated wi th \nPfirrmann grading system (Spearman ρ, 0.31-0.47) and Modic change \n(Spearman ρ, 0.24-0.50) with statistical significance (p < 0.0 01 for all \nPearson's chi-square and Spearman's correlation ana lysis). \nConclusion: The CEP abnormality in 3D UTE MRI may be associated  with the \nseverity of disc degeneration and the presence of M odic change. \nLimitations: The number of patients in the study is relatively s mall. MRI \nfindings did not correlate with pathological findin gs. The MRI findings did not \ncorrelate with the patient's clinical symptoms or p rognosis. This is a cross-\nsectional study that does not allow for a longitudi nal study of the patient. \nFunding for this study: None \nEthics committee - additional information: This study was designed \nprospectively, and was approved by the Inha Univers ity hospital's review \nboard, and informed consent was obtained from all t he participating patients. \nAuthor Disclosures:  \nSeunghun Lee: Nothing to disclose  \nYeoju Kim: Nothing to disclose  \nJang Gyu Cha: Nothing to disclose \n \n \nAdded Value of Color-Coded Fat-Calcium Dual-Energy CT in the \nDetection of Spine Occult Bone Metastasis – a Pilot  Study \n*J. Li*¹, J. Liu²; ¹Fujian/CN, ²Xiamen/CN \n(1508883851@qq.com) \n \nPurpose or Learning Objective: To assess the capability of color-coded Fat-\nCalcium dual-energy CT (DECT) in identifying spinal  occult bone metastases \n(S-OBMs). \nMethods or Background: DECT images of a consecutive series of lung \ncancer patients were retrospectively analyzed. Two radiologists reviewed \nconventional CT images and color-coded Fat-Calcium images, recording the \nlocations (diffuse infiltration, focal involvement of vertebral cancellous, \nvertebral edges, basivertebral venous plexus, and a ppendages) and number of \noccult bone metastases (OBMs) identified on the col or-coded Fat-Calcium \nimages. Diagnostic performance measures (sensitivit y, specificity, positive \npredictive value (PPV), negative predictive value ( NPV), and accuracy) were \nthen assessed. \nResults or Findings: A total of 24 patients were included, comprising 80  \nspinal occult bone metastases (S-OBMs). Color-coded  DECT images show \n100% sensitivity, PPV, and accuracy in diagnosing d iffusely invasive S-OBMs. \nThe overall sensitivity of color-coded DECT images for focal occult metastases \nwas measured at 96.1%. However, the PPV and accurac y of DECT for focal \nOBMs were influenced by the lesion’s location. The PPV and accuracy of \nOBMs in vertebral trabecular regions were higher th an those in vertebral \nedges, basivertebral venous plexus, and attachments  (PPV: 81%, 7.7%, 2.0%, \nand 6.0%, respectively; accuracy: 95.2%, 67.2%, 47. 4%, and 14%, \nrespectively). The diagnostic performance for verte bral cancellous regions was \nthe highest, with sensitivity, specificity, PPV, NP V, and accuracy of 94%, \n95.5%, 81%, 98.7%, and 95.2%, respectively. \nConclusion: Color-coded Fat-Calcium DECT significantly improves  the \ndetection of OBM in the spine. \nLimitations: Firstly, it was a retrospective study with a small sample size. \nSecondly, this study was based on non-enhanced DECT , and whether \nenhanced DECT could improve diagnostic efficiency r equires further study. \nFunding for this study: Natural Science Foundation of Fujian Province, Chin a \n(grant numbers: 2023J01181) \nEthics committee - additional information: Fujian Cancer Hospital Ethics \nCommittee (K2023-198-01) \nAuthor Disclosures:  \nJianfang Liu: Nothing to disclose \nJie Li: Nothing to disclose \n \n \nEx vivo and in vivo validation of dual-layer detect or spectral-CT fat \nquantification of vertebrae bone marrow \n*Y. F. Melzer*, G. Campbell, N. F. Schubert, I. Fie dler, B. Busse, I. Molwitz; \nHamburg/DE \n(f.schwietzer@uke.de) \n \nPurpose or Learning Objective: To evaluate and validate dual-layer detector \nspectral-CT fat quantification (dlCT) of the verteb rae bone marrow. \nMethods or Background: Isolated human cadaver vertebrae (n=14) of 10 \nbody donors were scanned within 72-96 hours after d eath using dlCT \n(CT7500) at 120 kV and a 3T MRI (Ingenia) (Philips Healthcare, the \nNetherlands). Spherical volumes of interest (VOIs, 11 mm diameter) were \nplaced in the center of all vertebral bodies. Addit ionally, n=13 patients were \nprospectively included (mean age 57±9 years; three females) who underwent \nprior to liver transplantation multiphase dlCT and MR imaging. VOIs were \ndefined at the third lumbar vertebrae. Within the V OIs fat was quantified in \ndlCT scans without contrast agent using three-mater ial decomposition for \nhydroxyapatite, red bone marrow, and fat. Reference  values for red bone \nmarrow were generated from the blood pool. MRI fat quantification was \nperformed using mDIXONquant sequences (TE shortest,  TR shortest, FA 3°). \nFor statistics, Pearson’s correlations and Bland Al tman analysis were \nemployed. \nResults or Findings: Ex vivo correlation between dlCT and MRI was high \n(r2=0.94, p<0.001) with a mean difference of -0.55 [95% intervals of agreement \n-11.0, 9.9]. In vivo, correlation between dlCT and MRI was moderate (r2=0.47, \np=0.01). The mean difference amounted to 14.4 [95% intervals of agreement -\n3.9, 32.7]. \nConclusion: Ex vivo dlCT fat quantification of the vertebral bo ne marrow \ndelivers reproducable results. In vivo measurements  require further calibration \nof dlCT using MRI and - due to challenges of MRI fa t quantification in the \npresences of bone - preferably histology as a refer ence. \nLimitations: Small sample size due to ongoing recruitment and ne cessary \nfurther calibration of dlCT fat quantification with  histological analyses, for which \nthe vertebra are currently prepared by formalin fix ation. \nFunding for this study: None. \nEthics committee - additional information: 2023-300414-WF (Ärztekammer \nHamburg) \nAuthor Disclosures:  \nBjörn Busse: Nothing to disclose \nIsabel Molwitz: Nothing to disclose \nGraeme Campbell: Nothing to disclose \nImke Fiedler: Nothing to disclose \nNiklas Ferdinand Schubert: Nothing to disclose \nYasmin Fede Melzer: Nothing to disclose \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 273  \nVERIFACT: Revealing the Hidden Epidemic of Undiagno sed Vertebral \nFractures in Routine CT Scans \nH. P. Dimai, J. Igrec, *J. Steiner*, R. Riedl, M. F uchsjäger; Graz/AT \n(jakob.steiner@medunigraz.at) \n \nPurpose or Learning Objective: Osteoporosis is a condition characterized by \nlow bone mass and increased fracture risk. Vertebra l fractures are the most \ncommon, often undetected despite serious health con sequences. Studies \nreveal that many fractures are missed in radiograph ic and CT reports, with \nfalse-negative rates ranging from 30% to 84%, highl ighting significant \nunderreporting in clinical practice. The study aims  to assess recognition rates \nin chest and abdominal CTs. \nMethods or Background: This retrospective analysis included 1,500 CT \nimages from 1,380 patients. Two independent board-c ertified radiologist with \nmulti-year-experience reviewed each scan for verteb ral fractures, noting the \nnumber, location, and type of fracture, and classif ying them based on the \nGenant classification system (Grades 2 and 3). Rate r agreement was \nassessed using the Kappa coefficient and AC1 statis tic. Fracture \ndocumentation rates were calculated based on whethe r fractures were noted in \nthe formal radiology report summary or only in the free-text narrative. Stratified \nanalysis was performed by patient gender, fracture location, and the presence \nof multiple fractures. \nResults or Findings: Vertebral fractures were found in 11.5% of patients . \nAgreement between the raters was excellent, with a Kappa value of 0.94 (95% \nCI: 0.92–0.97) and AC1 of 0.99 (95% CI: 0.98–0.99).  However, fractures were \ndocumented in the summary in only 14.7% of cases (9 5% CI: 9.8%–20.9%), \nwhile 35.3% were mentioned only in the narrative po rtion (95% CI: 28.1%–\n43.0%). Overall, 50% of the fractures were reported  in either the summary or \nthe narrative (95% CI: 42.3%–57.8%). \nConclusion: While interrater agreement on fracture detection wa s high, the \nrate of documentation, particularly in the structur ed summary, was low. These \nfindings suggest a need for better reporting protoc ols to ensure vertebral \nfractures are clearly communicated, improving patie nt outcomes and clinical \ndecision-making. \nLimitations: n/a \nFunding for this study: n/a \nEthics committee - additional information: Approved by local ethics \ncommittee \nAuthor Disclosures:  \nJakob Steiner: Nothing to disclose \nRegina Riedl: Nothing to disclose \nMichael Fuchsjäger: Nothing to disclose \nHans Peter Dimai: Nothing to disclose \nJasminka Igrec: Nothing to disclose \n \n \nSpectral Collagen Imaging: Assessment of Thoracic D isk Herniation and \nDegeneration \n*M. Dimitrova*, C. Booz, S. Mahmoudi, A. Gökduman, L. D. Grünewald,  \nS. Bernatz, E. Höhne, T. Vogl, I. Yel; Frankfurt/DE  \n \nPurpose or Learning Objective: This study investigates the diagnostic \nefficacy of Dual-Energy CT (DECT)-derived collagen maps in evaluating \nthoracic disk herniation and degeneration. \nMethods or Background: A retrospective analysis was conducted involving \n51 patients who underwent dual-source DECT (Somatom  Force; Siemens \nHealthineers) and MRI of the thoracic spine within a two-week timeframe. Two \nblinded radiologists assessed the presence and type  of herniation using the \nNorth American Spine Society's classification for i ntervertebral disk pathology, \nevaluating both conventional grayscale CT and DECT collagen maps. \nDiagnostic accuracy, sensitivity, and specificity w ere calculated with MRI as the \nreference standard. Additionally, subjective assess ments of diagnostic \nconfidence and image quality were performed, and in ter-reader reliability was \nevaluated using the intraclass correlation coeffici ent. \nResults or Findings: Analysis of 612 intervertebral disks showed that DE CT \ncollagen maps achieved significantly higher sensiti vity (98.0%), specificity \n(97.7%), and diagnostic accuracy (97.7%) compared t o conventional CT \n(sensitivity: 72.0%, specificity: 97.0%, diagnostic  accuracy: 96.0%; p < 0.001). \nSubstantial inter-reader reliability was noted (κ=0.76, p < 0.001), with DECT \ncollagen maps providing enhanced diagnostic confide nce and image quality (p \n< 0.001). For assessing disk degeneration, DECT col lagen maps demonstrated \nhigh sensitivity (83.0%), specificity (81.9%), and diagnostic accuracy (83.4%) in \ndistinguishing non/mild from moderate/severe degene ration, with inter-reader \nreliability also showing strong agreement (κ=0.82, p < 0.001). Subjective \nevaluations reported moderate to high diagnostic co nfidence (median 3.5) and \nmoderate to good image quality (median 3.5). \nConclusion: DECT-derived collagen maps significantly enhance th e detection \nof thoracic disk herniation and degeneration, offer ing improved diagnostic \naccuracy, reliability, confidence, and image qualit y over conventional CT. This \nimaging technique serves as a valuable alternative for patients who cannot \nundergo MRI. \nLimitations: retrospective study single-centre Study small patie nt group CT \nsystem from a specific vendor \nFunding for this study: No funding. \nEthics committee - additional information: Approval by local ethics \ncommittee. \nAuthor Disclosures:  \nChristian Booz: Nothing to disclose \nIbrahim Yel: Nothing to disclose \nMirela Dimitrova: Nothing to disclose \nThomas Vogl: Nothing to disclose \nScherwin Mahmoudi: Nothing to disclose \nAynur Gökduman: Nothing to disclose \nLeon David Grünewald: Nothing to disclose \nSimon Bernatz: Nothing to disclose \nElena Höhne: Nothing to disclose \n \n \nCervical spine motion in dynamic X-rays – approach,  results, conclusions \n*M. J. Łubiński*, A. Majos, P. Kowalski; Lodz/PL \n(marcin.lubinski92@gmail.com) \n \nPurpose or Learning Objective: Dynamic X-ray is very usefull tool to \nevaluate cevical spine motion. The aim of this stud y is to find reference ranges, \ncorrelations and practical tips which radiologists can use in cervical spine \nmobility assesment. \nMethods or Background: 288 patients aged 19-78 years old without cervical \nspine osteoarthritis or suffering from first degree  osteoarthritis in Kellgren-\nLawrance classtification were examined. We performe d lateral X-rays in three \nprojections – neutral, flexion and extension. In th e functional X-ray \nexaminations of the cervical spine, the following p arameters were assessed: \nthe Cobb angle C2-C7, angular and horizontal segmen tal mobility and \nsegmental cervical curvature in flexion and extensi on. \nResults or Findings: Reference ranges for extension and flexion C2-C7 Co bb \nangle, angular and horizontal segmental mobility we re found. Analysis showed \nthe biggest mobility of C4-C5 segment. Correlations  suggest that horizontal \ndisplacement index is the most universal parameter in cervical spine mobility \nassessment. \nConclusion: A large group of patients and multitude measurement s made it \npossible to find reliable and clinically useful ref erence ranges and parameters \nwhich can be used in routine evaluation of cervical  spine functional tests. \nLimitations: Not applicable. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nAgata Majos: Author: Co-researcher \nMarcin Janusz Łubiński: Author: Main researcher \nPiotr Kowalski: Author: Co-researcher \n \n \nLongitudinal assessment of structural abnormalities  in the lumbar spine \nof adolescent competitive alpine skiers over 48 mon ths \n*G. C. Feuerriegel*, D. Meyer, D. Fitze, J. Haniman n, C. Stern, S. Fröhlich,  \nJ. Scherr, J. Spörri, R. Sutter; Zurich/CH \n \nPurpose or Learning Objective: To longitudinally assess and compare spinal \nstructural abnormalities in adolescent competitive alpine skiers over 48 months \nand to compare MRI findings in asymptomatic and sym ptomatic skiers and \nexplore their clinical relevance. \nMethods or Background: Adolescent competitive alpine skiers recruited for a \ncross-sectional MRI investigation underwent a 3T MR  imaging of the lumbar \nspine at baseline and after 48 months. All MR image s were assessed for \nstructural changes occurring in the intervertebral disc, vertebral body and facet \njoints. At both baseline and follow-up, athletes' l ow back pain (LBP) symptoms \nwere assessed and Athletes were classified as sympt omatic if at least one \n'substantial' episode of health problems related to  back overuse had occurred \nin the 12 months prior to the MRI examination. The Wilcoxon signed-rank test \nand Pearson's chi-squared test were used to compare  the measurements. \nResults or Findings: A total of 63 athletes (mean age at follow-up 19.6± 1.2 \nyears, 25 female) were included in the study. A sig nificant increase in LBP \naffecting training and competition was observed at follow-up (P = 0.04). Of the \nathletes with LBP, 63% (n=16) reported recurrent LB P, 14% (n=4) reported \npermanent backpain, and 26% (n=7) reported one-time  LBP since baseline. \nAssessment of structural changes revealed a signifi cant increase in the \nnumber of athletes with disc dehydration (P < 0.001 ), disc protrusions (P = \n0.002) or disc extrusions (P = 0.04). Overall, stru ctural abnormalities did not \ncorrelate with LBP (P>0.05). \nConclusion: Overuse related structural changes progress during adolescence \nand are not self-limiting. However, structural chan ges are not directly \ncorrelated with LBP. This finding may facilitate th e development of appropriate \ntreatment and prevention strategies that do not foc us solely on spinal changes. \nLimitations: Structural abnormalities were only assessed by MRI and not \nconfirmed by other modalities. \n\n \n \nSunday \nAbstract-based Programme \n \n 274  \nFunding for this study: This study was generously supported by the Balgrist  \nFoundation. \nEthics committee - additional information: Cantonal Ethics Committee \nZurich \nAuthor Disclosures:  \nJohannes Scherr: Nothing to disclose \nJörg Spörri: Nothing to disclose \nStefan Fröhlich: Nothing to disclose \nChristoph Stern: Nothing to disclose \nJonas Hanimann: Nothing to disclose \nReto Sutter: Nothing to disclose \nGeorg Constantin Feuerriegel: Nothing to disclose \nDaniela Meyer: Nothing to disclose \nDaniel Fitze: Nothing to disclose \n \n \nMultiparametric quantitative MRI in Charcot-Marie-T ooth 1A inherited \nneuropathy: correlation with motor function and bal ance performance \n*D. Bianco*, F. Zaottini, S. Rinaldi, M. Pansecchi,  M. Hamedani, S. Massucco, \nE. Rovetta, C. Martinoli; Genova/IT \n(deborahbianco@yahoo.it) \n \nPurpose or Learning Objective: We aimed to evaluate the feasibility of a \nquantitative multiparametric MRI protocol of lumbo- sacral plexus and proximal \nsciatic nerve to differentiate patients affected by  Charcot-Marie-Tooth type 1A \n(CMT1A) neuropathy from controls and to correlate t hese imaging parameters \nwith clinical grading scale of disease's severity. \nMethods or Background: Patients with clinical, electrophysiological and \ngenetical proven CMT1a were prospectively enrolled.  The 3T MRI protocol \nincluded the following sequences: Diffusion Tensor Imaging, 2 points T2 \nDIXON, T1 mapping and T2 mapping. The MR parameters  were independently \nmeasured by two radiologists. The same day of MRI e xamination, the CMT1a \npatients were clinically assessed using CMTNS score  and Berg Balance Score \n(BBS). An age and sex matched control group without  clinical signs of \nneuropathy (NN) was enrolled. \nResults or Findings: n=11 patients (7 f,4 m), 47.57 yo ± 14.39 and n= 8 NN \ncontrols (5 f, 3 m) 46.5 yo ± 14.39 underwent MRI. The interobserver reliability \nof measurements was good (ICC=0,65). Lumbosacral pl exus roots and sciatic \nnerve cross sectional area, Fractional Anisotropy ( FA), T1 and T2 relaxation \ntime were significantly different between the two g roups (p<0,05). Bilateral L5 \nand S1 T1 relaxation values and FA significantly co rrelated with CMTNS \n(respectively R=0.86, p=0.013 and R=0,77, p=0,04) a nd BBS (respectively \nR=0.62, p=0.041 and R= 0.75, p=0,042). Sciatic Nerv e FA demonstrated \nstrong correlations with both CMTNS (R=0.92, p=0.01 0) and BBS (R=0.89, \np=0.019). \nConclusion: These findings suggest that FA and T1 relaxation ti me of the \nlumbosacral plexus and sciatic nerve are the MRI pa rameters that better \ncorrelate with balance performance and overall func tional disability in CMT1A \npatients, representing potential biomarker for dise ase severity and longitudinal \nevaluation. \nLimitations: Small sample size. \nFunding for this study: The study was funded by the Italian Ministry of hea lth \ntrough the public grant BANDO RICERCA FINALIZZATA 2 021. \nEthics committee - additional information: Comitato Etico Territoriale - \nRegione Liguria \nAuthor Disclosures:  \nDeborah Bianco: Nothing to disclose \nSimone Rinaldi: Nothing to disclose \nFederico Zaottini: Nothing to disclose \nCarlo Martinoli: Nothing to disclose \nMehrnaz Hamedani: Nothing to disclose \nEdoardo Rovetta: Nothing to disclose \nMichelle Pansecchi: Nothing to disclose \nSara Massucco: Nothing to disclose \n \n \nT2 relaxation times of the pubic symphysis in ostei tis pubis \n*N. Andjelic*¹, N. Holl², B. Klaan², M-A. Weber²; ¹ Sremska Kamenica/RS, \n²Rostock/DE \n(nikola.andjelic92@gmail.com) \n \nPurpose or Learning Objective: This study aimed to evaluate the T2 \nrelaxation times of the cartilage layer of the pubi c symphysis in male athletes \nwith osteitis pubis and examine the correlation bet ween these values, pubic \nbone marrow edema (BME), the pubic symphysis width,  and the presence of \ncleft injuries. \n \n \n \n \n \nMethods or Background: Sixty-two male athletes (median age, 28 years) \npresenting with groin pain were examined using a 3- Tesla MRI system. T2 \nmapping was applied using a T2W sequence to assess the pubic symphysis in \nthree ways: interpubic disc, articular cartilage, a nd the entire hyaline-\nfibrocartilage complex. T2 relaxation times were me asured, and BME and cleft \ninjuries were identified. Correlations between T2 v alues, symphyseal width, \nand the presence of BME and pubic cleft injuries we re explored. \nResults or Findings: The median T2 relaxation times for the hyaline-\nfibrocartilage complex, interpubic disc, and articu lar cartilage were 49.7 ms, \n54.6 ms, and 46.2 ms, respectively, with significan t differences between (p-\nvalue < 0.001), and a notable distinction between t he right and left sides of the \narticular cartilage. Athletes with BME had higher T 2 relaxation times for the \nhyaline-fibrocartilage complex and interpubic disc (p-value < 0.01) but not for \nthe articular cartilage. A moderate positive correl ation (r = 0.4) was found \nbetween symphysis width and T2 relaxation times. \nConclusion: T2 mapping provides valuable insights into the stru ctural \nchanges in the pubic symphysis in athletes with ost eitis pubis. Higher T2 \nrelaxation times in the interpubic disc and whole s ymphyseal \nhyaline/fibrocartilage complex are associated with BME, suggesting their \npotential use in evaluating osteitis pubis. \nLimitations: The study was limited to male patients, and the abs ence of an \nasymptomatic control group is a notable limitation.  Additionally, the pre-\nscreening of all participants introduces a potentia l selection bias due to the \nstudy design. \nFunding for this study: This research project was part of the ESOR Bracco \nResearch Fellowship 2024 \nEthics committee - additional information: The study was approved by the \nEthical Committee of Rostock University (approval N o. A 2020-0040) \nAuthor Disclosures:  \nNorman Holl: Nothing to disclose \nMarc-André Weber: Nothing to disclose \nBastian Klaan: Nothing to disclose \nNikola Andjelic: Nothing to disclose \n \n \nUltrasound-guided navigation system for spine surge ry \n*A. Lubansu*, P. Pandin; Brussels/BE \n(alphonse.lubansu@hubruxelles.be) \n \nPurpose or Learning Objective: Current image-guided navigation systems in \nspine surgery rely on ionizing radiation from intra operative fluoroscopy or CT \nscans. This study evaluates the feasibility and uti lity of fusing intraoperative \nultrasound (US) imaging with preoperative lumbar CT  scans to create a novel \nUS-guided spinal navigation system, aiming to reduc e radiation exposure. \nMethods or Background: Over one year, 25 patients undergoing lumbar \nspine surgery participated in this study. Cortical borders of spinal structures \nwere co-registered and fused with preoperative CT s cans. The accuracy and \ntime required for co-registration were assessed. Wh en navigation accuracy \nwas within 2mm, various spinal procedures, includin g screw removal or \nplacement and canal or foraminal recalibration, wer e performed using the \nsystem. \nResults or Findings: Accurate co-registration was achieved in under 10 \nminutes for all cases. Anatomical landmarks for co- registration varied \ndepending on the target region (sacrum, sacroiliac joint, lumbosacral junction, \nor lumbar vertebrae). The system facilitated percut aneous screw placement, \nintradural lesion localization, and optimal neural structure decompression. No \ncomplications related to US-guided navigation occur red. The total radiation \ndose was reduced compared to conventional non-navig ated procedures. \nConclusion: This preliminary experience suggests that US-guided  navigation \nfor spinal procedures is feasible, accurate, safe, and potentially beneficial in \nreducing radiation exposure. The system demonstrate d versatility across \nvarious spinal regions and procedures. Further rese arch with larger cohorts \nand more indications is needed to fully understand the potential advantages \nand limitations of this innovative navigation syste m. \nLimitations: This study is limited by its small sample size and single-center \ndesign. A direct comparison with conventional navig ation techniques was not \nassessed. The learning curve for implementing this new technique was not \nevaluated. \nFunding for this study: No \nEthics committee - additional information: Not applicable \nAuthor Disclosures:  \nAlphonse Lubansu: Nothing to disclose \nPierre Pandin: Nothing to disclose \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 275  \n09:30-11:00 Research Stage 2 \nResearch Presentation Session: Imaging \nInformatics and Artificial Intelligence \nRPS 2305 \nHealthy aging, body composition and \nprevention: the true potential of AI? \n \nModerator \nM. Pop; Tg.Mures/RO  \n \n \nAutomated coronary calcification assessment on unga ted unenhanced \nchest CT using an optimised nnUNet framework for pa tient \nprognostication in non-small-cell lung cancer \n*J. Y. Anifowose*, Z. Li, G. Agarwal, E. Aboagye, B . Ariff, S. Copley, M. Chen; \nLondon/UK \n(anifowoseolayinka68@gmail.com) \n \nPurpose or Learning Objective: To develop an automated software for \nassessing coronary calcification in non-small cell lung cancer (NSCLC) patients \nusing an optimised deep learning nnUNet framework f or disease \nprognostication. \nMethods or Background: Cardiovascular risk is higher in NSCLC patients \nthan in the general population, but is often underd iagnosed in clinical practice. \nAttenuation correction CTs from routinely acquired PET-CT staging scans are \nungated unenhanced studies which offer an opportuni ty to assess this risk \nwithout incurring additional radiation exposure or radiology workload. nnUNet \nis a state-of-the-art deep learning architecture de monstrating superior \nperformance in medical image segmentation applicati ons. We trained nnUNet \nmodels for coronary calcification on ungated unenha nced chest CTs (n = 100) \nfrom a public domain dataset (Stanford AIMI) and te sted them on independent \ndata: attenuation correction CTs of PET-CT scans of  NSCLC patients acquired \nat our multi-centre institution between 2012 and 20 18 (n = 287, age: 66.8 ± \n10.1, male: female 174:113). The reference truth se gmentations were drawn \nand verified by two radiologists of 8 and 2 years o f experience. Models with \nvarying batch sizes and convolutional filters were developed and \nbenchmarked; with the best performing one selected to develop a composite \nprognostic predictor, based on model-derived corona ry calcification score and \nsignificant NSCLC features. \nResults or Findings: The best performing nnUNet has a 3D_fullres \nconfiguration with batch size of 4 and patch size 2 8x224x224. All cases of \ncoronary calcifications were successfully detected.  Multivariable Cox showed \nstatistical significance of disease histology and s tage on patient survival. The \ncomposite predictor achieved statistically signific ant prognostic risk \nstratification (p-value < 0.05). \nConclusion: An optimised nnUNet framework can facilitate automa ted \ncoronary calcification assessment on ungated unenha nced CT to support a \ncomposite prognostic predictor in NSCLC patients. \nLimitations: Retrospective study. Single external validation coh ort. \nFunding for this study: Academy of Medical Sciences award SGL026 ∖1024. \nEthics committee - additional information: Retrospective observational \nstudy IRAS: 243592 REC: 18HH4616 \nAuthor Disclosures:  \nZechen Li: Nothing to disclose \nSusan Copley: Nothing to disclose \nBen Ariff: Nothing to disclose \nMitchell Chen: Nothing to disclose \nEric Aboagye: Nothing to disclose \nGirija Agarwal: Nothing to disclose \nJubril Yinka Anifowose: Nothing to disclose \n \n \nDeep learning-based biological age estimation from MRI predicts \ncardiometabolic events in the general population \n*M. Jung*¹, M. Reisert², H. Rieder², S. Rospleszcz² , M. T. Lu¹, F. Bamberg²,  \nV. Raghu¹, J. Weiß²; ¹Boston MA/US, ²Freiburg/DE \n(matthias.jung@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: Chronological age is one of the cornerstones \nof medical decision-making, but it's an imperfect m easure of health. We \npropose a deep learning framework (MRI-Age) for est imating biological age \nfrom MRI and investigated its value in predicting c ardiometabolic outcomes in \nthe general population beyond chronological age. \nMethods or Background: We used 30,389 individuals from the German \nNational Cohort (NAKO) to develop MRI-Age, which ta kes MRI-derived \nvolumetric body composition, including subcutaneous  (SAT), visceral (VAT), \nintramuscular adipose tissue (IMAT), and skeletal m uscle (SM) from the 1st to \n5th lumbar vertebra as input and outputs an age est imate in years. For \ndownstream analyses, we calculated MRI-Age accelera tion, defined as an age-\nspecific z-score of the age estimate. We then valid ated this framework in an \nexternal testing set of 36,317 individuals from the  UK Biobank (UKBB). Incident \noutcomes were diabetes, MACE, and all-cause mortali ty. Multivariable Cox \nregression assessed the association between MRI-Age  acceleration categories \n“negative” (MRI-Age acceleration <-1), “reference” (MRI-Age acceleration -1 to \n1), and “positive” (MRI-Age acceleration >1) and ou tcomes adjusted for \ntraditional cardiometabolic risk factors in the UKB B. \nResults or Findings: In 36,317 UKBB participants (65.1±7.8 years, 51.7% \nfemale; median follow-up 4.8 years), we found a hig her incidence of diabetes, \nMACE, and death in individuals with positive MRI-ac celeration. In multivariable-\nadjusted Cox regression, we observed a significant positive association \nbetween positive MRI-Age acceleration and diabetes (aHR: 1.87, 95% CI \n[1.56-2.25], p<0.001), MACE (aHR: 1.26, 95% CI [1.0 1-1.57], p=0.038), and \nall-cause mortality (aHR: 1.37, 95% CI [1.09-1.72],  p=0.007). \nConclusion: Deep learning-derived biological age from MRI predi cts \ncardiometabolic outcomes in the general population beyond chronological age \nand cardiometabolic risk factors. Individuals at hi gh MRI-Age could benefit \nfrom personalized prevention strategies, lifestyle interventions, and treatment \nplanning. \nLimitations: Limited age-range. Predominantly white population. \nFunding for this study: This project was conducted with data from the \nGerman National Cohort (NAKO) (www.nako.de). The NA KO is funded by the \nFederal Ministry of Education and Research (BMBF) [ project funding reference \nnumbers: 01ER1301A/B/C, 01ER1511D, and 01ER1801A/B/ C/D], federal \nstates of Germany, and the Helmholtz Association, t he participating \nuniversities and the institutes of the Leibniz Asso ciation. This research has \nbeen conducted using the UK Biobank Resource under Application Number \n80337. We thank all participants who took part in t he NAKO and UKBB study \nand the staff of these research initiatives. MJ was  funded by the Deutsche \nForschungsgemeinschaft (DFG, German Research Founda tion) - 518480401. \nVKR was funded by Norn Group Longevity Impetus Gran t, NHLBI \nK01HL168231, and AHA Career Development Award 93517 6. \nEthics committee - additional information: Informed consent was obtained \nfrom all participants in the UK Biobank and the Ger man National Cohort study. \nIn addition, we received local IRB approval (IRB of  the University of Freiburg: \n23-1316-S1-retro and 24-1099-S1-retro). \nAuthor Disclosures:  \nSusanne Rospleszcz: Nothing to disclose \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nMatthias Jung: Nothing to disclose \nFabian Bamberg: Nothing to disclose \nHanna Rieder: Nothing to disclose \nVineet Raghu: Nothing to disclose \nMichael T. Lu: Nothing to disclose \n \n \nBody Composition in the general population: MRI-der ived reference \ncurves from over 66,000 individuals and their assoc iation with \ncardiometabolic outcomes \n*M. Jung*¹, M. Reisert², H. Rieder², S. Rospleszcz² , M. Russe², M. T. Lu¹,  \nF. Bamberg², V. Raghu¹, J. Weiß²; ¹Boston MA/US, ²F reiburg/DE \n(matthias.jung@uniklinik-freiburg.de) \n \nPurpose or Learning Objective: Body composition (BC) plays an important \nrole in risk estimation in patients with cardiometa bolic disease and cancer, but \nreference curves are missing to place individual me asurements in context. We \ndeveloped a deep learning framework to quantify BC from MRI to calculate \nreference curves and investigated its value for pre dicting cardiometabolic \noutcomes. \nMethods or Background: BC extracted from MRI data of the UK Biobank \n(UKBB) and German National Cohort included 1) subcu taneous (SAT), 2) \nvisceral (VAT), 3) intramuscular adipose tissue (IM AT), 4) skeletal muscle \n(SM), and 5) SM fat fraction (SMFF). Reference curv es were generated using \ngeneralized additive models for each BC metric to c alculate age, sex, and \nheight-specific z-scores. Multivariable Cox regress ion assessed the association \nbetween z-score categories (low: z<-1; middle: z=-1 -1; high: z>1) and \noutcomes (incident diabetes; major adverse cardiova scular events [MACE]; \nand all-cause mortality) adjusted for traditional c ardiometabolic risk factors in \nthe UKBB. \nResults or Findings: Among 66,608 individuals (57.7±12.9 years; BMI: \n26.2±4.5 kg/m2, 48.3% female), we observed sex diff erences in BC volumes \nand distributions with SAT, VAT, SMFF, and IMAT pos itively and SM \nnegatively associated with age. We found graded ass ociations between BC z-\nscore categories and health outcomes in the UKBB. I n multivariable adjusted \nCox regression, z-score risk categories had hazard ratios of up to 2.69 for \nincident diabetes (high VAT), 1.41 for incident MAC E (high IMAT), and 1.49 for \nall-cause mortality (low SM) compared to middle cat egories. \n\n \n \nSunday \nAbstract-based Programme \n \n 276  \nConclusion: BC measures normalized for age, sex, and height are  associated \nwith cardiometabolic outcomes beyond traditional ri sk factors in the general \npopulation. We will provide open-source BC referenc e curves, which may \naccelerate the clinical translation of BC-based ris k assessment for \ncardiometabolic disease and support future BC resea rch. \nLimitations: Study population is predominantly white Western Eur opean \nadults. \nFunding for this study: This project was conducted with data from the \nGerman National Cohort (NAKO) (www.nako.de). The NA KO is funded by the \nFederal Ministry of Education and Research (BMBF) [ project funding reference \nnumbers: 01ER1301A/B/C, 01ER1511D, and 01ER1801A/B/ C/D], federal \nstates of Germany, and the Helmholtz Association, t he participating \nuniversities and the institutes of the Leibniz Asso ciation. This research has \nbeen conducted using the UK Biobank Resource under Application Number \n80337. We thank all participants who took part in t he NAKO and UKBB study \nand the staff of these research initiatives. MJ was  funded by the Deutsche \nForschungsgemeinschaft (DFG, German Research Founda tion) - 518480401. \nVKR was funded by Norn Group Longevity Impetus Gran t, NHLBI \nK01HL168231, and AHA Career Development Award 93517 6. \nEthics committee - additional information: Informed consent was obtained \nfrom all participants in the UK Biobank and the Ger man National Cohort study. \nIn addition, we received local IRB approval (IRB of  the University of Freiburg: \n23-1316-S1-retro and 24-1099-S1-retro). \nAuthor Disclosures:  \nSusanne Rospleszcz: Nothing to disclose \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nMatthias Jung: Nothing to disclose \nMaximilian Russe: Nothing to disclose \nFabian Bamberg: Nothing to disclose \nHanna Rieder: Nothing to disclose \nVineet Raghu: Nothing to disclose \nMichael T. Lu: Nothing to disclose \n \n \nAI-Driven MRI Biomarker Extraction and Machine Lear ning Analysis of \nTheir Association with Diabetes: A UK Biobank Study  \n*S. Kim*¹, D. W. Kim¹, C. Han¹, D. Kim², D. Yoon¹; ¹Seoul/KR, ²Daegu/KR \n(crown7699@yuhs.ac) \n \nPurpose or Learning Objective: To evaluate AI-derived imaging biomarkers \nfrom whole-body MRI in detecting and predicting dia betes mellitus (DM). \nMethods or Background: An open-source multi-label segmentation model \nwas applied to Dixon whole-body MRIs from the UK Bi obank to segment \norgans and body compositions. Volume indices (volum e/m³) and fat fractions of \neach structure were calculated automatically. For D M detection at the time of \nMRI, logistic regression was performed. Excluding b aseline DM, random \nsurvival forest analysis was performed for predicti ng future DM. Area under \ncurve (AUC) and Harrell’s C-index was used. Perform ance of imaging \nbiomarkers was compared to the Leicester Diabetes R isk Score. \nResults or Findings: Among the 2,924 participants, 149 had DM at baselin e. \nOf the 2,775 participants without baseline DM, 28 d eveloped DM and were \nincluded in the survival analysis (median follow-up  4.1 years, up to 8.9 years). \nFor DM detection, adrenal gland volume index, kidne y volume index, and \npancreatic fat fraction (AUC 0.748, 0.716, and 0.71 0 respectively) were the top \nclassifiers. The multivariate model, using 10 selec ted imaging features, \nachieved AUC of 0.802. In survival analysis, pancre atic fat fraction, adrenal \ngland volume index, and torso fat volume index (C-i ndex 0.713, 0.685, and \n0.678 respectively) were the top predictors. The mu ltivariate model with six \nselected imaging features achieved C-index 0.780, o utperforming the Leicester \nDiabetes Risk Score (C-index 0.651). When imaging f eatures were combined \nwith clinical features, performance further improve d (C-index 0.794). \nConclusion: AI-derived MRI biomarkers demonstrated strong perfo rmance in \ndetecting current DM and predicting future onset, h ighlighting their potential \nutility in opportunistic screening. \nLimitations: Further validation of the open-source segmentation model is \nnecessary to assess its quantitative and qualitativ e performance. \nFunding for this study: MD-PhD/Medical Scientist Training Program through \nthe Korea Health Industry Development Institute, fu nded by the Ministry of \nHealth & Welfare, Republic of Korea \nEthics committee - additional information: Our institution has received IRB \napproval for UK Biobank-related research, and any a dditional ethical \nconsiderations are adhered to in accordance with th is approval. \nAuthor Disclosures:  \nSongsoo Kim: Nothing to disclose \nDonghyun Kim: Nothing to disclose \nDong Won Kim: Nothing to disclose \nChangho Han: Nothing to disclose \nDukyong Yoon: Nothing to disclose \n \n \nCompositIA: an open-source pipeline for automated q uantification of \nbody composition scores from thoraco-abdominal CT s cans \n*R. F. Cabini*, A. Cozzi, S. Leu, B. Thelen, R. Kra use, F. Del Grande, S. Rizzo, \nD. U. Pizzagalli; Lugano/CH \n(raffaella.fiamma.cabini@usi.ch) \n \nPurpose or Learning Objective: This study aims to develop and validate \nCompositIA, an automated, open-source pipeline for quantifying body \ncomposition scores from thoraco-abdominal CT scans.  \nMethods or Background: CompositIA consists of three main steps: automatic \nidentification of the L1 and L3 vertebrae, segmenta tion of image slices at these \nspinal levels, and quantification of body compositi on indices. Two Deep \nLearning models were used: MultiResUNet for detecti ng CT slices intersecting \nthe L1 and L3 vertebrae, and UNet for segmenting th e corresponding axial \nslices. The pipeline was trained on 205 contrast-en hanced thoraco-abdominal \nCT scans and tested on an independent dataset of 54  scans. Manual \nsegmentation was performed by two radiology residen ts, who identified the \ncenters of the L1 and L3 vertebrae and segmented th e corresponding axial \nslices. Performance was evaluated via mean absolute  error (MAE) for L1/L3 \ndetection, volumetric Dice similarity coefficient ( vDSC) for segmentation, and \nmean percentage relative error (PRE), regression an alysis, and Bland–Altman \nplots for body composition indices estimation. \nResults or Findings: On the independent dataset CompositIA achieved a \nMAE of about 5 mm in detecting slices intersecting the L1 and L3 vertebrae, \nwith a MAE < 10 mm in at least 85% of cases, and a vDSC greater than 0.85 in \nsegmenting axial slices. Regression and Bland–Altma n analyses demonstrated \na strong linear relationship and good agreement bet ween automated and \nmanual scores (p values < 0.001 for all indices), w ith mean PREs ranging from \n5.13% to 15.18%. \nConclusion: CompositIA facilitated automated quantification of body \ncomposition scores, achieving high precision in ind ependent testing. \nLimitations: The main limitation of the study is the small size of the training \nset. \nFunding for this study: Raffaella F. Cabini, Benedikt Thelen, Rolf Krause a nd \nDiego U. Pizzagalli were supported financially by t he grants ExaTrain (to Rolf \nKrause), and FIR (to Diego U. Pizzagalli). \nEthics committee - additional information: This study was approved by the \nlocal Ethics Committee (Comitato Etico Cantonale, R epubblica e Cantone \nTicino, Switzerland; protocol code 2021-00943). All  patients whose CT scans \nwere included in the training set provided informed  consent for the participation \nin the study. \nAuthor Disclosures:  \nAndrea Cozzi: Nothing to disclose \nRolf Krause: Nothing to disclose \nStefania Rizzo: Nothing to disclose \nFilippo Del Grande: Nothing to disclose \nDiego Ulisse Pizzagalli: Nothing to disclose \nRaffaella Fiamma Cabini: Nothing to disclose \nBenedikt Thelen: Nothing to disclose \nSvenja Leu: Nothing to disclose \n \n \nA novel CT-based biological age model, based on aut omated abdominal \nCT biomarkers for accurate longevity prediction \n*J. Garrett*¹, M. Lee¹, A. Pyrros², R. Summers³, M.  Kattan⁴, P. J. Pickhardt¹; \n¹Madison, WI/US, ²Downers Grove, IL/US, ³Bethesda, MD/US,  \n⁴Cleveland, OH/US \n(jgarrett@uwhealth.org) \n \nPurpose or Learning Objective: To derive and test a CT-biological age \n(CTBA) model using explainable fully automated abdo minal CT-based tissue \nbiomarkers predictive of survival in a large adult population. \nMethods or Background: In this retrospective cohort study, an automated \nsuite of explainable CT-based AI algorithms quantif ying skeletal muscle (L3 \nlevel), fat (L3 level), aortic calcification, bone density, and solid organs \n(liver/spleen/kidney volume) was applied to a large  adult cohort undergoing \nabdominal CT between January 2001-January 2021. Mul tivariable Cox \nproportional hazards regression survival analysis w as performed to determine \nfinal CT biomarker selection based on index of pred iction accuracy (IPA). \nUsing all-cause mortality as a primary outcome, the  CTBA model informed only \nby CT biomarkers and blinded to demographics was co mpared to a model \nbased on demographic data (chronological age/sex/ra ce). The model was also \napplied to an external validation cohort of 40,718 adults. \nResults or Findings: 123,281 adults (mean age, 53.6 years [SD 17.4]; 47%  \nwomen) underwent abdominal CT during the study inte rval. Median post-CT \nfollow-up was 5.3 years (IQR,1.9-10.4 years). CT bi omarkers of greatest \nimportance to the model were (in descending order):  muscle attenuation, aortic \ncalcification, visceral fat attenuation, and bone d ensity. The CTBA model \nsignificantly outperformed demographic data for pre dicting longevity (IPA=29.2 \nvs. 21.7; 10-year AUC=0.880 vs. 0.779; p<0.001). Ag e- and sex-corrected \nsurvival HR for highest-vs-lowest risk CTBA quartil e was 8.73 (95% CI,8.14-\n9.36); HR for highest-risk vs remaining quartiles w as 3.13 (95% CI,3.04-3.23). \n\n \n \nSunday \nAbstract-based Programme \n \n 277  \nCTBA model performed well in the external validatio n cohort (IPA=28.6; \nAUC=0.888). \nConclusion: A novel CTBA model informed only by objective fully  automated \n\"opportunistically” derived abdominal CT biomarkers  outperformed a \ndemographics (CA/sex/race) based model and improves  survival prediction. \nLimitations: Data from a single large academic medical center we re used for \nmodel training. \nFunding for this study: None \nEthics committee - additional information: IRB Waiver of consent; \nretrospective analysis. \nAuthor Disclosures:  \nMichael Kattan: Nothing to disclose \nPerry J. Pickhardt: Advisory Board: Nanox AI \nMatthew Lee: Nothing to disclose \nJohn Garrett: Shareholder: NVIDIA Advisory Board: R adUnity \nAyis Pyrros: Nothing to disclose \nRonald Summers: Nothing to disclose \n \n \nEnabling Exchange of Quantitative CT-Assessed Body Composition Data \nusing FHIR: A First Step into Interoperable Body Co mposition Profiling \n*Y. Wen*, J. H. Eil, K. A. Borys, J. Kohnke, K. Arz ideh, J. Haubold, F. Nensa, \nO. Pelka, R. Hosch; Essen/DE \n \nPurpose or Learning Objective: This study aims to demonstrate the \nintegration of AI-generated body composition and or gan measurements from \nCT images with Fast Healthcare Interoperability Res ources (FHIR) to \nstandardise and enhance CT-derived biomarker intero perability. \nMethods or Background: FHIR is a widely used interoperability standard tha t \nenables health information exchange across differen t healthcare systems. With \nthe development of AI models, modern AI application s cannot only analyse the \ndata but also generate relevant data for patient mo nitoring and assessment, \nsuch as models for body composition analysis. The m issing step in advancing \npersonalised medicine is combining AI-generated hea lthcare results with an \ninteroperable and standardised format. Therefore, t his study integrated the \nresults of the Body and Organ Analysis (BOA) model into FHIR profiles, \nincluding measurements of 11 semantic body regions,  seven tissues, and 104 \nlandmarks, to streamline and provide interoperabili ty of CT-derived biomarkers \nin radiology. \nResults or Findings: Two FHIR profiles, Body Composition Analysis \nObservation and Body Structure Volume Observation p rofiles, have been \ndeveloped to capture body composition measurements and record the volume \nof body structure generated from the BOA model, inc orporating terminology \ncoding and references to related FHIR resources. \nConclusion: The presented FHIR profiles provide an interoperabl e format for \nAI-generated body composition data, standardising t he storage and exchange \nof AI-generated biomarkers derived from CT images. The contributed profiles \ncan also be extended in future work to support othe r radiological modalities \n(e.g. MRI) or other image-based AI model biomarkers  (e.g. CT-based bone \nmineral density). \nLimitations: The created profiles focus on tissue and organ volu metrics and \nshould be enhanced to include other available image -based markers and \nimaging modalities. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: This study does not require \nethics committee approval, since no identifiable or  sensitive patient data was \nused. \nAuthor Disclosures:  \nKatarzyna Anna Borys: Nothing to disclose \nJan Horst Eil: Nothing to disclose \nJohannes Haubold: Nothing to disclose \nJudith Kohnke: Nothing to disclose \nYutong Wen: Nothing to disclose \nKamyar Arzideh: Nothing to disclose \nRené Hosch: Nothing to disclose  \nFelix Nensa: Nothing to disclose \nObioma Pelka: Nothing to disclose \n \n \nTransfer of a CT-based 3D body composition analysis  segmentation \nmodel to MRI T2-weighted sequences using a generati ve adversarial \nnetwork \n*C. Bojahr*¹, J. Haubold¹, O. Pollok¹, C. S. Schmid t¹, K. A. Borys¹,  \nM. Mancino², L. Umutlu¹, F. Nensa¹, R. Hosch¹; ¹Ess en/DE, ²Rome/IT \n \nPurpose or Learning Objective: This study aims to adapt CT-based deep \nlearning (DL) models using CycleGAN-based style tra nsfer, enabling accurate \nbody composition analysis (BCA) without extensive m anual annotation on T2-\nweighted MRI sequences. \n \n \nMethods or Background: This study analyzed data from 120 patients (96 \ntrain, 24 test) who underwent whole-body CT and MRI  within 48 hours. A \nCycleGAN was trained to convert CT images to T2-wei ghted MRIs, producing \nsynthetic MRIs that preserve CT structures with MRI  styling. BCA was \nassessed on the corresponding CT scans using the Bo dy and Organ Analysis \n(BOA) framework, and 10 body-region class segmentat ions were transferred to \nsynthetic MRIs to train an initial nnU-Netv2 3D seg mentation model. This \nmodel was used to generate proposals for all 120 MR Is, which two trainees \nunder guidance of a radiologists (with 8 years of e xperience) refined. A second \nmodel was then trained on the refined segmentations , and evaluated by \ncomparing both models to expert annotations using t he Sørensen-DICE score. \nResults or Findings: The comparison between the two models (style transf er \nvs. expert refined) revealed the following DICE-sco res: subcutaneous tissue \n(0.835 vs. 0.978), muscle (0.845 vs. 0.965), abdomi nal cavity (0.943 vs. \n0.988), thoracic cavity (0.895 vs. 0.977), bone (0. 774 vs. 0.919), glands (0.576 \nvs. 0. 899), pericardium (0.697 vs. 0.945), mediast inum (0.731 vs. 0.914), brain \n(0.894 vs. 0.965), spinal canal (0.886 vs. 0.970) a nd the average of all classes \n(0.808 vs. 0.952). \nConclusion: The presented approach shows rapid adaptation of CT  BCA \nmodels to MRI without manual annotation, achieving notable segmentation \nperformance. When refined by experts, these metrics  are further enhanced, \nenabling precise body composition analysis with red uced annotation effort. \nLimitations: Validation of different MRI scanners is necessary t o ensure the \ngeneralizability and robustness of the proposed met hod. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: Informed consent was waived by \nthe ethics committee due to the retrospective setti ng. \nAuthor Disclosures:  \nChristian Bojahr: Nothing to disclose \nKatarzyna Anna Borys: Nothing to disclose \nOlivia Pollok: Nothing to disclose \nJohannes Haubold: Nothing to disclose \nLale Umutlu: Nothing to disclose \nCynthia Sabrina Schmidt: Nothing to disclose \nMatteo Mancino: Nothing to disclose \nRené Hosch: Nothing to disclose \nFelix Nensa: Nothing to disclose \n \n \nDeep Learning-Based Fully Automated Body Compositio n Analysis as a \nPrognostic Factor in ARDS Patients using CT-Based O pportunistic \nBiomarkers \n*J. Kohnke*, K. Schmidt, F. Espeter, K. Pattberg, J . Haubold, F. Nensa,  \nR. Hosch; Essen/DE \n \nPurpose or Learning Objective: Acute Respiratory Distress Syndrome \n(ARDS) is a severe condition with high morbidity an d mortality. Early risk \nassessment is crucial for improving outcomes and gu iding treatment. While \nbody composition parameters have recently emerged a s prognostic factors, \nthey are not commonly considered. However, image-ba sed Body Composition \nAnalysis (BCA) can help extract relevant informatio n about patients. By \nleveraging deep learning, these features can be eff ectively used for enhanced \nrisk stratification using detailed body information . \nMethods or Background: Thoracic CT scans from 960 ARDS patients (37.4 \n% female; median age = 54.7; interquartile range 43 .0 - 64.6), were analyzed. \nThe scans were obtained within two days before or a fter ICU admission. \nExtracted BCA features include lung volume and sarc openia marker (muscle \nvolume / bone volume). Based on the features, terti les were determined \nseparately for both genders (lower tertile vs. othe rs). Kaplan-Meier, Log-Rank, \nand Cox-regression analyses compared 30-days surviv al between the tertiles. \nResults or Findings: Kaplan-Meier analysis revealed significant differen ces in \nsurvival based on lung volume markers (p = 0.02 for  male; p = 0.09 female) \nand sarcopenia (p = 0.01 male; p = 0.52 female). Co x regression indicated that \nLung volume (p = 0.02) and gender (p = 0.01) had si gnificant effects on \nsurvival, while sarcopenia (p = 0.07) was slightly not statistically significant for \nsurvival. \nConclusion: The results suggest that image based BCA from routi ne CT \nimaging could improve risk predictions in ARDS pati ents by using additional \ninformation of the patient's body. \nLimitations: Although BCA parameters show promise, generalizabil ity is \nlimited as all data were from a single center, high lighting the need for validation \nin broader clinical settings. Furthermore, the diff erences between the results \ndepending on gender require further investigation. \nFunding for this study: None \nEthics committee - additional information: This study adhered to all \nguidelines defined by the approving institutional r eview board of the \ninvestigating hospital. The Institutional Review Bo ard waived written informed \nconsent due to the study's retrospective nature. Co mplete anonymization of all \ndata was performed before inclusion in the study. \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 278  \nAuthor Disclosures:  \nKarsten Schmidt: Nothing to disclose \nJohannes Haubold: Nothing to disclose \nJudith Kohnke: Nothing to disclose \nRené Hosch: Nothing to disclose \nKevin Pattberg: Nothing to disclose \nFelix Nensa: Nothing to disclose \nFlorian Espeter: Nothing to disclose \n \n \nEnhancing Autopsy Evaluations with AI-Driven Body C omposition \nBiomarkers from Post-Mortem CT Scans \n*J. Garrett*¹, M. Golden¹, M. Lee¹, S. Berry², N. A ppel³, H. Edgar³,  \nP. J. Pickhardt¹; ¹Madison, WI/US, ²Kalamazoo, MI/U S, ³Albuquerque, NM/US \n(jgarrett@uwhealth.org) \n \nPurpose or Learning Objective: To correlate fully automated post-mortem CT \n(PMCT)-based measures of aortic calcification, skel etal muscle, and intra-\nabdominal fat of decedents with causes of death and  comorbidities. \nMethods or Background: Retrospective study of the New Mexico Decedent \nImage Database (NMDID) with non-contrast PMCT scans  between 2010-2017. \nAn automated pipeline of AI-driven algorithms for q uantifying skeletal muscle, \nsubcutaneous fat, visceral fat, and aortic calcific ation (Agatston score) from the \nabdominal component of PMCT scans was used. Scans w ith more than \nminimal decomposition were excluded. Cause of death  was categorized as \nacute or chronic. CT-based biological age was deriv ed using a predetermined \nmodel. \nResults or Findings: The final cohort included 6638 decedents (mean age 50 \n± 18 years; 74% male). Deaths were classified as 80% acute, 10% chronic, \nand 10% uncertain. Muscle density and area at the L 3 level were higher in the \nacute group compared to the chronic group (26 HU vs . 18 HU, p<0.001; 192 \ncm² vs. 183 cm², p<0.001) and higher in those witho ut cancer (25 HU vs. 16 \nHU, p<0.001; 190 cm² vs. 169 cm², p<0.01). Aortic A gatston scores were \nhigher in heart disease deaths (5120 vs. 2098, p<0. 001). Diabetic patients had \nhigher L3 visceral fat area (227 cm² vs. 175 cm², p <0.001) and lower muscle \ndensity (17 HU vs. 25 HU, p<0.001). The chronic gro up had a larger biological-\nchronological age gap than the acute group (median age gap, 19 years vs. 10 \nyears; p<0.001). \nConclusion: Fully automated quantitative CT-based tissue biomar kers from \nPMCT scans match expectations based on previous stu dies on living patients \nand correlate with acuity of death and chronic co-m orbidities. \nLimitations: The process imperfect of categorizing decedents int o “acute” or \n“chronic” causes of death based on death certificat es is imperfect without \naccounting for all potential medical confounders. \nFunding for this study: None \nEthics committee - additional information: IRB Exempt study; non-human \nsubjects per HIPAA \nAuthor Disclosures:  \nPerry J. Pickhardt: Advisory Board: Nanox AI \nHeather Edgar: Nothing to disclose \nMatthew Lee: Nothing to disclose \nShamsi Berry: Nothing to disclose \nJohn Garrett: Shareholder: NVIDIA Advisory Board: R adUnity \nNicollette Appel: Nothing to disclose \nMax Golden: Nothing to disclose \n \n \nDeep Learning Models for Cardiomegaly Detection Ena bles Assessment \nof Cardiomegaly Prevalence in an International CT D ata Repository: \nInsights from AICT Consortium \n*U. Zidan*¹, N. Bi¹, A. Chandrashekar¹, M. Bown², E . Joviliano³, V. Grau¹,  \nE. R. Ranschaert⁴, R. Lee¹; ¹Oxford/UK, ²leicester/UK, ³São Paulo/BR , \n⁴Ghent/BE \n(usama.zidan@nds.ox.ac.uk) \n \nPurpose or Learning Objective: To develop high-performance ML/DL \npipelines for the detection and characterization of  cardiomegaly in a diverse \ninternational repository of CT scans. \nMethods or Background: The AICT consortium (www.aict.ai) consists of 10 \nclinical sites across 3 continents, contributing CT  scans in an agnostic fashion \nto a common research repository. The ultimate goal is to collect 1 million CT \nstudies, enabling ML/DL training at an unprecedente d scale. . This pilot \nanalysis includes the first 5487 unique individuals  encompassing 1978 chest \nCT scans performed from March 2017 to September 202 4. Two published \nmodels (Superem Total Segmentor) were used to detec t cardiomegaly. \n \n \n \n \n \n \nResults or Findings: Here we report the findings on cardiomegaly (define d as \na cardiothoracic ratio [CTR] > 0.50). Of the 1978 i ndividuals, 1577 did not \nexhibit cardiomegaly (784 males and 793 females), a nd 401 had cardiomegaly \n(178 males and 223 females). The overall prevalence  of cardiomegaly is 20%. \nThe prevalence is higher among females (22%) compar ed to males (19%) \n(p<0.05). The average age of those with cardiomegal y is on average 70 years \nold (range: 21-96) [m:65,(21-96); f:75,(36-95); p<0 .05]. The mean CTR in those \nwith cardiomegaly is 0.55 (±0.05) [m: 0.55 (±0.05); f: 0.56 (±0.06); p=ns]. \nConclusion: The AICT Consortium repository, combined with high- throughput \nML/DL analytic pipelines, provides novel insights i nto the prevalence and \ndemographic distribution of cardiomegaly in a conte mporary international \ncohort. This data enhances our understanding of car diomegaly epidemiology \nand supports the development of advanced detection methods. \nLimitations: [To be added based on study outcomes] \nFunding for this study: Horizon Europe and UK Research Innovation \nEthics committee - additional information: The study was approved by HRA \n(22/HRA/2302) \nAuthor Disclosures:  \nVicente Grau: Nothing to disclose \nAnirudh Chandrashekar: Nothing to disclose \nEdwaldo Joviliano: Nothing to disclose \nRegent Lee: Nothing to disclose \nUsama Zidan: Nothing to disclose \nMatt Bown: Nothing to disclose \nNing Bi: Nothing to disclose \nErik R. Ranschaert: Nothing to disclose \n \n \nDetection of osteoporotic vertebral body compressio n fracture in \ncomputed tomography scans of the chest and abdomen using artificial \nintelligence Nanox.AI \n*V. Mathew*, D. Pearce, N. Kate Rose, S. Saini, E. Bogoch; Toronto, ON/CA \n(vinumathew123@gmail.com) \n \nPurpose or Learning Objective: The detection of undiagnosed vertebral \ncompression fractures (VCFs) is critical due to the ir association with increased \nrisk of future fragility fractures. Primary objecti ve is to evaluate the performance \nof Nanox.AI HealthOST in identifying incidental VCF s on outpatient chest and \nabdomen CT scans by assessing sensitivity, specific ity, PPV, and NPV. \nSecondary objective is to quantify missed VCFs on b y initial reporting \nradiologist. \nMethods or Background: HealthOST is an AI solution from Nanox.AI, \nproviding automatic image analysis of the spine fro m CT images to support \nclinicians in the evaluation and assessment of indi cators of osteoporosis. \nRetrospective analysis on 590 outpatient CT cases f rom St. Michael’s Hospital \nat Unity Health Toronto. Two radiologists, includin g a senior musculoskeletal \nradiologist established a consensus “gold standard”  for comparison with AI \nresults. Two AI thresholds for vertebral height red uction were examined: mild \n(>20%) and moderate (>25%). Original radiologist re ports were reviewed to \nquantify missed VCFs on these scans. \nResults or Findings: At the 20% threshold, AI showed a sensitivity of 91 .1%, \nspecificity of 52.7%, PPV of 17.1%, and NPV of 98.2 %. At the 25% threshold, \nsensitivity decreased to 79.9%, while specificity i mproved to 94.2%, with a PPV \nof 50.7% and NPV of 98.4%. AI increased fracture de tection by 88% compared \nto initial radiologist findings at the 20% threshol d and 92% at the 25% \nthreshold. \nConclusion: Nanox.AI HealthOST shows potential as an effective tool for VCF \nscreening, with high sensitivity at the 20% thresho ld and improved specificity at \n25%. Given the variable specificity and substantial  rate of false positives, a \nsecondary review by radiologists is recommended for  accuracy. Increased \ndetection rate by the AI in comparison to the initi al radiologist report highlights \nthe AI's capability to assist in fracture detection  and enhancing diagnostic \naccuracy. \nLimitations: None \nFunding for this study: AMGEN Inc. \nEthics committee - additional information: Ethics committee approval REB# \n21-183 \nAuthor Disclosures:  \nNoah Kate Rose: Nothing to disclose \nVinu Mathew: Nothing to disclose  \nDawn Pearce: Nothing to disclose \nEarl Bogoch: Research/Grant Support: Amgen \nSidharth Saini: Nothing to disclose \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 279  \n09:30-11:00 Research Stage 3 \nResearch Presentation Session: Chest \nRPS 2304 \nLung cancer imaging: characterisation and \nprognosis \n \nModerator \nR.-I. Milos; Vienna/AT  \n(ruxandra-iulia.milos@meduniwien.ac.at) \n \n \nPulmonary adenocarcinoma: Correlation of Pathologic al Growth Pattern \nand Radiological Morphology on Computed Tomography \nL. Biggemann, H. Bohnenberger, J. Vincke, P. Kraus,  T. Overbeck,  \nA. Hammerstein-Eqourd, *J. Uhlig*; Göttingen/DE \n(johannes.uhlig@med.uni-goettingen.de) \n \nPurpose or Learning Objective: Pulmonary adenocarcinoma (AC) can \npresent with different pathological growth patterns . This study evaluates \nwhether these growth patterns correlate with the ra diological morphology of the \ntumor on computed tomography (CT). \nMethods or Background: Patients with surgically resected pulmonary AC and \npreoperative CT imaging were retrospectively includ ed. Cases were sampled \nto distribute growth patterns approximately evenly.  Pathological growth \npatterns were assessed on a representative patholog ical slice. The \npredominant growth pattern was defined as >=60%. Ra diological morphology \nwas assessed using preoperative thoracic CT scans a nd compared across \ngrowth patterns using the chi-square test. \nResults or Findings: A total of n=345 patients were included (43.8% fema le; \nmedian age 68 years). Pathological growth patterns were acinar (n=68), lepidic \n(n=51), micropapillary (n=58), papillary (n=50), mu cinous (n=51), and solid \n(n=67). While age was balanced across AC growth pat terns, lepidic and \nmicropapillary ACs were more common in women (56.9%  / 51.7%); and solid \nand acinar ACs more likely in men (71.6% / 57.4%; o verall p <0.05). \nPathological growth patterns demonstrated specific morphologies on CT \nregarding nodule type, margin, ground glass opaciti es, central low attenuation, \nair bronchogram, associated lymphadenopathy and loc ation of distant \nmetastases (each variable p<0.05). For example, lep idic ACs commonly \npresented as ground-glass or partially solid nodule s (13.7% / 35%); acinar ACs \nwith associated ground-glass opacities (77.9%); sol id ACs with contact to \npulmonary fissures (50.7%) and central low attenuat ion (58.2%); and mucinous \nACs with air bronchogram (41.2%). \nConclusion: Radiological morphology of pulmonary ACs on CT corr elates well \nwith pathological growth patterns, which could aid guiding diagnostic and \ntreatment patterns. \nLimitations: Limitations include that only a representative path ological slice of \nthe pulmonary AC was assessed, whereas growth patte rns might vary in the \nfull tumor volume; and that CT-imaging was performe d on different CT-\nscanners, introducing heterogeneity. \nFunding for this study: This study has been supported by Siemens \nHealthineers. \nEthics committee - additional information: Ehtics committee of the \nUniversity Medical Center Goettingen \nAuthor Disclosures:  \nHanibal Bohnenberger: Nothing to disclose \nTobias Overbeck: Nothing to disclose \nPaul Kraus: Nothing to disclose \nJohannes Uhlig: Investigator: Siemens Healthineers Advisory Board: Bayer \nJan Vincke: Nothing to disclose \nLorenz Biggemann: Nothing to disclose \nAlexander Hammerstein-Eqourd: Nothing to disclose \n \n \nAI-Based Computer-Aided Volumetry for Invasiveness Evaluation in Lung \nAdenocarcinoma: Influence of Radiation Dose Reducti on and \nReconstruction Algorithms on High-Definition CT \n*Y. Ozawa*, D. Takenaka, H. Nagata, T. Ueda, M. Nom ura, T. Yoshikawa,  \nY. Ohno; Toyoake/JP \n(ykiooster@gmail.com) \n \nPurpose or Learning Objective: To determine the influence of radiation dose \nand reconstruction method on artificial intelligenc e (AI)-based computer-aided \nvolumetry (CADv) for nodule component measurement a nd diagnostic \nperformance to evaluate invasiveness in lung adenoc arcinoma on high-\ndefinition CT (HDCT). \nMethods or Background: 112 consecutive patients with 181 lung \nadenocarcinomas underwent thin-section HDCTs at sta ndard-dose (SDCT: \n9.0±1.8 mGy), reduced-dose (RDCT: 1.7±0.2 mGy) and ultra-low-dose \n(ULDCT: 0.8±0.1 mGy) levels. All HDCT data were rec onstructed with hybrid-\ntype iterative reconstruction (IR) and deep learnin g reconstruction (DLR). \nThen, standard references for solid and GGO compone nts and consolidation-\nto-tumor ratio (CTR) were computationally determine d with the simultaneous \ntruth and performance level estimation (STAPLE) met hod from annotated CT \ndata by three board-certified chest radiologists. T hen, each component volume \nand CTR on all HDCT data were measured by AI-based CADv software. Each \ncomponent volume and consolidation-to-tumor ratio ( CTR) were correlated \nbetween CADv measurement on each CT data and standa rd reference. Then, \nMeasurement differences of each index between stand ard reference and each \nCADv measurement were compared among all CT data by  Tukey’s HSD test. \nFinally, diagnostic performance of invasiveness was  compared among all CTR \nmeasurements by ROC analysis. \nResults or Findings: There were significant correlations for each compon ent \nand volume and CTR on all HDCTs (hybrid-type IR: 0. 71≤r≤0.88, p<0.0001; \nDLR: 0.71≤r≤0.88, p<0.0001). Mean differences between each CADv  \nmeasurement and standard reference had no significa nt differences among all \nHDCTs (p>0.05). Area under the curve (AUC) of each HDCT with DLR (SDCT: \nAUC=0.98, RDCT: AUC=0.98, ULDCT: AUC=0.97) was sign ificantly larger \nthan all HDCTs with hybrid-type IR (AUC=0.95, p<0.0 5). \nConclusion: Reconstruction method rather than radiation dose re duction had \nsignificantly affected to diagnostic performance of  AI-based CADv for \ninvasiveness evaluation in lung adenocarcinoma on H DCT. \nLimitations: Number of subjects \nFunding for this study: Reconstruction method is more important than \nradiation dose reduction for keeping diagnostic per formance of AI-based CADv \nfor invasiveness evaluation in lung ademocarcinoma on high-definition CT. \nEthics committee - additional information: Research grants from Canon \nMedical Systems Corporation and Smoking Research Fo undation \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Smoking Rese arch Foundation \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grants-in-Ai d for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Canon Medica l Systems \nCorporation \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Smoking Res earch Foundation \nResearch/Grant Support: Canon Medical Systems Corpo ration \n \n \nVariability of Enlarged Mediastinal Lymph Node Meas urement in CT: \nManual vs. Automatic Assessment \nA. Olesinski, R. Lederman, Y. Azraq, J. Sosna, *L. Joskowicz*; Jerusalem/IL \n(josko@cs.huji.ac.il) \n \nPurpose or Learning Objective: Detection of enlarged mediastinal lymph \nnodes (LNs) in CECT chest scans is necessary for th e assessment of oncology \npatients. It is performed by manually measuring the  short axis length (SAL) of \nthe LNs. We aimed to quantify the interobserver agr eement and variability of \nthe SAL measurements and to compare them to automat ically computed SALs \nfrom volumetric LN delineations. \nMethods or Background: We retrospectively analyzed 40 CECT chest scans \nfrom patients with mediastinal lymphadenopathy. Two  radiologists with 25 \nyears and 35 years of experience independently meas ured the SALs of 471 \nLNs by: identifying the scan slice in which the LN appears largest, drawing the \nline segment along its short axis; and classifying the LN as normal \n(SAL<10mm) or enlarged (SAL≥10mm). Four weeks later, one of the \nradiologists manually segmented the volume of the L N's along its contours. \nThe LNs were automatically classified as normal/enl arged based on their SALs \ncomputed from the delineation (COMP). Confusion mat rices were computed, \nas well as the differences between the SALs. \nResults or Findings: The normal/enlarged LN overall agreement (371 norma l \nLNs, 49 enlarged LNs) between both radiologists was  94.8% (420/471). For \nagreement/disagreement, the SAL differences (std) w ere 1.2(1.1)mm and \n3.5(3.2)mm. Note that the SALs difference is nearly  twice as large for \ndisagreement as for agreement. The normal/enlarged overall agreement \nbetween the manual and the automatically computed S ALs and both \nradiologists was 94.8%(421/471) and 92.4%(411/471).  \nConclusion: Identification of enlarged mediastinum lymph nodes in chest \nCECT based on short axis measurements derived manua lly or computed from \nlymph nodes delineations has high accuracy. The agr eement of the computed \nSALs from LN contour delineations is within the man ual SALs interobserver \n\n \n \nSunday \nAbstract-based Programme \n \n 280  \nvariability. Accurate automatic LN segmentations ca n be obtained with existing \nmethods. \nLimitations: Single institution and single contour delineation. \nFunding for this study: None. \nEthics committee - additional information: Yes, Helsinki Committee of the \nHadassah University Medical Center \nAuthor Disclosures:  \nRichard Lederman: Consultant: HighRAD \nLeo Joskowicz: Consultant: HighRAD \nJacob Sosna: Consultant: HighRAD \nAlon Olesinski: Consultant: HighRAD \nYusef Azraq: Nothing to disclose \n \n \nMultiparametric 18F-FDG PET/MRI based on restrictiv e spectrum imaging \nand amide proton transfer-weighted imaging facilita tes the assessment of \nlymph node metastases in non-small cell lung cancer  \n*N. Meng*¹, X. Liu¹, J. Pan¹, X. Yu¹, Y. Wu¹, Y. Ya ng², Z. Wang³, M. Wang¹; \n¹Zhengzhou/CN, ²Bei Jing/CN, ³Shanghai/CN \n(821157922@qq.com) \n \nPurpose or Learning Objective: To investigate the value of multiparametric \n18F-FDG PET/MRI based on tri-compartmental restrict ive spectrum imaging \n(RSI), amide proton transfer-weighted imaging (APTW I), and diffusion-\nweighted imaging (DWI) in the assessment of lymph n ode metastases (LNM) \nof non-small cell lung cancer (NSCLC) \nMethods or Background: A total of 152 NSCLC patients were enrolled. 18F-\nFDG PET- derived parameter (SUVmax), RSI-derived pa rameters (f1, f2, and \nf3), APTWI-derived parameter (MTRasym(3.5ppm)), DWI -derived parameter \n(ADC), and were calculated. Logistic regression ana lysis was used to identify \nindependent predictors, and combined diagnostics. A UC, calibration curves \nand decision curve analysis (DCA) were employed to assess the performance \nof the combined diagnostics. \nResults or Findings: MTRasym(3.5ppm), SUVmax, f2, and f3 were higher \nand ADC and f1 were lower in LNM-positive group tha n in LNM-negative group \n(all P < 0.05). Maximum lesion diameter, f1, MTRasy m(3.5ppm) , SUVmax, \nand ADC were independent predictors of LNM status i n NSCLC patients, and \nthe combination of them had an optimal diagnostic e fficacy (AUC = 0.978; \nsensitivity = 95.35 %; specificity = 90.91 %), whic h was significantly higher \nthan maximum lesion diameter, f1, MTRasym(3.5ppm), SUVmax, and ADC \n(AUC = 0.774, 0.810, 0.832, 0.834, and 0.783, respe ctively, and all P < 0.01). \nThe combined diagnosis showed a good performance (A UC = 0.968) in the \nbootstrap (1000 samples)-based internal validation.  Calibration curves and \nDCA demonstrated that the combined diagnosis not on ly provided better \nstability, but also resulted in a higher net benefi t for the patients involved. \nConclusion: Multiparametric 18F-FDG PET/MRI based on RSI, APTWI , and \nDWI is beneficial for the non-invasive assessment o f LNM status in NSCLC, \nand the combination of maximum diameter, f1, MTRasy m(3.5ppm), SUVmax, \nand ADC may serve as a promising biomarker. \nLimitations: This study was conducted at a single institution wi th a relatively \nsmall sample size \nFunding for this study: The National Key R&D Program of China \n(2023YFC2414200), the National Natural Science Foun dation of China \n(82371934), the Joint Fund of Henan Province Scienc e and Technology R&D \nProgram (225200810062). \nEthics committee - additional information: The current study received \napproval from the local ethics committee, and all p articipants provided written \ninformed consent for participation. \nAuthor Disclosures:  \nXue Liu: Nothing to disclose \nYang Yang: Nothing to disclose \nJiayin Pan: Nothing to disclose \nNan Meng: Nothing to disclose \nMeiyun Wang: Nothing to disclose \nYaping Wu: Nothing to disclose \nZhe Wang: Nothing to disclose \nXuan Yu: Nothing to disclose \n \n \nDiagnostic Accuracy in NSCLC Lymph Node Staging wit h Total-Body and \nConventional PET/CT \n*C. Mingels*¹, M. H. Madani¹, F. Sen¹, H. Nalbant¹,  Y. G. Abdelhafez¹,  \nM. Guindani², R. Badawi¹, B. A. Spencer¹, L. Nardo¹ ; ¹Sacramento/US,  \n²Los Angeles, CA/US \n \nPurpose or Learning Objective: To characterize diagnostic accuracy for \nnodal (N)-staging with [18F]FDG Total-Body (TB) and  short-axial field-of-view \n(SAFOV) PET/CT in non-small cell lung cancer (NSCLC ) patients. \nMethods or Background: In this prospective, randomized, single center head -\nto-head comparative study 48 patients underwent TB and SAFOV PET/CT. \n700 nodal levels (1R/L, 2R/L, 3a/p, 4R/L, 5, 6, 7, 8R/L, 9R/L, 10-14R/L) of 28 \npatients could be associated to histopathological f indings, imaging after \nlocalized or systemic treatment, which allowed calc ulation of sensitivity, \nspecificity, positive (PPV) and negative predictive  value (NPV). Thresholds for \nmaximum standardized uptake value (SUVmax), tumor-t o-background ratio \n(TBR), metabolic tumor volume (MTV) and total-lesio n glycolysis (TLG) were \ncalculated. \nResults or Findings: TB and SAFOV PET/CT showed high diagnostic \naccuracy indices for patient-based N-staging. Sensi tivity and specificity were \n86.0% (CI: 77.0-95.0%) and 98.3% (CI: 97.3-99.3%) f or TB; 77.2% (CI: 66.3-\n88.1%) and 97.4% (CI: 96.1-98.6%) for SAFOV PET. PP V was higher for TB \n(81.7%, CI: 71.9-91.5%) compared to SAFOV PET (72.1 %, CI: 60.9-83.4%), \nhowever, this finding was not statistically signifi cant (p=0.08). NPV for TB \n(98.6%, CI: 97.9-99.6%) and SAFOV PET/CT (98.0%, CI : 96.9-99.1%) were \ncomparable (p=0.22). Overall, NSCLC N-staging was a ffected in six cases on \nSAFOV and only in one case on TB PET/CT. Semi-quant itative analysis \nrevealed a SUVmax-threshold of 3.0 to detect TP les ions on both scanners. \nTBR, MTV and TLG thresholds were lower on TB compar ed to SAFOV PET \n(TBR: 1.2 vs. 1.7, MTV: 0.5 ml vs. 1.0 ml and TLG: 1.0 ml vs. 3.0 ml). \nConclusion: TB and SAFOV PET/CT showed high diagnostic accuracy  for N-\nstaging in NSCLC. Sensitivity and PPV on TB PET/CT were slightly higher \ncompared to SAFOV PET/CT. TB PET/CT showed lower ra te of incorrect N-\nstaging and lower semi-quantitative thresholds. \nLimitations: Small sample size, composite reference standard wit h imaging \nFunding for this study: Research reported in this publication was supported  \nby the National Institutes of Health under award nu mber R01CA249422. The \nwork was also supported by the In Vivo Translationa l Imaging Shared \nResources with funds from NCI P30CA093373 and by th e Fred and Julia \nRusch Foundation for Nuclear Medicine Research and Education. Hande \nNalbant’s funding is partially provided by United I maging Health’s UIH \nFellowship Gift. \nEthics committee - additional information: This study was approved by the \nUC Davis institutional review board (IRB 1506448). Written informed consent \nfor inclusion was obtained. The study was performed  in accordance with the \nDeclaration of Helsinki. \nAuthor Disclosures:  \nFatma Sen: Nothing to disclose \nHande Nalbant: Nothing to disclose \nClemens Mingels: Nothing to disclose \nMichele Guindani: Nothing to disclose \nMohammad H. Madani: Nothing to disclose \nRamsey Badawi: Nothing to disclose \nLorenzo Nardo: Nothing to disclose \nBenjamin A. Spencer: Nothing to disclose \nYasser Gaber Abdelhafez: Nothing to disclose \n \n \nWhole-lesion iodine map histogram analysis versus s ingle-slice spectral \nCT parameters for determining of visceral Pleural I nvasion in NSCLC \n*K. Zhu*, J. Zhou; Lanzhou/CN \n(18835359804@163.com) \n \nPurpose or Learning Objective: To evaluate and compare the performances \nof whole-lesion iodine map histogram analysis to th ose of single-slice spectral \nCT parameters in discriminating of visceral pleural  invasion in NSCLC. \nMethods or Background: A total of 99 NSCLC patients underwent \npreoperative spectral CT and were divided into two groups: VPI and non-\nVPI.There were 65 men and 34 women with a mean age of 59.33 ± 8.62 \n(standard deviation) years(range:37-79 years) .The whole-lesion iodine map \nhistogram parameterswere measured for each NSCLC pa tient. By placing \nregions of interest at representative levels of the  tumor and normalizing them, \nspectral CT parameters IC and NIC were obtained. Di scriminating capabilities \nof spectral CT and histogram parameters were assess ed and compared using \narea under the ROC curve (AUC)and logistic regressi on models. \nResults or Findings: The SD, Variance and CV of the iodine map histogram  \nanalysis,and iodine concentration and normalized io dine concentration of \nsingle-slice spectral CT parameters were significan tly different of visceral \npleural invasion in NSCLC (P < 0.001 to P = 0.03). The CV of histogram \nparameters (AUC=0.65; 95% [CI]: 0.54-0.76) and norm alized iodine \nconcentration (AUC=0.75; 95% CI: 0.64-0.85) from sp ectral CT parameters \nhad the best performance for distinguishing whether  visceral pleural invasion \noccurred in NSCLC. At ROC curve analysis no signifi cant differences in AUC \nwere found between histogram parameters (AUC = 0.84 ; 95% CI: 0.76-0.91) \nand spectral CT parameters (AUC = 0.66; 95% CI: 0.5 5-0.77) (P = 0.24). \nConclusion: Both whole-lesion iodine map histogram analysis and  single-slice \nspectral CT parameters help discriminate whether of  visceral pleural invasion \nin NSCLC, and the single-slice spectral CT paramete rs performed better in \nterms of diagnostic efficacy. \nLimitations: This finding has not been validated in an independe nt population, \nlimiting their generalizability. Future prospective  studieswith larger and external \npatient cohorts are warranted. \nFunding for this study: This research has been supported by the National \nNatural Science Foundation of China (82371914). \n\n \n \nSunday \nAbstract-based Programme \n \n 281  \nEthics committee - additional information: The study was approved by the \nInstitutional Ethics Committee (2021A-498) and exem pted from patient \ninformed consent. \nAuthor Disclosures:  \nJunlin Zhou: Nothing to disclose \nKaibo Zhu: Nothing to disclose \n \n \nMRI-Based Molecular Imaging vs. FDG-PET/CT: Capabil ity for \nPostoperative Recurrence Prediction with FDG-PET/CT  in Stage I NSCLC \nPatients \n*Y. Ozawa*, H. Nagata, T. Ueda, M. Nomura, T. Yoshi kawa, D. Takenaka,  \nY. Ohno; Toyoake/JP \n(ykiooster@gmail.com) \n \nPurpose or Learning Objective: To compare the prediction capability for \npostoperative recurrence among FDG-PET/CT and MRI-b ased molecular \ninformation from chemical exchange saturation trans fer (CEST) imaging and \ndiffusion-weighted imaging (DWI) in stage I non-sma ll cell lung cancer \n(NSCLC) patients. \nMethods or Background: 79 pathologically diagnosed and surgically treated \nNSCLC patients who underwent CEST imaging, DWI and FDG-PET/CT, \nfollow-up and pathological examinations were includ ed in this study. According \nto the follow-up and pathological examination resul ts, all patients were divided \nas recurrence (n=13) and non-recurrence (n=66) grou ps. In each lesion, \nmagnetization transfer ratio asymmetry at 3.5ppm (M TRasym), apparent \ndiffusion coefficient (ADC) and SUVmax of each nodu le were assessed by ROI \nmeasurements. To compare all indexes between two gr oups, Student’s t-test \nwas performed. To determine the significant predict ors, multiple logistic \nregression analysis was performed. Then, ROC analys is was performed to \ncompare distinguishing two groups among all indexes  and combined significant \npredictors. Finally, sensitivity (SE), specificity (SP) and accuracy (AC) were \ncompared among all methods by McNemar’s test. \nResults or Findings: There was significant difference of each index betw een \ntwo groups (p<0.05). Multiple logistic regression a nalyses determined \nMTRasym (Odds ratio [OR]: 1.31, p=0.03) and ADC (OR : 0.002, p=0.008) as \nsignificant predictors. When applied each threshold  value, SPs and ACs of \nMTRasym (SP: 81.8%, AC: 82.2%), ADC (SP: 87.9%, AC:  86.1%) and \ncombined predictors (SP: 89.4%, AC: 89.9%) were sig nificantly higher than \nthose of SUVmax (SP: 69.7%, p<0.05; AC: 72. %, p<0. 05). Moreover, AC of \ncombined predictors was significantly higher than t hat of MTRasym (p=0.03). \nConclusion: MRI-based molecular information has better predicti on capability \nfor postoperative recurrence than FDG-PET/CT in sta ge I NSCLC patients. \nLimitations: Limited study cohort number and follow-up periods i n some \npatients are considered as limitations in his study . \nFunding for this study: Canon Medical Systems Corporation \nEthics committee - additional information: Fujita Health University Hospital \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Smoking Rese arch Foundation \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grant-in-Aid  for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Canon Medi cal Systems \nCorporation Research/Grant Support: Grants-in-Aid f or Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Canon Medic al Systems \nCorporation Research/Grant Support: Smoking Researc h Foundation \n \n \nInterstitial lung abnormalities are significant poo r prognostic factors in \nresected clinical stage Ⅰ non-small cell lung cancer \n*T. Hino*, T. Akamine, T. Hida, K. Sagiyama, Y. Yam asaki, K. Tabata,  \nK. Ishigami; Fukuoka/JP \n(hino.takuya.372@m.kyushu-u.ac.jp) \n \nPurpose or Learning Objective: Interstitial lung abnormalities (ILA) are \nknown to be associated with increased mortality; ho wever, the impact of ILA on \npost-operative prognosis of non-small cell lung can cer (NSCLC) remains \nunclear. The aim of study is to assess the associat ion between mortality in \npost-operative clinical stage Ⅰ NSCLC patients and the presence of ILA. \n \n \n \n \nMethods or Background: We retrospectively evaluated the patients who \nunderwent chest CT, followed by radical resection f or clinical stage Ⅰ NSCLC \nfrom 2006 to 2018. The presence of ILA was evaluate d on high-resolution CT \nimages by two radiologists and one thoracic surgeon . Five-year overall survival \n(OS), recurrence-free survival (RFS), and cumulativ e incidence of other cause \nof death stratified by the presence of ILA were ass essed. \nResults or Findings: Among 709 patients included in this study, 80 patie nts \nhad ILA (11.2%). Five-year OS and RFS were signific antly lower in patients \nwith ILA than those without ILA (43.6% vs. 90.1 %, log-rank test p<0.001; \n44.7% vs. 80.6 %, log-rank test p<0.001). Multivari able analysis demonstrated \nthat the presence of ILA was an independent poor pr ognostic factor in both OS \nand RFS (HR: 1.45, 95% CI: 1.14–1.84, p=0.003; HR: 1.37, 95% CI: 1.08–\n1.74, p=0.010, respectively). Five-year cumulative incidence of other cause of \ndeath was significantly higher in patients with ILA  than those with non-ILA \n(21.1% vs. 5.4%, Gray’s test p<0.001). \nConclusion: The presence of ILA affected the cause of other dea th and was \nan independent poor prognostic factor in clinical s tage Ⅰ NSCLC. \nLimitations: All the participants were retrospectively collected . ILA was not \nconformed with pathological specimens. \nFunding for this study: N/A \nEthics committee - additional information: Kyushu University Hospital \nIRB No. 23240-00 \nAuthor Disclosures:  \nKosuke Tabata: Nothing to disclose \nKousei Ishigami: Nothing to disclose \nYuzo Yamasaki: Nothing to disclose \nAkamine Takaki Akamine: Nothing to disclose \nKoji Sagiyama: Nothing to disclose \nTakuya Hino: Nothing to disclose \nTomoyuki Hida: Nothing to disclose \n \n \nIntratumoral and peritumoral CT radiomics in predic ting anaplastic \nlymphoma kinase mutations and survival in patients with lung \nadenocarcinoma: a multicenter study \n*G. Lin*, W. Chen, J. Ji; Lishui/CN \n \nPurpose or Learning Objective: To explore the value of intratumoral and \nperitumoral radiomics in preoperative prediction of  anaplastic lymphoma kinase \n(ALK) mutation status and survival in patients with  lung adenocarcinoma. \nMethods or Background: We retrospectively collected data from 505 eligible  \npatients with lung adenocarcinoma from four hospita ls (training and external \nvalidation sets 1–3). The CT-based radiomics featur es were extracted \nseparately from the gross tumor volume (GTV) and GT V incorporating \nperitumoral 3-, 6-, 9-, 12-, and 15-mm regions (GPT V3, GPTV6, GPTV9, \nGPTV12, and GPTV15), and screened the most relevant  features to construct \nradiomics models to predict ALK (+). The combined m odel incorporated \nradiomics scores (Rad-scores) of the best radiomics  model and clinical \npredictors was constructed. Performance was evaluat ed using receiver \noperating characteristic (ROC) analysis. Survival o utcomes were examined \nusing the Cox proportional hazards model. \nResults or Findings: The GPTV3 radiomics model using a support vector \nmachine (SVM) algorithm achieved the best predictiv e performance, with the \nhighest average AUC of 0.811 in the validation sets . Clinical TNM stage and \npleural indentation were independent predictors. Th e combined model \nincorporating the GPTV3-Rad-score and clinical pred ictors achieved higher \nperformance than the clinical model alone in predic ting ALK (+) in three \nvalidation sets (AUC: 0.855 vs. 0.648, 0.882 vs. 0. 634, 0.810 vs. 0.663). The \nprediction score of the combined model could strati fy survival outcome in \npatients receiving ALK-TKI therapy (P=0.026) and im munotherapy (P=0.012). \nConclusion: The presented combined model based on GPTV3 effecti vely \nmined tumor features to predict ALK mutation status  and stratify survival in \npatients with lung adenocarcinoma. \nLimitations: This is a retrospective study and may have varying degrees of \nselection bias. \nFunding for this study: This research was funded by the National Natural \nScience Foundation of China (Grant No.82072026 to J iansong Ji), Key Project \nof Joint Construction by Provincial and Ministerial  Authorities (Grant No.WKJ-\nZJ-2452 to Minjiang Chen), Medical and Health Gener al Project of Zhejiang \nProvince (Grant No. 2024KY568 to Weiyue Chen, and N o. 2023KY425 to \nGuihan Lin). \nEthics committee - additional information: This study was approved by the \nInstitutional Review Boards, and the requirement fo r informed consent was \nwaived due to retrospective nature. \nAuthor Disclosures:  \nJiansong Ji: Nothing to disclose \nWeiyue Chen: Nothing to disclose \nGuihan Lin: Nothing to disclose \n \n\n \n \nSunday \nAbstract-based Programme \n \n 282  \n09:30-11:00 Research Stage 4 \nResearch Presentation Session: \nGenitourinary \nRPS 2307 \nMalignant lesions of the female pelvis: \nadvances in imaging techniques, \ndiagnostic and follow-up \n \nModerator \nA. M. Hötker; Zurich/CH  \n \n \nA Machine Learning Model Based on Endometrium MRI R adiomics to \nPredict Histological Diagnosis From Biopsy in Subje cts at Risk of \nEndometrial Cancer: Pilot study \n*R. V. Ninkova*, M. Gennarini, V. Miceli, A. Cupert ino, F. Curti, C. Catalano,  \nL. Manganaro; Rome/IT \n(roberta.ninkova@gmail.com) \n \nPurpose or Learning Objective: Aim of this study was to develop a machine \nlearning model based on Magnetic Resonance Imaging (MRI) to stratify the \nsingle-subject risk of endometrial cancer (EC). \nMethods or Background: From September 2023 to July 2024, we collected \nMRI images from 41 patients. Among these subjects, 15 patients (36.6%) \nbelonged to class \"microsatellite instability (MSI) \" and 26 patients (63.4%) \nbelonged to class \"microsatellite stability (MSS)\",  according to histological \ndiagnosis from biopsy. This image set was used for the training and cross-\nvalidation of different machine learning models. A robust radiomic approach \nwas applied, under the hypothesis that radiomic fea ture could be able to \ncapture the disease heterogeneity among the two gro ups. \nResults or Findings: Three models consisting of 3 ensembles of machine-\nlearning classifiers (random forests, support vecto r machines and k-nearest \nneighbor classifiers) were developed for the binary  classification task of \ninterest (“MSI” vs. “MSS”), based on supervised lea rning, using histological \ndiagnosis from biopsy as reference standard. The be st model showed ROC-\nAUC mean value of 84 % [78.4-89.8], accuracy mean v alue of 75.6% [69.6-\n81.7], sensitivity mean value of 68.9 % [59.3-78.5] , specificity mean value of \n79.5 % [74-85], PPV mean value of 71.66 %[58.1-73.9 ], and NPV mean value \nof 81.6% [76.3-86.9] (p <0.005 mean value). \nConclusion: The radiomics-based machine learning model achieved  a high \ndiagnostic performance in the molecular stratificat ion of patients with EC, \nwhich could improve risk stratification and support  clinical therapeutic decision. \nThis is a preliminary study which could provide a n ew perspective for patients \nwith EC, allowing a complete and accurate identific ation of the disease and \npromoting personalized treatment. \nLimitations: The primary limitation of this study is the small p atient cohort. \nExpanding the sample size and investigating additio nal molecular pathological \nprofiles will be necessary for further validation. \nFunding for this study: None \nEthics committee - additional information: The study was performed in line \nwith the principles of the Declaration of Helsinki.  \nAuthor Disclosures:  \nMarco Gennarini: Nothing to disclose \nValentina Miceli: Nothing to disclose \nFederica Curti: Nothing to disclose \nAngelica Cupertino: Nothing to disclose \nRoberta Valerieva Ninkova: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLucia Manganaro: Nothing to disclose \n \n \nFIGO 2023 staging of endometrial cancer: is there s till a role for \nradiology? \n*A. Rame*¹, S. Bottazzi¹, G. Avesani¹, M. Bonatti²,  V. Celli¹, E. Perrone¹,  \nT. Pasciuto¹, B. Gui¹, E. Sala¹; ¹Rome/IT, ²Bolzano /IT \n(annecopper95@gmail.com) \n \nPurpose or Learning Objective: To investigate the role of MRI in the \npreoperative evaluation of endometrial cancer (EC) considering the new FIGO \n2023 staging system. \nMethods or Background: Patients diagnosed with EC from two institutions \nbetween 2019 and 2023 were retrospectively included . Inclusion criteria were \nthe availability of preoperative MRI and biopsy, mo lecular data and definitive \nhistopathological data obtained after surgery. Two radiologists retrieved MRI \nfindings from reports and combined them with biopsy  results (grading and \nhistology) to determine the FIGO 2023 preoperative stage. Definitive \nhistopathological and molecular data served as the gold standard (final \nstaging). The preoperative evaluation was compared with final staging (FIGO \nstages I, II, III and IV as independent categories)  and discrepancies were \nrecorded. \nResults or Findings: 231 patients were included. The agreement between \npreoperative evaluation and final staging was 74% ( 171/231), while the FIGO \nstage was discordant at 26% (60/231). Causes of dis cordance were: lymph \nnode (LN) involvement in 26.7% (16/60), changes in grading or histological \nsubtype between biopsy and surgical specimen in 23. 3% (14/60), metastases \nand carcinosis in 15% (9/60: 5 in the upper abdomin al peritoneum, 1 in the \npelvic peritoneum and 3 distant), the presence of s ubstantial lymphovascular \nspace invasion (LVSI) in 18.2% (10/55). Less common  causes of \ndiscrepancies were myometrial invasion (6.7%; 4/60) , cervical stroma invasion \n(5%; 3/60), ovarian involvement (3.3%; 2/60) and va ginal invasion (1.7%; \n1/60). Overall, parameters not assessable preoperat ively (LVSI and definitive \ngrading and histological subtype) accounted for sta ging discrepancies in 10.4% \n(24/231) of cases. \nConclusion: Our study showed that the inclusion of parameters n ot \nassessable preoperatively in the new FIGO 2023 stag ing system does not \nsignificantly diminish the role of MRI in the preop erative evaluation of EC, \nresulting in staging discrepancy in only approximat ely 10% of cases. \nLimitations: No limitations. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: ID 6194, approved on 7/12/2023 \nAuthor Disclosures:  \nMatteo Bonatti: Nothing to disclose \nTina Pasciuto: Nothing to disclose \nGiacomo Avesani: Nothing to disclose \nEmanuele Perrone: Nothing to disclose \nBenedetta Gui: Nothing to disclose  \nVeronica Celli: Nothing to disclose \nSilvia Bottazzi: Nothing to disclose  \nEvis Sala: Nothing to disclose \nAnna Rame: Nothing to disclose \n \n \nValue of enhanced t1 mapping MR imaging in the eval uation of the depth \nof myometrial invasion in endometrial cancer: compa red with dynamic \ncontrast-enhanced MR imaging \n*X. Liu*, Z. Yuan, Y. Li, J. Ren, Y. He, H. Xue, Z.  Jin; Beijing/CN \n(liuxinyu960815@gmail.com) \n \nPurpose or Learning Objective: To compare the diagnostic efficiency of \nenhanced T1 mapping MR imaging and dynamic contrast -enhanced (DCE) \nimaging for assessing the depth of myometrial invas ion in patients with \nendometrial cancer. \nMethods or Background: 46 women diagnosed with endometrial cancer \nunderwent preoperative MR imaging. Two radiologists  independently assessed \nthe depth of myometrial invasion, categorized as no  myometrial invasion, \nsuperficial myometrial invasion, or deep myometrial  invasion, on \nT2WI+DWI+DCE MR imaging, followed by T2WI+DWI+enhan ced T1 mapping \nMR imaging, four weeks later. The findings were the n compared to \nhistopathological examinations. The diagnostic perf ormance comparison was \nconducted using the chi-square test. \nResults or Findings: The overall accuracy for accessing depth of myometr ial \ninvasion on T2WI+DW+DCE and T2WI+DW+enhanced T1 map ping were \n76.1%, 80.4% for reader 1, and 78.3%, 80.4% for rea der 2, respectively. The \nincrement was not statistically significant for eit her reader. While assessing the \nabsence of myometrial invasion, the precision, sens itivity, and specificity \nachieved by both radiologists using DCE were 100%, 16.7%, and 100%, \nwhereas for enhanced T1 mapping the precisions were  100%, with sensitivities \nof 50% and 33%, and specificities of 100%. In evalu ating superficial \nmyometrial invasion, these values using DCE were 82 .4%/82.8%, \n84.8%/87.9%, and 53.8%; using enhanced T1 mapping, these values were \n87.5%/85.3%, 84.8%/87.9%, and 69.2%/61.5%. Assessme nt of deep \nmyometrial invasion achieved these values of 54.5%/ 60%, 85.7%, and \n87.2%/89.7% with DCE; 54.5%/60%, 85.7%, 87.2%/89.7%  with enhanced T1 \nmapping. Inter-reader agreement, measured with kapp a values, was 0.831 \nwith DCE-MRI and 0.899 with enhanced T1 mapping. Bo th radiologists \nconcurred that enhanced T1 mapping substantially im proved diagnostic \nconfidence over DCE-MRI. \nConclusion: Enhanced T1 mapping demonstrates superior diagnosti c \nefficiency in the evaluation of myometrial invasion  in endometrial cancer \ncompared with DCE MR imaging. \nLimitations: This was a pilot study with a small sample size. \nFunding for this study: This work was supported by grants from Natural \nScience Foundation of China (grant No. 82271886). \nEthics committee - additional information: This retrospective study was \napproved by the institutional review board, which w aived the requirement of \ninformed consent. \n\n \n \nSunday \nAbstract-based Programme \n \n 283  \nAuthor Disclosures:  \nJing Ren: Nothing to disclose \nYuan Li: Nothing to disclose \nZhilin Yuan: Nothing to disclose \nZhengyu Jin: Nothing to disclose \nHuadan Xue: Nothing to disclose \nYonglan He: Nothing to disclose \nXinyu Liu: Nothing to disclose \n \n \nMRI in early-stage cervical cancer after cone biops y: can it predict the \npresence of residual tumor? \n*A. Amerighi*, M. Dolciami, A. Napoletano, I. Isufi , G. Avesani, E. Sala, B. Gui; \nRome/IT \n \nPurpose or Learning Objective: Cervical cancer (CC) is increasingly \ndiagnosed at earlier stages due to cancer screening  programs. Cone biopsy is \noften necessary to rule out invasive disease and ca n sometimes suffice for \ntreatment; however, determining residual disease is  essential for treatment \nplanning. Despite its crucial role in managing CC, the role of Magnetic \nResonance Imaging (MRI) in determining residual dis ease in patients after \nconization remains poorly understood, with only a f ew studies focusing on this \ntopic. We aimed to assess MRI accuracy in detecting  residual disease after \ncervical conization for early-stage cancer and comp are the diagnostic \nperformance of MRI with and without a contrast agen t. \nMethods or Background: We retrospectively enrolled all patients with early -\nstage CC who received conization before MRI and the n surgically treated \n(hysterectomy, trachelectomy, or re-conization). Tw o radiologists evaluated \nMRI scans for residual disease in the cervix, blind ed to surgical outcomes. \nResults or Findings: 119 patients were included in the study. MRI showed  an \naccuracy of 78,18%, sensitivity of 65.45%, specific ity of 90.91%, positive \npredictive value (PPV) of 87.80%, and negative pred ictive value (NPV) of \n72.46%. There was no significant change in the MRI performance with and \nwithout contrast medium, in accuracy (81,05% vs 75, 80%, respectively), \nsensitivity (76,92% vs 55,17%), specificity (85,19%  vs 96,43%), PPV (83,33% \nvs 94,12%), and NPV (79,31% vs 67,50%). \nConclusion: MRI showed good accuracy in assessing residual tumo r after \nconization with high specificity and PPV; however, the main problem remains \nthe high number of false negatives. The use of cont rast medium did not \nsignificantly affect the MRI's performance, as it l ed to a slight increase in true \nnegatives but also an increase in false positives. \nLimitations: Retrospective study \nFunding for this study: None \nEthics committee - additional information: ID prot. 7073 prot. 0022234/24 \nAuthor Disclosures:  \nAndrea Amerighi: Nothing to disclose \nGiacomo Avesani: Nothing to disclose \nMiriam Dolciami: Nothing to disclose \nIna Isufi: Nothing to disclose \nAnna Napoletano: Nothing to disclose \nBenedetta Gui: Nothing to disclose \nEvis Sala: Nothing to disclose \n \n \nDiagnostic performance of DWI and ADC in characteri zing the N \nparameter classified according to the Node-RADS sco re in patients with \ncervical cancer \n*M. Gennarini*, R. V. Ninkova, V. Miceli, A. Cupert ino, S. Riccardi, F. Curti,  \nC. Cutonilli, C. Catalano, L. Manganaro; Rome/IT \n(marco.gennarini@hotmail.it) \n \nPurpose or Learning Objective: This study evaluates the diagnostic accuracy \nof the Node-RADS score in Magnetic Resonance Imagin g (MRI) and assesses \nthe significance of the Apparent Diffusion Coeffici ent (ADC) in identifying actual \nlymph node neoplastic involvement in patients with cervical cancer (CC). \nMethods or Background: A retrospective analysis was performed on cervical \ncancer patients who underwent preoperative MRI and radical surgery with \nlymphadenectomy from February 2018 to July 2024. Ly mph node involvement \nrisk was evaluated for the primary pelvic lymph nod es, assigning scores \nranging from 1 to 5: 1 (very low), 2 (low), 3 (uncl ear), 4 (high), and 5 (very \nhigh). The mean ADC, relative ADC (rADC), and corre cted ADC (cADC) for \nlymph nodes rated as Node-RADS 3, 4, and 5 were mea sured and compared \nagainst lymph node histology results. \nResults or Findings: In this study, 156 lymph nodes from 54 patients, wi th a \nNode-RADS score greater than 2, were included. Of t hese, 108/156 (69.2%) \nwere histologically confirmed as positive, while 48 /156 (30.8%) were negative. \nThe mean ADC value proved most statistically signif icant, showing a sensitivity \nof 87.0%, specificity of 82.6%, a positive predicti ve value (PPV) of 92.2%, and \na negative predictive value (NPV) of 73.1%. The are a under the curve (AUC) \nvalues for Node-RADS >2 and Node-RADS 3 were 0.892 and 0.677, \nrespectively, with ADC thresholds of <0.963×10^-3 m m^2/s and <0.983×10^-3 \nmm^2/s. \nConclusion: The ADC measurement of lymph nodes provides crucial  data that \naids in the accurate classification of patients wit h cervical cancer. Utilizing an \nADC cut-off of 0.963×10^-3 mm^2/s, the MRI demonstr ated high diagnostic \nsensitivity. Thus, ADC serves as a valuable tool fo r enhancing the diagnostic \nprocess in cervical cancer management. \nLimitations: Single center retrospective design study. \nFunding for this study: None \nEthics committee - additional information: N/A \nAuthor Disclosures:  \nMarco Gennarini: Nothing to disclose \nClaudia Cutonilli: Nothing to disclose \nValentina Miceli: Nothing to disclose \nFederica Curti: Nothing to disclose \nSandrine Riccardi: Nothing to disclose \nAngelica Cupertino: Nothing to disclose \nRoberta Valerieva Ninkova: Nothing to disclose \nCarlo Catalano: Nothing to disclose \nLucia Manganaro: Nothing to disclose \n \n \nAssessing the feasibility of Magnetic Resonance Ima ging Compilation for \ndetermining the treatment strategies and predicting  the recurrence risk \nfactors and short-term efficacy in cervical cancer \n*X. Ou*, Y. Li, Y. Pei, W. Li; Hunan/CN \n(2631141379@qq.com) \n \nPurpose or Learning Objective: To investigate the feasibility of MAGiC (sy-\nT2WI; sy-T1, sy-T2, and sy-PD maps) to determine tr eatment plan and predict \nrecurrence risk factors (RRF) ，short-term treatment efficacy (STE) in cervical \ncancer (CC) patients using high-resolution T2-weigh ted (hr-T2WI）and \ndiffusion-weighted imaging (DWI) as reference stand ards. \nMethods or Background: 194 patients suspected of CC were enrolled. For \nCC underwent CCRT, hrT2WI performed in 2 months for  evaluating STE \n(completed response and no-CR ). MAGiC can generate  synthetic morphologic \nimages (syT2WI) and quantitative synthetic images ( sy-T1, T2 and PD maps). \nFor syT2WI, The evaluation of image quality and sta ging using sy-T2WI and \nhr-T2WI was conducted. The accuracy, sensitivity an d specificity of sy-T2WI \nwere analyzed for making treatment strategies (IB-I IA: surgery; IIB-IVA: \nCCRT). The AUC was used to predict RRF and STE, use  quantitative sy-T1, \nT2, PD maps. \nResults or Findings: 69 out of 119 CC(IIB－IVA) received CCRT . 50 out of \n119 CC(IA-IIA) received surgery. SyT2WI was no sign ificant differences with \nhrT2WI in four aspects (P＞0.05). The accuracy, sensitivity and specificity of  \nsy-T2WI was 0.908, 0.908 ,0.999 for differentiating  IB-IIA from IIB-IVA, and an \nexcellent agreement between them (k = 0.935; p < 0. 001). T2, T1 and ADC \nvalues had a significant differences to identified CR from no-CR and identified \nRRF from no-RRF (P < 0.05) but PD.The diagnostic pe rformance of ADC was \ninferior to T2 for STE, which was similar to T2 for  RRF. Furthermore, T1 ＋T2 \nwas superior to ADC for predicting RRF (AUC: 0.980 vs. 0.776; p = 0.005) and \nforecasting STE (AUC: 0.982 vs. 0.737; p < 0.001) .  \nConclusion: MAGiC is a promising technique for deciding therape utic planning \nand predicting RRF and STE in CC , which is similar  and even superior to hr-\nT2WI and DWI. \nLimitations: Not applicable \nFunding for this study: Not applicable \nEthics committee - additional information: the local Research Ethics \nCommittee \nAuthor Disclosures:  \nWenzheng Li: Nothing to disclose \nYigang Pei: Nothing to disclose  \nYue Li: Nothing to disclose \nXiaorong Ou: Nothing to disclose \n \n \nA retrospective comparative study of diagnostic alg orithms for the \ndifferential diagnosis of uterine leiomyoma and sar coma: preliminary \nresults \n*C. Vercelli*¹, F. Rosa², C. Martinetti¹, D. Schett ini¹, A. Gastaldo², N. Gandolfo¹; \n¹Genova/IT, ²Savona/IT \n(catevercelli@gmail.com) \n \nPurpose or Learning Objective: In 2022, an important Consensus Statement \nand diagnostic algorithm for assessing the risk of uterine leiomyosarcoma \n(LMS) using MRI was published (Hindman et al.). The  aim of our study is to \ncompare the diagnostic performance of the recently proposed algorithms \navailable in the literature for the differential di agnosis of uterine mesenchymal \ntumors on MRI. \nMethods or Background: After a literature review, three diagnostic algorit hms \nwere identified: Hindman et al. (2022), Rosa et al.  (2023), and Wahab et al. \n(2020). A retrospective evaluation was conducted on  50 MRIs performed for \nsuspected uterine mesenchymal masses. Each lesion w as categorized in a \n\n \n \nSunday \nAbstract-based Programme \n \n 284  \ndouble-blind manner by an expert Radiologist in gyn ecological imaging \naccording to the three algorithms. Histological exa mination or appropriate \nfollow-up was used as reference standard for the fi nal diagnosis. \nResults or Findings: The study included 52 uterine masses: 37 benign \nlesions (leiomyomas) and 15 malignant lesions (7 LM S, 2 STUMP, 2 \nendometrial stromal sarcomas, 1 adenosarcoma, and 3  non-sarcomatous \nmalignant lesions). The three algorithms demonstrat ed equal specificity in \ndiagnosing LMS (97.3%). However, when analyzing the  performance in the \ndifferential diagnosis between malignant and benign  lesions (including rarer \nhistotypes), the Rosa et al. algorithm showed super ior sensitivity (93.33% vs \n73.33%) and diagnostic accuracy (96.15% vs 90.38%).  \nConclusion: The differential diagnosis of uterine mesenchymal l esions \npresents a diagnostic challenge with significant im plications for management \nand outcome. The three algorithms demonstrated high  diagnostic accuracy; in \nparticular, the Rosa et al. algorithm proved more e ffective in identifying also \nrarer malignant histotypes \nLimitations: Retrospective study; Small cohort with high prevale nce of \nsarcomas. \nFunding for this study: No funds \nEthics committee - additional information: N. Registro CER Liguria: \n78/2023 - DB id 12883 \nN. CET - Liguria: 104/2024 - DB id 13756 \nAuthor Disclosures:  \nNicoletta Gandolfo: Nothing to disclose \nDaria Schettini: Nothing to disclose \nCarola Martinetti: Nothing to disclose \nAlessandro Gastaldo: Nothing to disclose \nFrancesca Rosa: Nothing to disclose \nCaterina Vercelli: Nothing to disclose \n \n \nClues for a new MR scoring system of uterine mesenc hymal tumors \nE. Zlotykamien Taieb, *D. Gherman*; Vincennes/FR \n(dia.gherman@yahoo.com) \n \nPurpose or Learning Objective: To externally validate a previous MRI-based \nexpert consensus algorithm and evaluate the potenti al improvement of an MR-\nscoring system's accuracy in diagnosing uterine mes enchymal tumors. \nMethods or Background: A bicentric retrospective observational cohort stud y \nwas conducted from January 2018 to December 2023 in cluding women with a \npathological diagnosis of myometrial tumor followin g a pelvic MRI within six \nmonths. Clinical and MR criteria were recorded blin dly by two radiologists. \nContinuous variables were analyzed using a Mann–Whi tney test, and \ncategorical variables using Fisher’s exact test. Od ds ratios for predicting \nmalignancy were calculated with 95% confidence inte rvals and p-values. \nResults or Findings: The cohort included 455 women with mesenchymal \ntumors: 437 leiomyomas, 2 STUMPs, and 16 leiomyosar comas. Using initial \ncriteria (pelvic hypertrophic lymph nodes, T2W sign al intensity, DW signal \nintensity compared to endometrium, and ADC cutoff v alue of 0.9 × 10⁻³ \nmm²/sec), the model accurately classified 420 out o f 455 cases (Accuracy: \n80.9%, sensitivity was 61.1% and specificity 93.6%.  A refined approach added \n“irregular tumor margins” and menopausal status, mo dified DW signal \ncompared to bladder, and an ADC cutoff value of 1.2 3 × 10⁻³ mm²/s, improving \nclassification to 445 out of 455 cases (Accuracy: 9 2.5%; sensitivity: 83.3% ; \nspecificity: 98.4%. The refined algorithm significa ntly improved accuracy, \nallowing the development of a 5-category scoring sy stem. \nConclusion: MR imaging effectively differentiates leiomyosarcom a from other \nuterine tumors. The new algorithm increases diagnos tic accuracy, helping \nprevent morcellation risks in women with uterine le iomyosarcoma \nLimitations: Our study is retrospective, which does not permit a voidance of all \nbiases. Even though the cohort was bicentric, the p revalence of malignancy \nremained low. Due to the sample size, differentiati on between STUMP and \nfrankly invasive UMT was not possible. No external validation of our modified \nscore is yet available, necessitating further studi es. \nFunding for this study: None \nEthics committee - additional information: Institutional ethics committee \napproval and waiver of informed consent (CRM-2405-4 10) \nAuthor Disclosures:  \nDiana Gherman: Nothing to disclose \nEva Zlotykamien Taieb: Author: conceiving and writi ng this study \n \n \nMagnetic resonance spectroscopy integration with mu ltiparametric MRI: \nEnhancing diagnostic precision in sonographically i ndeterminate adnexal \nmasses \n*D. Garg*, R. Kaur, R. Bedi, R. Gupta, B. Goel, U. Handa; Chandigarh/IN \n(gargdollphy@gmail.com) \n \nPurpose or Learning Objective: Proton magnetic resonance spectroscopy \n(¹H-MRS) is a non-invasive imaging technique that o ffers insights into \nbiochemical metabolism. While its role in brain and  prostate malignancies is \nwell-established, its application in evaluating adn exal masses is still in its early \nstages. This study investigates the role of ¹H-MRS in characterising adnexal \nmasses and its value, in conjunction with dynamic c ontrast-enhanced MRI \n(DCE-MRI) and diffusion-weighted imaging (DWI), in enhancing the diagnostic \naccuracy of conventional MRI for differentiating ad nexal masses. \nMethods or Background: We conducted a prospective study including 62 \nhistologically confirmed adnexal masses (19 benign and 43 malignant), which \nwere indeterminate on ultrasound. Patients underwen t conventional MRI, DWI \n(apparent diffusion coefficient), DCE-MRI (time int ensity curves), and ¹H-MRS. \nSingle-voxel spectroscopy analysed resonance peaks for choline, N-acetyl \naspartate (NAA), creatine, lactate, and lipids. Cho line-to-creatine ratios were \ncompared between benign and malignant tumours, and ROC curves were \nused to define optimal thresholds. \nResults or Findings: Conventional MRI showed sensitivity, specificity, p ositive \npredictive value (PPV), and negative predictive val ue (NPV) of 97.67%, \n57.89%, 84%, and 91.67%, respectively. We detected choline peak in 100% of \nmalignant and 47.4% of benign masses, NAA in 79.1% of malignant and \n31.6% of benign masses, as well as lipid peaks in 3 6.8% of benign and 20.9% \nof malignant masses. The mean choline-to-creatine r atio is 1.05+0.55 in benign \nand 12.18+12.38 in malignant tumours, statistically  significant (p<0.05). With a \ncholine-to-creatine threshold of 2.01, sensitivity,  specificity of 100% is \nachieved. The addition of ¹H-MRS, DWI, and DCE-MRI improved the \ndiagnostic accuracy of conventional MRI, with 100% sensitivity, 94.74% \nspecificity, 97.73% PPV, and 100% NPV. \nConclusion: 1H-MRS has promising role in characterising adnexal  masses \nand in conjunction with DWI and DCE-MRI, enhances t he diagnostic accuracy \nof conventional MRI. \nLimitations: The sample size is limited. \nFunding for this study: No funding was provided for this study. \nEthics committee - additional information: The ethics committee notification \ncan be found under the number ECR/658/Inst/PB/2014/ RR-20 \nAuthor Disclosures:  \nUma Handa: Nothing to disclose \nRekha Gupta: Nothing to disclose \nRaveena Bedi: Nothing to disclose \nDollphy Garg: Nothing to disclose \nRavinder Kaur: Nothing to disclose  \nBharti Goel: Nothing to disclose \n \n \n11:30-12:30 Research Stage 1 \nResearch Presentation Session: \nAbdominal and Gastrointestinal \nRPS 2401 \nAcute abdominal diseases and imaging of \nthe bowel \n \nModerator \nM. Zins; Paris/FR  \n(mzins@hpsj.fr) \n \n \n0.5-mSv ultra-low dose appendiceal CT using deep le arning–based \ndenoising algorithm: a comparison with conventional  2.0-mSv low dose \nCT \n*B. J. Choi*, W. Chang, J. H. Hwang, J. Cho, Y. J. Lee, Y. H. Kim, S. H. Park, \nJ. Y. Choi; Seongnam-si/KR \n \nPurpose or Learning Objective: To demonstrate that 0.5-mSv ultra-low dose \nCT using deep learning–based denoising algorithm (D LA) has non-inferiority in \ndiagnosing acute appendicitis, compared to conventi onal 2.0-mSv low dose \nCT. \nMethods or Background: We used 2.0-mSv CT images of 30 patients with \nsuspected appendicitis from the prior prospective s tudy. The original 2.0-mSv \nCT were reconstructed using iterative model reconst ruction (IMR). We \nsimulated 0.5-mSv CT images from the original 2.0-m Sv CT. Then we applied \nIMR and DLA, resulting in three CT image groups per  patient (IMR 2.0-mSv, \nIMR 0.5-mSv, and DLA 0.5-mSv groups). Six radiologi sts (three abdominal and \nthree non-abdominal radiologists) rated the likelih ood of appendicitis on a five-\npoint Likert scale. Primary end point was compariso n of the pooled area under \nthe receiver operating characteristic curve (AUC) b etween DLA 0.5-mSv CT \nand IMR 2.0-mSv CT, with a non-inferiority margin o f 0.06. Secondary end \npoints included comparison of AUC between DLA 0.5-m Sv CT and IMR 0.5-\nmSv CT, diagnostic sensitivity/specificity. \nResults or Findings: The AUC of DLA 0.5-mSv CT was non-inferior to that of \nIMR 2.0-mSv CT [AUC difference: 0.003 (95% CI: -0.0 11, 0.017)]. The AUC of \n\n \n \nSunday \nAbstract-based Programme \n \n 285  \nDLA 0.5-mSv CT was slightly higher among the non-ab dominal radiologists \ncompared to IMR 0.5-mSv CT [AUC difference: 0.034 ( 95% CI: -0.130, 0.198)]. \nDiagnostic sensitivity/specificity were 100% (9/9)/ 95% (20/21) for all readers \nwith both DLA 0.5-mSv CT and IMR 2.0-mSv CT. Howeve r, the sensitivities of \ntwo non-abdominal radiologists with IMR 0.5-mSv CT were mildly \ncompromised [78% (7/9) and 89% (8/9), respectively] . \nConclusion: 0.5-mSv ultra-low dose CT using DLA was non-inferio r to \nconventional 2.0-mSv low dose CT using IMR in diagn osing acute appendicitis. \nLimitations: The limitations of the study are 1) 0.5-mSv CT imag es were \nsimulations, not real data and 2) small sample size , collected from a single \ntertiary hospital. \nFunding for this study: Funding was provided by the National Research \nFoundation of Korea(NRF) grant funded by the Korea government(MSIT) \n(NRF-2022R1F1A1072570). \nEthics committee - additional information: The institutional review board \napproved this study, and the requirement for inform ed consent was waived (B-\n2401-879-112). \nAuthor Disclosures:  \nSo Hyun Park: Nothing to disclose \nJungheum Cho: Nothing to disclose \nJi Young Choi: Nothing to disclose \nYoon Jin Lee: Nothing to disclose \nWon Chang: Nothing to disclose \nYoung Hoon Kim: Nothing to disclose \nByung Jin Choi: Nothing to disclose \nJin Hee Hwang: Nothing to disclose \n \n \nUpright CT vs. Supine CT: Diagnostic Capabilities f or Inguinal Hernias \nand Subtypes in Emergency Department \n*T. Yoshikawa*, H. Nagata, T. Ueda, M. Nomura, D. T akenaka, Y. Ozawa,  \nY. Ohno; Toyoake/JP \n(yoshikawa0816@aol.com) \n \nPurpose or Learning Objective: Upright CT is recently and clinically set in \nour institution, and the purpose of this study was to directly compare \ncapabilities for diagnosis and subtype classificati on of inguinal herniation \nbetween upright CT (uCT) and conventional supine CT  (sCT). \nMethods or Background: 258 consecutive patients who suspected inguinal \nhernia underwent sCT and uCT within a week, surgica l treatment or follow-up \nexamination. From this cohort, 120 inguinal hernias  and computationally \nselected 120 out of 396 non-inguinal hernia were vi sually assessed by two \nboard certified general and abdominal radiologists by 5-point scales as well as \nsubtypes of hernia. Inter-observer agreements for p robability of hernia and \nsubtype were assessed by kappa statistics with χ2 test. Then, ROC analysis \nwas performed to compare diagnostic performance bet ween two CTs. Then, \nsensitivity (SE), specificity (SP) and accuracy (AC ) were compared each other \nby McNemar’s test. Moreover, subtype classification  accuracy (SAC) was also \ncompared between uCT and sCT by McNemar’s test. \nResults or Findings: Inter-observer agreement for probability of hernia were \ndetermined as significant and almost perfect on uCT  (κ=0.84, p<0.0001) and \nsubstantial on sCT (κ=0.77, p<0.0001), and that for subtype classificati on were \nalso significant and almost perfect on both CTs (uC T: κ=0.83, p<0.0001; sCT: \nκ=0.81, p<0.0001). Area under the curve (AUC) of uCT  (AUC=0.99) were \nsignificantly larger than that of sCT (AUC=0.97, p< 0.05). SE and AC of uCT \n(SE=92.5%, AC=96.3%) were significantly higher than  those of sCT (SE: \n80.8%, p<0.0001; AC: 90.4%, p<0.0001). SAC of uCT ( 87.5%) was \nsignificantly higher than that of sCT (73.3%, p<0.0 001). \nConclusion: Upright CT has better diagnostic performance for in guinal hernia \nand subtype classification than conventional supine  CT in routine clinical \npractice. \nLimitations: Lack of clinical outcome evaluation \nFunding for this study: Research grant from Canon Medical Systems \nCorporation \nEthics committee - additional information: Fujita Health University Hospital \nAuthor Disclosures:  \nYoshiyuki Ozawa: Research/Grant Support: Grant-in-A id for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Smoking Rese arch Foundation \nMasahiko Nomura: Nothing to disclose \nTakahiro Ueda: Research/Grant Support: Grant-in-Aid  for Scientific Research \nfrom the Japanese Ministry of Education, Culture, S ports, Science and \nTechnology \nDaisuke Takenaka: Nothing to disclose \nHiroyuki Nagata: Research/Grant Support: Grants-in- Aid for Scientific \nResearch from the Japanese Ministry of Education, C ulture, Sports, Science \nand Technology Research/Grant Support: Canon Medica l Systems \nCorporation \nTakeshi Yoshikawa: Nothing to disclose \nYoshiharu Ohno: Research/Grant Support: Smoking Res earch Foundation \nResearch/Grant Support: Canon Medical Systems Corpo ration \n \nDual-Energy CT of Gastrointestinal Bleeding - Influ ence on diagnostic \naccuracy and reader confidence \n*M. Oberparleiter*, H-C. Breit, J. Vosshenrich, P. Hehenkamp, A. C. Seifert,  \nA. Kobe, C. J. Zech, M. Obmann; Basel/CH \n \nPurpose or Learning Objective: Current guidelines suggest replacing \nunenhanced images with DECT-derived virtual non-con trast images (VNC) in \nsuspected upper GI bleeding based on only two clini cal studies. Our study \ncompares diagnostic accuracy, reader confidence, an d reading time of a \nconventional triphasic versus a dual-energy CT prot ocol in patients with upper \nand lower GI bleeding. \nMethods or Background: This retrospective study included 52 patients with \nactive GI bleeding (22 upper, 30 lower) and 52 cont rols who underwent non-\ncontrast, arterial, and portal-venous phase abdomin al CT. For each case, a \ntriphasic conventional CT dataset and a DECT datase t with VNC, iodine \nimages, and arterial and portal venous phase images  were created. Two \nresidents and two fellowship-trained abdominal radi ologists evaluated all cases \nfor active GI bleeding. Radiation dose and reading time were recorded. \nDiagnostic confidence was rated on a 5-point Likert  scale. Inter-reader \nagreement was assessed using Fleiss' kappa. Sensiti vity and specificity were \ncompared using McNemar's test, reading time, and re ader confidence with the \nWilcoxon signed-rank test. \nResults or Findings: Inter-reader agreement was substantial ( \u0007=0.80). \nSensitivity and specificity for detecting GI bleedi ng using conventional CT did \nnot differ from DECT (91% and 95%, vs. 93% and 96%,  p=0.30 and p=0.77, \nrespectively). Subgroup analysis of lower GI bleedi ng showed a sensitivity of \n88% in conventional CT versus 93% in DECT (p=0.18).  Diagnostic confidence \nincreased from 4(IQR, 4-5) to 5(IQR, 4-5) when usin g DECT (p<0.01). Mean \nreading time per case was 102 s for both datasets ( p=0.62). Total DLP without \ntrue unenhanced images was 21% lower. \nConclusion: DECT-derived VNC and iodine images can replace true  non-\ncontrast images when searching for GI-bleeding. Gui delines should be \nextended to include lower-GI-bleeding. \nLimitations: Sample size was moderate, the study had a single-ce nter design, \nand only dual-source and split-beam DECT scanners w ere used. \nFunding for this study: This research received no specific grant from any \nfunding agency in the public, commercial, or not-fo r-profit sectors. \nEthics committee - additional information: The need for informed consent \nwas waived due to the retrospective nature of this study. \nAuthor Disclosures:  \nMarkus Obmann: Nothing to disclose \nHanns-Christian Breit: Nothing to disclose \nChristoph Johannes Zech: Nothing to disclose \nAdrian Kobe: Nothing to disclose \nPaul Hehenkamp: Nothing to disclose \nJan Vosshenrich: Nothing to disclose \nAlina Carolin Seifert: Nothing to disclose \nMoritz Oberparleiter: Nothing to disclose \n \n \nCorrelation of CT-derived Quantitative Image Featur es and Inflammatory \nLaboratory in Pyelonephritis \n*A. W. Marka*, M. Graf, S. Ziegelmayer, M. R. Makow ski, A. Sauter, T. Huber; \nMunich/DE \n(alexander.marka@gmail.com) \n \nPurpose or Learning Objective: To investigate the relationship between \ninflammatory laboratory markers and quantitative CT -derived imaging features \nin patients with acute pyelonephritis (APN). \nMethods or Background: In this single-center retrospective study, we \nevaluated patients with clinical symptoms of APN at  our institution from \nDecember 2018 to April 2024. Inclusion criteria com prised APN symptoms, \nelevated inflammatory markers (CRP and/or WBC), and  CT morphologic signs \nof APN. Exclusion criteria included concomitant acu te pathology and poor \nimage quality. A total of 102 patients (mean age 60 .1±18.8) were initially \nidentified; 14 were excluded due to acute pathology  and 5 due to poor image \nquality, leaving 83 for final analysis. CT scans fo llowed a standardized \nprotocol, and two radiologists blinded to clinical and lab data conducted image \nanalysis. Inflammatory markers were collected on th e scan day. Statistical \nanalyses included Spearman correlation, Mann-Whitne y-U tests, and linear \nregression. \nResults or Findings: Spearman correlation analysis revealed strong posit ive \ncorrelations between CRP levels and both total volu me (r=0.76, p<0.001) and \npercentage (r=0.71, p<0.001) of renal perfusion def icit. Multivariate linear \nregression showed total perfusion deficit volume ex plained 54.5% of CRP \nvariability (p<0.001). WBC count also correlated si gnificantly with total volume \n(r=0.379, p<0.001) and percentage (r=0.374, p<0.001 ) of perfusion deficit. \nProcalcitonin levels moderately correlated with fat  stranding area (r=0.482, \np=0.0014) but not other CT features. Locoregional l ymphadenopathy was \nsignificantly associated with elevated CRP and WBC counts, but not \nprocalcitonin levels. \n\n \n \nSunday \nAbstract-based Programme \n \n 286  \nConclusion: Quantitative CT-derived features, particularly rena l perfusion \ndeficits, are significantly associated with inflamm atory markers in APN. These \nfindings suggest CT imaging can serve as a surrogat e for inflammation \nseverity, potentially guiding clinical management. Further research is needed \nto explore the clinical implications of these assoc iations. \nLimitations: -Small cohort -Scarcity of procalcitonin levels -No  correlation with \nduration of a patient's hospitalization \nFunding for this study: None \nEthics committee - additional information: Data collection, processing, and \nanalysis were approved by the institutional review board (protocol number \n180/17S), and informed consent was waived. \nAuthor Disclosures:  \nAndreas Sauter: Nothing to disclose \nMarkus Graf: Nothing to disclose \nThomas Huber: Nothing to disclose \nMarcus R. Makowski: Nothing to disclose \nAlexander Wolfgang Marka: Nothing to disclose \nSebastian Ziegelmayer: Nothing to disclose \n \n \nFast abdominopelvic T2-weighted imaging with deep l earning \nreconstruction for acute abdomen: feasibility study  \nJ. Xu, L. Zhu, W. Liu, *Y. Lu*, J. Liu, C. Ma, Y. Z hang, X. Wang, F. Feng; \nBeijing/CN \n(lytong9815@163.com) \n \nPurpose or Learning Objective: To evaluate the image quality and diagnostic \nperformance of single-shot fast spin-echo T2 weight ed imaging with deep \nlearning reconstruction (SSFSE-DL) in volunteers an d patients with acute \nabdomen, in comparison to SSFSE without deep learni ng reconstruction \n(SSFSE-nonDL) and conventional PROPELLER sequences.  \nMethods or Background: Thirty-five healthy volunteers, as well as 35 patie nts \nwith acute abdominal pain from emergency room were prospectively enrolled. \nAbdominopelvic MRI at 3T was performed using three T2-weighted imaging \nsequences: SSFSE-DL (acquisition time: 34s), SSFSE- nonDL, and \nconventional PROPELLER (acquisition time: 2-3min), in random order. Two \nblinded radiologists independently evaluated the ov erall image quality, noise, \nmotion artifacts and clarity of major abdominopelvi c organs. Diagnostic \nconfidence for the presence or absence of common ab dominopelvic diseases \nwas rated on a 1-5 Likert scale. Signal-to-noise ra tio (SNR), contrast-to-noise \nratio (CNR) as well as image noise for the liver, p ancreas and spleen were also \nquantified. Intra- and inter- observer agreement we re assessed, and \ncomparisons of image quality and diagnostic perform ance between the three \nsequences were made. \nResults or Findings: Intra- and inter- observer agreement for the qualit ative \nanalysis and diagnostic performance were good to ex cellent (0.776-0.967). \nSSFSE-DL yielded significant higher SNR and CNR, an d lower noise than \nSSFSE-nonDL and PROPELLER in both volunteers and pa tients (all P<0.05). \nSSFSE-DL obtained significantly higher image qualit y and lower noise than \nSSFSE-nonDL and PROPELLER (both P<0.05). SSFSE-DL a nd SSFSE-\nnonDL had significantly lower motion artifacts and better clarity of major \nabdominopelvic organs than PROPELLER (both P<0.05).  The AUC for \ndetecting common abdominopelvic diseases in SSFSE-D L (0.977-1) and \nSSFSE-nonDL (0.887-1) were significantly higher tha n PROPELLER (0.585-\n0.953). \nConclusion: SSFSE-DL achieved superior image quality and diagno sis \nperformance for volunteers and patients with acute abdomen. \nLimitations: The number of patients with positive diagnosis for each specific \ndisease was relatively small. \nFunding for this study: This study was funded by the National Natural \nScience Foundation of China (82371950) \nEthics committee - additional information: This prospective single-center \nstudy was approved by the local institutional revie w board, and informed \nconsent was obtained from all participants prior to  inclusion in the study. \nAuthor Disclosures:  \nWei Liu: Nothing to disclose \nFeng Feng: Nothing to disclose \nYitong Lu: Nothing to disclose \nChenxue Ma: Nothing to disclose \nXuan Wang: Nothing to disclose \nJia Xu: Nothing to disclose  \nLiang Zhu: Nothing to disclose \nJingjuan Liu: Nothing to disclose  \nYifei Zhang: Nothing to disclose \n \n \n \n \n \n \n \nDiagnostic performance of low-dose abdominal CT wit h artificial \nintelligence iterative reconstruction for acute pan creatitis \n*X. Zhang*¹, S. Zhong², G. Zhang², X. Zhou¹; ¹Chong qing/CN, ²Shanghai/CN \n(db421521@163.com) \n \nPurpose or Learning Objective: To characterize the diagnostic performance \nof low-dose (LD) abdominal CT combined with artific ial intelligence iterative \nreconstruction (AIIR) for assessing acute pancreati tis based on CT severity \nindex (CTSI). \nMethods or Background: A total of 30 patients with acute pancreatitis who \nunderwent follow-up CT examination were prospective ly enrolled. All patients \nunderwent standard-dose (SD) CT followed by LD-CT i n the same breath hold, \nwhere an immediate LD-scan was added in the portal venous phase. The SD-\nprotocol was 120 kVp, ref. 141 mAs, and hybrid iter ative reconstruction (HIR), \nwhereas the LD-protocol was 120kVp, ref. 50mAs and AIIR reconstruction. To \nobtain the CTSI, SD- and LD-CT images at portal ven ous phase were \nindependently scored by two radiologists for assess ing pancreatic \ninflammation, necrosis, and extrapancreatic complic ations. Signal-to-noise \nratio (SNR) and contrast-to-noise ratio (CNR) of he althy and inflamed \nparenchyma were measured and calculated. \nResults or Findings: Compared to SD-CT, LD-CT examination achieved a \n65.18% reduction in effective radiation dose for th e portal venous phase (6.06 \n± 1.28 mSv vs. 2.11 ± 0.45 mSv, p < 0.05). Based on CTSI scoring, LD-AIIR \nwas found comparable to SD-HIR in evaluating the se verity of acute \npancreatitis (5.02 ± 1.42 vs. 5.11 ± 1.78, p = 0.86). Inter-observer agreement \nfor assessing the severity of acute pancreatitis wa s excellent (k = 0.89). LD-\nAIIR showed superior conspicuity compared to SD-HIR  for both the inflamed \n(SNR: 1.96 ± 0.83 vs. 1.74 ± 0.98; CNR: 4.17 ± 1.58 vs. 2.11 ± 0.87; both p < \n0.05) and the healthy parenchyma (SNR: 8.03 ± 1.71 vs. 4.91 ± 1.04; CNR: \n3.06 ± 1.84 vs. 1.78 ± 1.03; both p < 0.05). \nConclusion: AIIR allows for significant radiation dose reductio n without \ncompromising image quality or diagnostic performanc e for the evaluation of \nacute pancreatitis. \nLimitations: n/a \nFunding for this study: n/a \nEthics committee - additional information: Ethics committee of Chongqing \nUniversity Jiangjin Hospital \nAuthor Disclosures:  \nXue Zhou: Nothing to disclose \nXiufu Zhang: Nothing to disclose \nGuozhi Zhang: Nothing to disclose \nSihua Zhong: Nothing to disclose \n \n \nSBOM-AI TRIAL: Set up and validation of AI-based au tomatic total Small \nBowel length Measurement using CT and MRI in Obese patients \ncandidates for metabolic surgery \n*M. Zerunian*¹, S. Nardacci¹, N. Petrucciani¹, I. T oniolo², D. De Santis¹,  \nD. Caruso¹, C. G. Fontanella², G. Silecchia¹, A. La ghi¹; ¹Rome/IT, ²Padova/IT \n(marta.zerunian@gmail.com) \n \nPurpose or Learning Objective: Total small bowel length(TSBL) is crucial to \nachieve successful metabolic/bariatric surgery. A n on-invasive measurement of \nthe TSBL will impact on surgical strategy to avoid short-bowel syndrome after \nsurgery. Cross-sectional imaging(CSI) might play an  important role by \nmeasuring TSBL non-invasively. We aimed to set up a  reliable AI-based \nautomatic method using preoperative CSI to measure the TSBL in candidate to \nbariatric/metabolic surgery. \nMethods or Background: This multicentre prospective TRIAL included \npatients eligible for bariatric surgery(BMI >35 kg/ m2 and at least one obesity-\nrelated comorbidity,BMI>40 kg/m2) underwent the sam e day MRI-and CT-\nenterography before bariatric surgery. TSLB assesse d right before the surgery \nat the operation table and, TSLB <250 cm considered  the cut-off as risk for \ndeveloping short-bowel syndrome. TSLB obtained on M RI and CT by manual \nsegmentation(Slicer3D). A Convolutional Neural Netw ork with U-NET \ndeveloped, consisting in a contracting path followe d by an expansive path, \nRELU activation and a softmax activation function t o reduce the feature map. \nThe Adam optimization algorithm with a constant lea rning rate was used for the \nlearning process. Training stopped after 100 epochs . DICE coefficient were \ncalculated to quantify the accuracy of the predicti on compared to annotated-by-\nradiologist images. \nResults or Findings: Fifty patients enrolled (27 female,age range 26-54 years \nold,mean BMI 41.47). Patients underwent surgery sho wed a TSBL mean \nmeasured intraoperatively of 652.85±87.37cm. CSI showed good concordance \nwith TSBL mean of 589.08±82.95cm(k= 0.70).DICE coef ficient of the training \nset showed a DICE score ranging between 35%-45%, co nfirmed in the \nvalidation set. All methods correctly categorized t he patients according to the \ncut-off considered as risk factor to develop short- bowel syndrome. \nConclusion: Automatic AI-based segmentation of small bowel on n on-invasive \nCSI might be a useful tool to assess obese patients  to personalize the \ntreatment and reduce complications. \nLimitations: Small sample size \n\n \n \nSunday \nAbstract-based Programme \n \n 287  \nFunding for this study: Italian Ministry of University and Research (MUR) \nResearch Projects of Significant National Interest – PRIN (ID: MUR \n2022MPAE29_003) \nEthics committee - additional information: Multicenter interventional study \nAuthor Disclosures:  \nNiccolò Petrucciani: Nothing to disclose \nDamiano Caruso: Nothing to disclose \nMarta Zerunian: Nothing to disclose \nDomenico De Santis: Nothing to disclose \nIlaria Toniolo: Nothing to disclose \nGianfranco Silecchia: Nothing to disclose \nAndrea Laghi: Nothing to disclose \nChiara Giulia Fontanella: Nothing to disclose \nStefano Nardacci: Nothing to disclose \n \n \nVascular enhancement in single-pass abdominal CT: E ffects of a fixed \ninjection duration in patients with non-traumatic a cute abdomen \n*A. Stanzione*, V. Arpaia, A. E. Antonini, R. Liuzz i, L. Sommella, L. Mannacio, \nA. Brunetti, L. Camera; Naples/IT \n(arnaldostanzione@yahoo.it) \n \nPurpose or Learning Objective: To evaluate the effects of a fixed injection \nduration (FID) on vascular enhancement in a Single- Pass (SP) abdominal CT \nperformed in patients with nontraumatic acute abdom en (ANTA). \nMethods or Background: 100 patients (58M, 42F; aged 52±20 yrs ) with \nANTA underwent a SP contrast-enhanced CT (Somatom D rive, Siemens) \nperformed using a Single Source at either 80 kVp (G roup A; BMI 19±3), 100 \nKvp (Group B; BMI 25±4) or 120 Kvp (Group C; BMI 30 ±2.5) . In all groups a \nnon-ionic iodinated contrast media (370 mgI/ml) was  administered as follows: \nGroup A (0.37 grI/Kg); Group B (0.52 grI/Kg); Group  C (0.63 grI/Kg). All \npatients underwent a SP protocol with a FID (50 sec ) and a tailored scan delay \n(SD). In all patients Signal- (SNR) and Contrast-to -Noise Ratios (CNR) were \ncalculated for the abdominal aorta (AA) and the mai n portal vein (MPV) using \nthe psoas muscles as reference tissue. Statistical analysis was performed with \nANOVA (p < 0.05). \nResults or Findings: No significant differences were observed in the \ndemographics of either Group A (20M/12F; 47±20 yrs) or B (36M/22F; 54±19 \nyrs) whereas a female preponderance was observed in  Group C (2M/8F; \n50±18 yrs). Despite significant differences (p < 0.001) were observed in both \nthe volumes (53±11 vs 103±17 vs 141±19 ml) as well as the injection rates \n(1.1±0.2 vs 2.1±0.4 vs 2.8±0.4 ml/sec) of the contrast media for Group A, B \nand C, respectively, SNR and CNR were not significa ntly different for both AA \nand MPV. \nConclusion: SP performed with a FID results in a consistent vas cular \nenhancement. \nLimitations: Unbalanced sample size \nFunding for this study: None \nEthics committee - additional information: Local IRB approval \nAuthor Disclosures:  \nLuigi Mannacio: Nothing to disclose \nLaura Sommella: Nothing to disclose \nValerio Arpaia: Nothing to disclose \nLuigi Camera: Nothing to disclose \nRaffaele Liuzzi: Nothing to disclose \nAndrea Ennio Antonini: Nothing to disclose \nArnaldo Stanzione: Nothing to disclose \nArturo Brunetti: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n11:30-12:30 Research Stage 2 \nResearch Presentation Session: Cardiac \nRPS 2403 \nCardiac imaging: insights from trials \n \nModerator \nN. Lama; Athens/GR  \n(niklampatr@gmail.com) \n \n \nAssociation between metabolic phenotypes and MRI-de rived cardiac \nfunction parameters, major adverse cardiovascular e vents, and mortality: \nInsights from the UK Biobank \n*B. Bogner*, M. Jung, M. Reisert, J. Maushagen, S. Rospleszcz, C. L. Schlett, \nF. Bamberg, J. Weiß, J. Taron; Freiburg/DE \n \nPurpose or Learning Objective: To evaluate the association between \nmetabolic phenotypes, MRI-derived cardiac function,  major adverse \ncardiovascular events (MACE), and all-cause mortali ty. \nMethods or Background: We analyzed 22,348 UK Biobank (UKBB) \nparticipants who underwent cardiac MRI. Obesity was  defined as BMI \n≥30kg/m²; metabolically unhealthy as the presence of  >=1 metabolic syndrome \ncomponents or diabetes resulting in four phenotypes : metabolically healthy \nnon-obese (MHN), metabolically healthy obese (MHO),  metabolically unhealthy \nnon-obese (MUN), metabolically unhealthy obese (MUO ). Associations \nbetween MRI-derived left ventricular ejection fract ion (LVEF, %), cardiac index \n(cardiac output/body surface area [L/min/m2]), MACE , and all-cause mortality \nwere investigated using uni-/multivariable linear a nd Cox regression analyses \nadjusted for age, sex, and smoking status. \nResults or Findings: Among 22,348 UKBB participants (mean age 64.1±7.5 \nyears, 48.8% male), 45.4% were MHN, 6.0% MHO, 36.5%  MUN and 12.1% \nMUO. Over a median follow-up of 5.2 years, 52 MACE events and 371 deaths \noccurred. Compared to MHN, both obese phenotypes sh owed a significant \nnegative association with LVEF whereas a positive a ssociation was found for \nMUN (p<=0.02). For cardiac index, only MUO showed a  significant positive \nassociation compared to MHN (p=0.03). Cox regressio n revealed a >5-fold \nhigher risk of MACE for both obese phenotypes (MHO (HR 5.53 [95% CI 2.10-\n14.58], p<0.001; MUO (HR 5.24 [95% CI 2.37-11.57], p<0.001) compared to \nMHN. A similar pattern was seen for all-cause morta lity (MHO (1.57 [1.04, \n2.36], p=0.03; MUO (1.61 [1.20, 2.17]; p=0.001). \nConclusion: Alterations in metabolic health are linked to diffe rences in cardiac \nfunction, MACE and all-cause mortality risk indepen dent of age, sex, and \nsmoking. This suggests distinct mechanisms affectin g cardiac health and \nemphasizes the need to consider both metabolic heal th and obesity for \npersonalized risk assessment. \nLimitations: Findings may not be generalizable to non-UK populat ions or \nethnic groups underrepresented in the UKBB. \nFunding for this study: None \nEthics committee - additional information: Approved by the UK Biobank. \nAuthor Disclosures:  \nSusanne Rospleszcz: Nothing to disclose \nChristopher L. Schlett: Nothing to disclose \nMarco Reisert: Nothing to disclose \nJakob Weiß: Nothing to disclose \nBalazs Bogner: Nothing to disclose \nMatthias Jung: Nothing to disclose \nJana Taron: Nothing to disclose \nJuliane Maushagen: Nothing to disclose \nFabian Bamberg: Nothing to disclose \n \n \nThe aging heart: Associations between cardiac struc ture, function, and \ndemographic factors in a population-based study \n*B. J. Kerber*¹, T. Küstner², S. Gatidis³; ¹Zurich/ CH, ²Tübingen/DE,  \n³Stanford, CA/US \n \nPurpose or Learning Objective: Exploring the relationship between \ndemographic factors and quantitative imaging phenot ypes based on cardiac \nMR (CMR) in the NAKO population study to provide in sights into the aging \nheart and advance the understanding of cardiovascul ar disease. \nMethods or Background: Steady-state free precession short-axis CMR full-\ncycle sequences from 29,104 participants of the NAK O study, aged 19 to 74 \nyears (16,201 male and 12,903 female, aged 47.9 +/-  12.4, resp. 48.7 +/- 12.2 \nyears), were analyzed. A custom-trained nnU-Net mod el was used to segment  \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 288  \nimages into left ventricle (LV), leftventricular my ocardium, and right ventricle \n(RV). From these segmentations, key cardiac metrics  including LV mass \n(LVM), LV and RV end-diastolic and end-systolic vol umes \n(LVEDV/RVEDV/LVESV/RVESV), ejection fractions (LVEF /RVEF), stroke \nvolumes (LVSV/RVSV) and cardiac output (LVCO) were calculated. Correlation \nanalysis and group comparisons were performed to ex amine associations \nbetween the participants' demographic and quantitat ive imaging phenotypes. \nResults or Findings: In this study, LVEDV/RVEDV (r=-0.22/r=-0.23), \nLVESV/RVESV (r=-0.17/r=-0.19), LVSV/RVSV, (r=-0.21/ r=-0.20) and LVCO \n(r=-0.19) significantly decreased with age, with ma les showing a higher \nbaseline and steeper decline. LVM and LVEF were nea rly stable \n(r=0.01/r=0.01). LVEF was consistently higher in fe males (p<0.001). RVEF \nincreased with age for females and decreased for ma les (r=0.07, r=-0.02). \nHypertension, diabetes, smoking and high blood lipi ds were associated with \nsignificantly higher LVM and lower LVEF/RVEF (p<0.0 01), while high HDL was \nnegatively correlated with LVM (r=-0.36), LVEDV/RVE DV (r=-0.20/r=-0.22) and \npositively with LVEF/RVEF (r=0.08/r=0.14). \nConclusion: A CMR full-cycle segmentation model was developed a nd \napplied to a large population study spanning a broa d age range. The resulting \nimaging phenotypes revealed strong associations wit h age, disease, and \ndemographic factors. \nLimitations: The analysis was performed automatically using only  basic \nquality control. \nFunding for this study: No specific funding. \nEthics committee - additional information: The study protocol, participant \ninformation, and consent forms for the NAKO study w ere reviewed by all local \nethics committees of the participating institutions . \nAuthor Disclosures:  \nSergios Gatidis: Nothing to disclose \nBjarne Jonas Kerber: Nothing to disclose \nThomas Küstner: Nothing to disclose \n \n \nVariability of coronary artery calcium score in the  multicentre \nDISCHARGE trial: Agreement between readings at the core laboratory \nand the clinical centres \n*F. Biavati*¹, M. Mohamed¹, S. Tsogias¹, B. Föllmer ¹, M. Bosserdt¹, J. Dodd², \nM. Dewey¹; ¹Berlin/DE, ²Dublin/IE \n \nPurpose or Learning Objective: To assess the agreement between core \nlaboratory and the clinical centre measurement of c oronary artery calcium \n(CAC) score using the Agatston method. \nMethods or Background: At each of the 26 clinical centres across 16 \nEuropean countries [NCT02400229], radiologists meas ured the patients’ CAC \nscore based on noncontrast coronary computed tomogr aphy (CT). Two \nreaders measured the CAC score at the core laborato ry blinded to the clinical \ncentre reading. Bland-Altman analysis of the CAC sc ores values and Cohen’s \nkappa of the CAC score risk categories (I: 0, II: 1 -400, III: >400) was \nperformed. \nResults or Findings: 1550 patients (mean age, 59 years ± 10 [SD], 56.3% \nwomen) were included. The Bland-Altman analysis sho wed a mean absolute \ndifference of 2.0 and limits of agreement of ± 93.3 between the core laboratory \nand the clinical centres. There was agreement in th e CAC score risk categories \nin 96.6% (1498 of 1550) of patients between the cor e laboratory and the \nclinical centre reading. Discrepancies between CAC score risk categories \noccurred mostly between categories I and II (88.5%,  46 of 52). Agreement \naccording to Cohen’s kappa was excellent (0.94, 95%  CI: 0.93, 0.96; p<.001). \nMost disagreements in the assignment of risk catego ries were between \ncategory I and II (88.5%). \nConclusion: CAC score measurements had good agreement between t he \ncore laboratory and clinical centres in a Pan-Europ ean multicentre trial \nsuggesting that CAC score measurements can be widel y implemented as part \nof cardiac CT based on its reproducibility. \nLimitations: Our study has limitations. It included only stable chest pain \npatients. The population was from 26 European centr es, limiting global \napplicability. Some CT scans were excluded due to m issing data or non-\nmatching reconstruction parameters, and only filter ed back projection \nreconstruction was used. \nFunding for this study: Funding was provided by grants from the EU-FP7 \nFramework Program (FP 2007-2013, EC-GA 603266). \nEthics committee - additional information: The study was approved by The \nGerman Federal Office for Radiation Protection and the local or national \nauthorities at each trial site. The reference numbe r is: EA1/294/13. \nAuthor Disclosures:  \nJonathan Dodd: Author: Receives royalties as a co-a uthor of book chapters in \nthe Stat-Dx book Series Diagnostic Imaging – Cardio vascular and the textbook \nCT and MRI in Cardiology, Elsevier. Other: Associat e Editor for Radiology, a \nmember of the Editorial Board for Radiology Cardiot horacic Imaging, and an \nAssociate Editor for the Quarterly Journal of Medic ine. All non-paid. Grant \nRecipient: Received funding from EU-FP7 Framework P rogram (DISCHARGE \nEU FP EC-GA 603266). \n \nMahmoud Mohamed: Nothing to disclose \nMarc Dewey: Board Member: M.D. is European Society of Radiology (ESR) \nPublications Chair (2022-2025); the opinions expres sed in this presentation are \nthe author’s own and do not represent the view of E SR. Other: Hands-on \ncardiac CT courses (www.ct-kurs.de) Author: Cardiac  CT (Springer Nature). \nGrant Recipient: EU (EC-GA 603266 in HEALTH.2013.2. 4.2-2) DFG (DE \n1361/14-1, DE 1361/18-1, BIOQIC GRK 2260/1, Radiomi cs DE 1361/19-1 \n[428222922] and 20-1 [428223139] in SPP 2177/1), GU IDE-IT (DE 1361/24-1), \nBerlin University Alliance (GC_SC_PC 27), G-BA (01N VF23002), Berlin \nInstitute of Health (Digital Health Accelerator). R esearch/Grant Support: \nSiemens, General Electric, Philips, Canon. Patent H older: Patent on fractal \nanalysis of perfusion imaging (jointly with Florian  Michallek, EPO 2022 \nEP3350773A1, and USPTO 2021 10,991,109, approved). \nSotirios Tsogias: Nothing to disclose \nFederico Biavati: Nothing to disclose \nMaria Bosserdt: Nothing to disclose \nBernhard Föllmer: Nothing to disclose \n \n \nPatient Acceptance of Coronary CT Angiography Versu s Invasive \nCoronary Angiography in Patients with Stable Chest Pain \nM. Bosserdt, *K. Schulze*, M. Mohamed, A-M. Stantie n, M. Dewey,  \nE. Schöneberger; Berlin/DE \n \nPurpose or Learning Objective: Patient preference between coronary \ncomputed tomography (CT) and invasive coronary angi ography (ICA) in a \nmulticentre analysis in Europe is unknown. Therefor e, we compare patient \npreference for CT and ICA in a European multicentre  randomised controlled \ntrial. \nMethods or Background: A total of 3561 patients with a clinical indication  for \nICA with stable chest pain and an intermediate like lihood of obstructive \ncoronary artery disease were analysed in this presp ecified secondary analysis \nfrom the randomised DISCHARGE trial (NCT02400229) c onducted between \nOctober 2015 and April 2019 in 26 European centres.  Patient preference using \na previously validated questionnaire completed at l east 24 hours after CT or \nICA, including preparation for the tests, anxiety, comfort, level of helplessness, \npain, willingness to undergo the tests again, overa ll satisfaction and \npreference. \nResults or Findings: The questionnaire was completed by 89.7% in the CT \ngroup (1622/1808) and 89.4% (1567/1753, P=.75) in t he ICA group. Patients \nreported significantly higher satisfaction with CT (mean (SD): CT: 1.37 (0.53) \nvs. 1.48 (0.60); score: 1-5; P<.0001), were more wi lling to undergo CT again \n(mean (SD): CT: 1404/1622 (86.6%) vs. ICA: 1127/156 7 (71.9%); P<.0001), \nand were better prepared to CT (mean (SD): CT: 1.40  (0.56) vs. 1.49 (0.64); \nscore: 1-5; P=.0005). They were less anxious before  CT (mean (SD): CT: 1.07 \n(1.17) vs. ICA: 1.62 (1.33); score: 1-4; P<.0001), felt less helpless during CT \n(mean (SD): CT: 0.54 (0.72) vs. ICA: 0.91 (0.90); s core: 0-4; P<.0001), and felt \nmore comfortable (mean(SD): CT: 1.64 [0.66] vs. ICA : 1.90 [0.80]; score: 1-5; \nP<.0001). \nConclusion: In this multicentre, randomised trial of patients r eferred for ICA \nwith stable chest pain and an intermediate likeliho od of obstructive coronary \nartery disease, patient preference was in favour of  coronary CT angiography. \nLimitations: Not applicable. \nFunding for this study: This study was funded by grants from the EU-FP7 \nFramework Program (FP 2007-2013, EC-GA 603266). \nEthics committee - additional information: The study was approved by \nethics committee at Charité (EA1/294/13). \nAuthor Disclosures:  \nKenrick Schulze: Nothing to disclose \nMahmoud Mohamed: Nothing to disclose \nEva Schöneberger: Nothing to disclose \nAnne-Marieke Stantien: Nothing to disclose \nMarc Dewey: Board Member: M.D. is European Society of Radiology (ESR) \nPublications Chair (2022-2025); the opinions expres sed in this presentation are \nthe author’s own and do not represent the view of E SR. Grant Recipient: EU \n(EC-GA 603266 in HEALTH.2013.2.4.2-2) DFG (DE 1361/ 14-1, DE 1361/18-1, \nBIOQIC GRK 2260/1, Radiomics DE 1361/19-1 [42822292 2] and 20-1 \n[428223139] in SPP 2177/1), GUIDE-IT (DE 1361/24-1) , Berlin University \nAlliance (GC_SC_PC 27), G-BA (01NVF23002), Berlin I nstitute of Health \n(Digital Health Accelerator). Author: Cardiac CT (S pringer Nature) Patent \nHolder: Patent on fractal analysis of perfusion ima ging (jointly with Florian \nMichallek, EPO 2022 EP3350773A1, and USPTO 2021 10, 991,109, approved) \nResearch/Grant Support: Siemens, General Electric, Philips, Canon. Other: \nHands-on cardiac CT courses (www.ct-kurs.de) Instit utional research \nagreements: Siemens, General Electric, Philips, Can on. Patent on fractal \nanalysis of perfusion imaging (jointly with Florian  Michallek, EPO 2022 \nEP3350773A1, and USPTO 2021 10,991,109, approved) \nMaria Bosserdt: Nothing to disclose \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 289  \nCAD-Man EXTEND: Long-term clinical results of a sin gle centre \nrandomised controlled trial comparing CT with ICA \n*A-M. Stantien*¹, F. Biavati¹, A-C. Stahl¹, S. Chim ed¹, M. Mohamed¹,  \nL. M. Serna Higuita², M. Bosserdt¹, M. Dewey¹; ¹Ber lin/DE, ²Tübingen/DE \n \nPurpose or Learning Objective: To investigate major adverse cardiovascular \nevents (MACE) after 10 years of follow-up in patien ts with an intermediate \nprobability of coronary artery disease (CAD) underg oing computed tomography \n(CT) or invasive coronary angiography (ICA). \nMethods or Background: This is the 10-year long-term clinical follow-up of  \nthe single centre randomised CAD-Man (Coronary Arte ry Disease \nManagement) trial comparing CT with ICA in patients  with atypical angina or \nchest pain clinically referred for ICA. Clinical fo llow-up was done by \ninterviewing patients using questionnaires asking a bout MACE (myocardial \ninfarction, stroke, unstable angina, (surgical) re- /revascularization, cardiac \ndeath) in the past 10 years. Additionally, every pa tient was offered a cardiac \nCT scan and blood samples were taken. The associati on between \nrandomisation group and MACE was assessed using a m ultivariate Cox \nproportional hazards model. \nResults or Findings: Out of 329 patients included in CAD-Man 106 patient s \ncompleted the clinical long-term follow-up resultin g in a total median follow-up \nof 4.9 years. Additional 18 MACE cases occurred, 7/ 49 (14%) of which in the \nCT group and 11/57 (19%) in the ICA group. There wa s no statistically \nsignificant difference in MACE between the two rand omisation arms. The HR \nwas 0.86 (95% CI 0.42–1.74) in the CT group. \nConclusion: After 10 years of follow-up, the survival and the o ccurrence of \nMACE was similar in the CT group compared to the IC A group for patients \nreferred for ICA because of atypical angina or stab le chest pain and an \nintermediate pretest probability of CAD. \nLimitations: The limitations of the study are the single centre design and the \nlow number of MACE, limiting the generalisability o f our findings. \nFunding for this study: This study was funded by a grant of the Heisenberg \nprogramme. \nEthics committee - additional information: The study was approved by \nethics committee at Charité (EA1/124/23). \nAuthor Disclosures:  \nMahmoud Mohamed: Nothing to disclose \nAnne-Marieke Stantien: Nothing to disclose \nLina Maria Serna Higuita: Nothing to disclose \nSurenjav Chimed: Nothing to disclose \nMarc Dewey: Research/Grant Support: Institutional r esearch agreements with \nSiemens, General Electric, Philips, Canon. Grant Re cipient: EU (EC-GA \n603266 in HEALTH.2013.2.4.2-2) DFG (DE 1361/14-1, D E 1361/18-1, BIOQIC \nGRK 2260/1, Radiomics DE 1361/19-1 [428222922] and 20-1 [428223139] in \nSPP 2177/1), GUIDE-IT (DE 1361/24-1), Berlin Univer sity Alliance \n(GC_SC_PC 27), G-BA (01NVF23002), Berlin Institute of Health (Digital Health \nAccelerator). Patent Holder: Patent on fractal anal ysis of perfusion imaging \n(jointly with Florian Michallek, EPO 2022 EP3350773 A1, and USPTO 2021 \n10,991,109, approved) Author: Cardiac CT (Springer Nature). Other: Hands-on \ncardiac CT-couses (www.ct-kurs.de) Board Member: Eu ropean Society of \nRadiology (ESR) Publications Chair (2022-2025) \nAnn-Christine Stahl: Nothing to disclose \nFederico Biavati: Nothing to disclose \nMaria Bosserdt: Nothing to disclose \n \n \nImproved Prediction of Obstructive Coronary Artery Disease by \nDISCHARGE Trial Pretest Calculator Combined with Ca rdiac CT \n*M. Mohamed*¹, M. Dewey¹, V. Wieske¹, P. Schlattman n², R. Haase¹, J. Dodd³; \n¹Berlin/DE, ²Jena/DE, ³Dublin/IE \n \nPurpose or Learning Objective: To evaluate the accuracy of pretest \nprobability (PTP) calculations alone and in combina tion with computed \ntomography angiography (CTA) results for the diagno sis of CAD in stable chest \npain. \nMethods or Background: Individual patient data (IPD) meta-analysis of 65 \nprospective diagnostic accuracy studies of patients  clinically referred to \ninvasive coronary angiography (ICA) with stable che st pain in 22 countries. \nThree clinical probability models a PTP model based  on age, sex and chest \npain type (termed the updated DISCHARGE trial PTP c alculator), a CTA alone \nmodel and the updated DISCHARGE Trial PTP calculato r and CTA model \ncombined were constructed. The models were built by  multivariable logistic \nregressions with a dataset-specific random intercep t and were compared using \nthe area under the receiver-operating-characteristi c curve (AUC) and the \ndecision curve analysis (DCA). \nResults or Findings: 5332 stable chest pain patients with clinically ind icated \nICA were included. The updated DISCHARGE pretest pr obability calculator \nwas more accurate than the original Diamond-Forrest er model (AUC: 0.68, \n95% CI: 0.66-0.69 versus 0.63, 95% CI: 0.62-0.65). The combination of \nDISCHARGE pretest probability calculator with CTA f indings significantly \nimproved accuracy compared with either model alone (AUC: 0.86, 95% CI: \n0.85-0.87 versus 0.81, 95% CI: 0.80-0.82). The impr oved prediction of CAD by \ncombining CTA with the updated DISCHARGE Trial PTP calculator prediction \nmodel was consistent in the DCA with an increased n et benefit for all chest \npain types and was almost equally seen in patients with typical or atypical \nangina (0.85, 95% CI: 0.84-0.86) and nonanginal or other chest discomfort \n(0.88, 95% CI: 0.86-0.89). \nConclusion: Combining the updated DISCHARGE trial PTP calculato r with \nCTA provides more accurate prediction than CTA alon e for the diagnosis of \nobstructive CAD. \nLimitations: The ICA indication resulted in a relatively high CA D prevalence of \n48.3%. \nFunding for this study: The COME-CCT Consortium is funded by a joint \nprogram of the German Research Foundation and the G erman Federal \nMinistry of Education and Research (01KG1110) and t he Digital Health \nAccelerator of the Berlin Institute of Health to Ma rc Dewey. All researchers are \nindependent of the funding bodies. The funding bodi es had no role in the study \ndesign; in the collection, analysis and interpretat ion of data; in the writing of the \nreport; and in the decision to submit the manuscrip t for publication. \nEthics committee - additional information: This study was approved by the \nlocal research ethics committee of Charité (EA-1-08 0-08) and the German \nFederal Office for Radiation Protection (Z5-22462/2 -2008-048). All patients \ngave written informed consent before randomisation.  \nAuthor Disclosures:  \nJonathan Dodd: Nothing to disclose \nMahmoud Mohamed: Nothing to disclose \nPeter Schlattmann: Nothing to disclose \nViktoria Wieske: Nothing to disclose \nMarc Dewey: Nothing to disclose  \nRobert Haase: Nothing to disclose \n \n \nPericoronary adipose tissue volume but not attenuat ion is associated \nwith quantitative coronary plaque metrics on corona ry CT angiography: \nInsights from the PROMISE trial \n*M. C. Langenbach*¹, I. Hadzic¹, T. Mayrhofer¹, J. Karady¹, S. Shah², M. T. Lu¹, \nM. Ferencik¹, P. Douglas², B. Foldyna¹; ¹Boston, MA /US, ²Durham/US \n(marcel.langenbach@me.com) \n \nPurpose or Learning Objective: Pericoronary adipose tissue (PCAT) is \nrelated to pericoronary inflammation and contribute s to atherogenesis and \nadverse outcomes. We investigated the association b etween PCAT and \nadvanced plaque characteristics in patients with st able chest pain. \nMethods or Background: PCAT was quantified around all three major \nepicardial vessels on non-contrast CT images from t he PROMISE trial using a \nvalidated deep-learning algorithm. Quantitative cor onary plaque metrics \nincluded total plaque volume and burden (TPV, mm³; TPB, %), and plaque \ncomposition (calcified plaque (CP), non-calcified p laque (NCP), and low-\ndensity NCP (LD-NCP; <30HU). Multivariable linear r egression analyses \nrelated global PCAT density (per 10HU) and BSA-inde xed PCAT volume (per \n10 cm³/m²) to plaque metrics (per 10mm³/1%), adjust ed for signal-to-noise \nratio, tube voltage, risk score, and BMI. \nResults or Findings: In 3,620 participants (age: 60±8 years; women: \n1,945(51.2%), mean total heart PCAT volume and dens ity were \n12.9±3.1cm³/m² and -81.4±6.4HU. Greater PCAT volume related to higher CP \nvolume (Coef. 2.24, 95%CI:0.40–4.08, p=0.017), LD-N CP volume (Coef. 0.28, \n95%CI:0.05–0.52, p=0.018), and TPV (Coef. 3.84, 95% CI:-0.21–7.88, p=0.06). \nA 10 cm³/m² increase in PCAT volume was associated with a 7% higher TPB \n(Coef. 6.93, 95%CI:3.58–10.28, p<0.001), 4% higher CP burden (Coef. 3.97, \n95%CI:2.42–5.51, p<0.001), 3% higher NCP burden (Co ef. 2.96, 95%CI:0.33–\n5.59, p=0.027), and 1% higher LD-NCP burden (Coef. 0.68, 95%CI:0.34–1.01, \np<0.001). PCAT density showed no significant associ ation with plaque volume \nand burden including composition. \nConclusion: PCAT volume is associated with plaque volume, burde n, and \ncomposition suggesting a relationship with both ath erogenesis and plaque \narchitecture. In contrast, PCAT density, a known me asure of inflammation, was \nnot associated with either. These findings emphasiz e the complex physiology \nof pericoronary fat and underscore the need to furt her investigate PCAT’s \npotential role as a target for treatment interventi ons. \nLimitations: Secondary analysis \nFunding for this study: DFG (project number: 502109212) \nEthics committee - additional information: Massachusetts General Hospital \n(2009P002231) \nAuthor Disclosures:  \nIbrahim Hadzic: Nothing to disclose \nPamela Douglas: Nothing to disclose \nThomas Mayrhofer: Nothing to disclose \nBorek Foldyna: Nothing to disclose \nMaros Ferencik: Nothing to disclose \nMarcel Christian Langenbach: Grant Recipient: DFG r esearch grant \nJulia Karady: Nothing to disclose \nMichael T. Lu: Nothing to disclose \nSvati Shah: Nothing to disclose \n \n\n \n \nSunday \nAbstract-based Programme \n \n 290  \nEvaluating Radiation Exposure in Patients with Stab le chest Pain in the \nDISCHARGE trial \nM. Bosserdt¹, M. Mohamed¹, M. C. Williams², M. Dewe y¹, *J. Knape*¹; \n¹Berlin/DE, ²Edinburgh/UK \n \nPurpose or Learning Objective: To assess 3.5 years of cumulative radiation \ndoses of cardiovascular imaging to participants und ergoing computed \ntomography (CT) or invasive coronary angiography (I CA) for suspected \ncoronary artery disease. \nMethods or Background: This is a prespecified analysis of a multicentre, \nrandomised DISCHARGE trial involving 3561 participa nts with stable chest and \nwho were referred for ICA, conducted between Octobe r 2015 and April 2019 in \n26 European centres. Participants were randomised t o either CT (1808) or ICA \n(1753). Radiation dose from CT (dose-length product ), SPECT (injected \nactivity), PET-CT (injected activity) and ICA (kern -area product) was assessed \nfor 3.5 years after randomisation. Effective dose w as calculated using \nconversion factors appropriate for the imaging moda lity. Missing data were \nimputed using the mean. Wilcoxon rank sum test was used to assess group \ndifferences. \nResults or Findings: Over a median follow-up period of 3.5 years, a tota l of \n1845 (CT: 1796 vs. ICA: 49) CT scans, 1939 (CT: 344  vs. ICA: 1595) ICA \nwithout PCI, 584 (CT: 269 vs. ICA: 315) ICA with PC I, 75 (CT: 46 vs. ICA: 29) \nSPECT and 81 (CT: 66 vs. ICA: 15) PET-CT were perfo rmed. Total per-\nparticipant cumulative dose was higher in the CT gr oup (median, 6.1 mSv; \nIQR, 3.9-10.3 mSv) compared with ICA group (median,  4.4 mSv; IQR, 2.2-9.1 \nmSv, P<0.001). The cumulative dose varied across pa rticipating centre, with a \nmean 9-fold for CT-group and 4-fold for ICA-group v ariation between the \nhighest and lowest dose. \nConclusion: The cumulative radiation dose over 3.5 years was hi gher in the \nCT group compared to the ICA group. However, the to tal radiation dose for the \nCT and ICA group differed greatly among the partici pating centres. \nLimitations: At a follow-up radiation dose was not available for  all procedures \nFunding for this study: This study was funded by grants from the EU-FP7 \nFramework Program (FP 2007-2013, EC-GA 603266) \nEthics committee - additional information: The study was approved by \nethics committee at Charité (EA1/294/13)). \nAuthor Disclosures:  \nMahmoud Mohamed: Nothing to disclose \nMarc Dewey: Grant Recipient: EU (EC-GA 603266 in HE ALTH.2013.2.4.2-2) \nDFG (DE 1361/14-1, DE 1361/18-1, BIOQIC GRK 2260/1,  Radiomics DE \n1361/19-1 [428222922] and 20-1 [428223139] in SPP 2 177/1), GUIDE-IT (DE \n1361/24-1), Berlin University Alliance (GC_SC_PC 27 ), G-BA (01NVF23002), \nBerlin Institute of Health (Digital Health Accelera tor). Other: Editor: Cardiac CT \n(Springer Nature). Hands-on cardiac CT courses (www .ct-kurs.de) Institutional \nresearch agreements: Siemens, General Electric, Phi lips, Canon. Patent on \nfractal analysis of perfusion imaging (jointly with  Florian Michallek, EPO 2022 \nEP3350773A1, and USPTO 2021 10,991,109, approved) M .D. is European \nSociety of Radiology (ESR) Publications Chair (2022 -2025); the opinions \nexpressed in this presentation are the author’s own  and do not represent the \nview of ESR. \nMichelle Claire Williams: Nothing to disclose \nMaria Bosserdt: Nothing to disclose \nJakob Knape: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n11:30-12:30 Research Stage 3 \nResearch Presentation Session: \nInterventional Radiology \nRPS 2409 \nHepato-biliary interventions in benign \nconditions \n \nModerator \nG. Maleux; Leuven/BE  \n(Geert.Maleux@uzleuven.be) \n \n \nMulticenter outcomes analysis of self-expandable bi odegradable stents \nfor the management of benign biliary strictures in 81 patients with \npediatric liver transplantation \n*P. Marra*¹, D. Barnés Navarro², L. F. Fernández Ro dríguez³, G. Barbiero⁴,  \nS. Mcguirk⁵, C. Gonzalez-Junyent², T. Hernández Cabrero³, M. C . Minà⁴,  \nS. Sironi¹; ¹Bergamo/IT, ²Barcelona/ES, ³Madrid/ES,  ⁴Padova/IT, \n⁵Birmingham/UK \n(pmarra@asst-pg23.it) \n \nPurpose or Learning Objective: Percutaneous transhepatic cholangiography \n(PTC), billioplasty and biliary drainage are routin e treatments for benign \nstrictures after pediatric liver transplantation (p LT). This multicentric study \nevaluated the efficacy and safety of biodegradable biliary stents. \nMethods or Background: ELLA (Ella-CS Ltd) is self-expandable stent made \nof polydioxanone which degrades in 3-6 months. We e valuated a total of 81 \npatients (39 females, median age 4 y/o; 42 males, m edian age 5 y/o) with \nbenign biliary strictures developed after pLT in fi ve European centers. All the \npatients underwent percutaneous bilioplasty followe d by ELLA stent placement \nbetween October 2014 and March 2024. Stricture feat ures and treatment \ntiming were assessed. Efficacy in terms of freedom from stricture recurrence \nand safety in terms of complications were analyzed.  \nResults or Findings: Regarding stricture features, 42.7% of strictures h ad and \nextension <1 cm, 57.3% ≥1 cm; 69.7% of patients had strictures at the \nhepaticojejunostomy, 2.5% had intrahepatic strictur es and 28% had both \nintrahepatic and anastomotic strictures. The time b etween PTC and stent \nplacement varied from 0 to 744 days (median of 36 d ays, IQR 61 days) with \nmaintainance of a drainage; 53/81 (65.4%) patients underwent additional \nbilioplasty sessions before ELLA placement. Success ful stent placement was \nachieved in 100% of cases and complications occurre d in 23.5% of cases, all \nminor, mostly infectious cholangitis. Eighteen pati ents had a stricture \nrecurrence (22.2%) during a median follow-up of 784  days (IQR 1200 days); of \nthese, 55.6% were retreated with ELLA. \nConclusion: Biodegradable self-expandable biliary stent are be safe and \neffective for the treatment of benign biliary strictures after pLT. Further studies \ninvestigating factor predisposing to stent failure with standardized protocols are \nrequired to define the ideal candidates and the bes t timing for stenting. \nLimitations: Retrospective design; variable protocols among cent ers \nFunding for this study: None \nEthics committee - additional information: Comitato Etico of Bergamo - \nMulticenter ELLA \nAuthor Disclosures:  \nSandro Sironi: Nothing to disclose \nTeresa Hernández Cabrero: Nothing to disclose \nGiulio Barbiero: Nothing to disclose \nPaolo Marra: Nothing to disclose \nSimon Mcguirk: Nothing to disclose \nLucia Fernández Fernández Rodríguez: Nothing to dis close \nCarla Gonzalez-Junyent: Nothing to disclose \nMaria Carla Minà: Nothing to disclose \nDaniel Barnés Navarro: Nothing to disclose \n \n \n \n \n \n \n \n \n \n \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 291  \nPercutaneous portal vein recanalization of non-cirr hotic extrahepatic \nportal vein obstruction: technical considerations a nd clinical outcomes in \n15 children \n*P. Marra*, R. Muglia, F. S. Carbone, L. Dulcetta, L. D'Antiga, S. Sironi; \nBergamo/IT \n(pmarra@asst-pg23.it) \n \nPurpose or Learning Objective: Portal hypertension resulting from \nextrahepatic portal vein obstruction (EHPVO) in chi ldren has been managed \nprimarily through Meso-Rex bypass. The aim of the s tudy is to report a \npreliminary series of patients who underwent attemp ts at portal vein \nrecanalization (PVR) prior to other types of interv ention. \nMethods or Background: A cohort of consecutive patients with EHPVO from \n2021-2024 was retrospectively collected. After tran sjugular wedge hepatic \nvenography for the study of the native intrahepatic  portal system, percutaneous \nPVR was attempted via transhepatic and/or transplen ic access. Clinical and \nprocedural data, technical and clinical success, co mplications and follow up \ndata were recorded. Technical success was considere d at least the partial \nrevascularization of the native portal system. \nResults or Findings: Fifteen patients (7 males; median age 8 years) with  \nsevere portal hypertension due to EHPVO underwent 2 0 percutaneous \ntranshepatic (n=1), transplenic (n=12) or simultane ous transhepatic/transplenic \n(n=7) attempts at portal vein recanalization. Rex v ein was judged patent ad \nwedge hepatic venography in 2/15 (13%). Successful recanalization was \nachieved in 9/15 patients (60%). No major adverse e vents were observed. \nAfter successful angioplasty, 8/9 patients required  metal stenting to obtain \nsustained patency. None of the failed patients was considered eligible for \nMeso-Rex bypass and underwent TIPS (n=2), splenecto my (n=1), surgical \nshunt (n=1). Two patients were followed-up without further interventions. After \na median follow-up of 6 months patency of the main portal vein was \ndemonstrated in all the patients who achieved PVR, with clinical and laboratory \nimprovement of portal hypertension. \nConclusion: Our preliminary experience suggests that 60% of chi ldren with \nEHPVO can restore the portal flow by endovascular t reatment, even with \nobliterated Rex vein. Thanks to its low invasivenes s, PVR may be regarded as \nthe primary intervention, before considering surger y. \nLimitations: Retrospective study with limited sample \nFunding for this study: None \nEthics committee - additional information: EC of Bergamo - Portal-01 \nAuthor Disclosures:  \nSandro Sironi: Nothing to disclose \nPaolo Marra: Nothing to disclose \nFrancesco Saverio Carbone: Nothing to disclose \nRiccardo Muglia: Nothing to disclose \nLorenzo D'Antiga: Nothing to disclose \nLudovico Dulcetta: Nothing to disclose \n \n \nComparison of different techniques for transjugular  intrahepatic \nportosystemic shunt creation in a retrospective ser ies of 51 paediatric \npatients \n*A. Princi*, P. Marra, F. S. Carbone, L. Dulcetta, R. Muglia, S. Sironi; \nBergamo/IT \n(angeloprinci1@gmail.com) \n \nPurpose or Learning Objective: TIPS creation with the standard technique \nmay be challenging in children with low body weight , unusual anatomy, and \nliver grafts. This study analyses different approac hes for TIPS creation in a \nretrospective cohort of paediatric patients. \nMethods or Background: The retrospective single-centre cohort included 47 \npatients who received TIPS either with the standard  (sTIPS; n:30 median age \n10 years IQR8-16) or with a transhepatic/transpleni c ‘’hybrid’’ (hTIPS; hTIPS \nn:17 13y IQR7-16) approaches from 2005 to 2023. Inc lusion criteria were age \n≤ 18 years or liver graft transplanted in paediatric  age. All the variables \nbetween the sTIPS and the hTIPS group were compared . \nResults or Findings: Technical, hemodynamic, and clinical success were \n100%, 100%, and 83% in sTIPS (8 permanent, 22 as br idge) while 100%, \n97%, and 81% in hTIPS (10 permanent, three bridge).  Patients with liver grafts \nwere 4 (13%) in the sTIPS and 5 in the hTIPS (30%) groups, respectively. \nPortal cavernoma, acute portal vein thrombosis, and  Budd-Chiari syndrome \nwere significantly more prevalent in the hTIPS grou p (76% vs 33%, p<0.05). \nIndications were comparable. Covered stents were re spectively employed in \n90% sTIPS and 88% hTIPS, with primary-assisted pate ncy of 100% and 93% \nafter a median follow-up of 34 and 14 months (IQR 2 1-52 and 12-20). \nComplications included one hemoperitoneum in each g roup conservatively \nmanaged, one liver failure (hTIPS group) requiring urgent transplantation and \none septic shock (sTIPS group). Eight shunt dysfunc tions in sTIPS and 3 in \nhTIPS were all successfully treated. \nConclusion: In a retrospective cohort of children, hybrid techn iques for TIPS \ncreation achieved high success rates comparable to the standard technique \ndespite unfavourable baseline characteristics. Furt her studies may investigate \nthe role of these approaches in expanding TIPS indi cation for complex \nscenarios in paediatric patients. \nLimitations: Retrospective nature. \nFunding for this study: None. \nEthics committee - additional information: Not applicable. \nAuthor Disclosures:  \nSandro Sironi: Nothing to disclose \nPaolo Marra: Nothing to disclose \nFrancesco Saverio Carbone: Nothing to disclose \nAngelo Princi: Nothing to disclose \nRiccardo Muglia: Nothing to disclose \nLudovico Dulcetta: Nothing to disclose \n \n \nTransjugular Intrahepatic Portosystemic Shunt (TIPS ): early laboratory \nchanges and correlations with short-term mortality \n*F. Schön*, T. Helmberger, M. Berning, S. F. U. Blu m, C. Radosa, S. Löck,  \nR-T. Hoffmann, J-P. Kühn; Dresden/DE \n \nPurpose or Learning Objective: To investigate early laboratory changes \nfollowing Transjugular Intrahepatic Portosystemic S hunt (TIPS) and their \nassociations with short-term mortality. \nMethods or Background: TIPS procedures from 2017 to 2023 were enrolled \nretrospectively. Laboratory parameters (INR, ALAT, ASAT, GGT and bilirubin) \nwere assessed once pre-procedurally, on post-proced ural days 1 or 2, and \nonce again between days 5 and 7. Percentage changes  from baseline were \ncalculated for each parameter. Temporal changes of the parameters were \nassessed using Kruskal-Wallis tests, and comparison s regarding 30-day \nmortality were evaluated using Mann-Whitney U tests . \nResults or Findings: A total of 245 TIPS procedures (161 men, mean age \n59.8 +/- 10.9 years) were enrolled, with a technica l success rate of 95.5% \n(234/245). All laboratory parameters significantly increased post-procedurally \n(p < 0.001). ALAT and ASAT revealed the highest inc rease within the first two \npost-procedural days (+374 +/- 1118%, and +450 +/- 1079%, respectively), \nfollowed by decreases on days 5-7 (+279 +/- 568%, a nd +125 +/- 233%, \nrespectively). Patients who died within 30 days (n = 17) had significantly higher \nbaseline INR (p = 0.009), ASAT (p = 0.014) and bili rubin (p = 0.011), while \nGGT was lower (p = 0.012). 30-day mortality was ass ociated with a higher \nincrease of ASAT and GGT on days 1/2 (+1361 +/- 220 2% vs. +415 +/- 954%, \np = 0.029; and +56 +/- 101% vs. +21 +/- 51%, p = 0. 034, respectively). \nConclusion: TIPS significantly impacts liver function, with mar ked early \nincreases of ALAT and ASAT levels. Pronounced incre ase of laboratory \nparameters within the first two post-procedural day s might help to identify high-\nrisk patients in terms of short-term mortality. \nLimitations: Retrospective study design with a relatively small number of \npatients. \nFunding for this study: Not applicable. \nEthics committee - additional information: The present study was approved \nby the local ethics committee (BO-EK-501122023). \nAuthor Disclosures:  \nRalf-Thorsten Hoffmann: Nothing to disclose \nMarco Berning: Nothing to disclose \nJens-Peter Kühn: Nothing to disclose \nFelix Schön: Nothing to disclose \nChristoph Radosa: Nothing to disclose \nSophia Freya Ulrike Blum: Nothing to disclose \nThomas Helmberger: Nothing to disclose \nSteffen Löck: Nothing to disclose \n \n \nSuccessful approach to giant hydatid cysts of liver  \n*U. Koç*, C. Aydın, M. Özdemir; Ankara/TR \n \nPurpose or Learning Objective: We aimed to evaluate cases of giant liver \nhydatid cysts. \nMethods or Background: Between December 2020 and January 2023, out of \n100 liver hydatid cysts treated with the percutaneo us approach in our \ninterventional radiology department, 35 cases were more than 10 cm in one of \nthe diameters. These giant hydatid cysts were treat ed with the catheterization \napproach which includes trochar style puncture, asp iration of the cavity, and \ninstallation with hypertonic saline solution. Then,  ethanol installation of the cyst \nis the next step if the cavity is not connected wit h the biliary tree. \nResults or Findings: Out of 35 patients, 27 (78%) patients were treated \nsuccessfully with catheter approach solely without the need of ERCP related \ninterventions. 8 patients (22%) needed further inte rventions with ERCP \nbecause of cystobiliary fistula; 2 out of 8 patient s had biliary passage \nconfirmed after contrast installation at catheteriz ation. Other 6 patients did not \nhave opaque drainage into the biliary system, rathe r they were suspected of \nhaving a fistula because of yellow content or bioch emistry showing high \nbilirubin. These 8 patients had extended days of ex ternal catheter duration, \nabout 20 days on average; while other patients had the external catheter \nremoved approximately 2 days later. 1 patient had t o go through cystectomy \n\n \n \nSunday \nAbstract-based Programme \n \n 292  \noperation because of insufficient drainage after re peated ERCP interventions, \nand this patient had the longest hospitalization du ration of 100 days; whilst \nother ERCP patients had 20 days of hospitalization on average. \nConclusion: Giant hydatid cysts are manageable with a percutane ous \napproach. Cystobiliary fistula must be kept in mind  especially for the giant \nhydatid cysts, since size is an important predictor  for this communication. \nLimitations: As a limitation, this study did not compare the met hod with others, \nbut a future study will address this. \nFunding for this study: None \nEthics committee - additional information: None \nAuthor Disclosures:  \nMustafa Özdemir: Nothing to disclose \nUral Koç: Nothing to disclose \nCeren Aydın: Nothing to disclose \n \n \nIncidence of Bleeding Between Percutaneous vs. Endo scopic Biliary \nDiversion in Patients with Biliary Tract Obstructio n: A Systematic Review \nand Meta-Analysis \n*E. D. L. A. Salazar Perez*, E. E. Lozada Hernandez ,  \nB. E. E. Retamoza Rojas; Leon/MX \n(estrella.asp.16.02@gmail.com) \n \nPurpose or Learning Objective: This meta-analysis examines the incidence \nof bleeding in patients undergoing biliary diversio n via two alternatives to \nERCP: percutaneous and endoscopic methods. The anal ysis aims to establish \na foundation for identifying the most suitable trea tment by evaluating \neffectiveness and safety. The primary goal of manag ing biliary tract obstruction \nis to achieve safe and effective drainage, with pro per patient selection and \ndiagnosis being essential for determining the best procedure. \nMethods or Background: Eleven studies were analyzed regarding the \nincidence of bleeding, pancreatitis, and reinterven tions associated with \nendoscopic or percutaneous biliary drainage procedu res from January 2010 to \n2023. The evaluated aspects included sensitivity, r isk of bias ratio, odds ratio, \nand their 95% confidence interval using a random-ef fects model, with effects \nconsidered statistically significant if the confide nce interval was at 95%. The I² \nstatistic was calculated to assess heterogeneity. B ias analysis was reported \nusing funnel plot tables. Data processing was perfo rmed using the R \nprogramming language within the RStudio environment  (version 4.1.0 CRAN). \nAny p-value less than 0.05 was considered statistic ally significant. \nResults or Findings: The study evaluated complications and reinterventio ns \nfor two techniques of biliary diversion: the percut aneous method, involving \n2,058 patients from 11 studies, and the endoscopic method, with 7,959 \npatients. The overall odds ratio for bleeding was 1 .81 (95% CI 0.43-7.60), with \nheterogeneity (I² = 74%). For acute pancreatitis, t he overall odds ratio was 0.15 \n(95% CI 0.05-0.47) with p = 0.03. Reinterventions s howed an odds ratio of \n0.25, with a wide confidence interval [0.06; 1.51],  reflecting high heterogeneity \nand variability among the studies. \nConclusion: The percutaneous technique, as a first-line option compared to \nmany other emerging techniques, remains the ideal c hoice in many referral \ncenters for diseases presenting with biliary tract obstruction, showing minimal \ncomplication rates. \nLimitations: Heterogeneity. \nFunding for this study: There is no funding or conflict of interest. \nEthics committee - additional information: The registration for the approval \nof the hospital was the CEI-004-2022 \nAuthor Disclosures:  \nBeatriz Elena Elena Retamoza Rojas: Nothing to disc lose \nEstrella De Los Angeles Salazar Perez: Nothing to d isclose  \nEdgard Efren Lozada Hernandez: Nothing to disclose \n \n \nClinical Outcomes of Separate versus Single Tract T echniques in \nPercutaneous Radiologic Gastrostomy with Single Gas tropexy:  \nA Multi-Center Retrospective Analysis \n*H. N. Lee*¹, S-J. Park², Y. Cho³, S. Lee¹; ¹Cheona n/KR, ²Ansan/KR, \n³Gangneung/KR \n(kenshin_007_@naver.com) \n \nPurpose or Learning Objective: To compare the clinical outcomes of \nseparate versus single tract techniques and to inve stigate predictors of \ncomplications during percutaneous radiologic gastro stomy with single \ngastropexy. \nMethods or Background: Between January 2018 and January 2024, 241 \nconsecutive patients (mean age: 68.8 ± 13.5 years; male: 73.4%) who \nunderwent percutaneous radiologic gastrostomy with single gastropexy were \nenrolled. The patients were divided into two groups  based on the anchoring \nmethod: the separate tract group (n = 136) and the single tract group (n = 105). \nResults or Findings: The technical success rate was 99.3% in the separat e \ntract group and 98.1% in the single tract group (p = 0.582). Four patients \n(3.81%) in the single tract group experienced intra -procedural anchor \ndislodgment. In 3 of these cases, technical success  was achieved by deploying \na second anchor. The 30-day complication rate was 1 2.5% in the separate \ntract group and 15.2% in the single tract group (p = 0.324). There was no \nprocedure-related mortality. BMI (odds ratio 0.885,  p = 0.021) was a negative \npredictor, while unfavorable anatomy on CT (odds ra tio 2.4, p = 0.033) was a \npositive predictor for complication. \nConclusion: Although anchor dislodgment was a complication uniq ue to the \nsingle tract technique, the two groups showed no si gnificant difference in \noverall clinical outcomes. This study highlights th at BMI and CT findings are \nmore crucial predictors of complications than the c hoice of technique. \nLimitations: The retrospective nature of the study leads to seve ral biases. \nFunding for this study: This study was not supported by any funding. \nEthics committee - additional information: The Institutional Review Board of \ntertiary care hospitals approved this retrospective  study and waived written \ninformed consent for using clinical and imaging dat a. \nAuthor Disclosures:  \nSung-Joon Park: Nothing to disclose \nYoungjong Cho: Nothing to disclose \nHyoung Nam Lee: Nothing to disclose \nSangjoon Lee: Nothing to disclose \n \n \n11:30-12:30 Research Stage 4 \nResearch Presentation Session: \nMusculoskeletal \nRPS 2410 \nImaging of musculoskeletal tumours \n \nModerator \nI.-M. Noebauer-Huhmann; Vienna/AT  \n(iris.noebauer@meduniwien.ac.at) \n \n \nDual-energy computed tomography parameters for the differentiation of \nvertebral small osteolytic metastases (SOMs) and SO M-mimics \n*J. Li*¹, J. Liu²; ¹Fujian/CN, ²Xiamen/CN \n(1508883851@qq.com) \n \nPurpose or Learning Objective: To evaluate the value of dual energy \ncomputed tomography (DECT) quantitative parameters for the differentiation of \nsmall osteolytic metastases (SOMs) and SOM-mimics s uch as osteopenia, \nosteoporosis, and Schmorl's nodule. \nMethods or Background: Fat(HAP), fat(calcium), hydroxyapatite(fat), and \ncalcium(fat) densities [Dfat(HAP), Dfat(calcium), D HAP(fat), and Dcalcium(fat)], \nas well as CT value were collected. Comparisons wer e made using the \nindependent sample T test. Diagnostic performance w as assessed in terms of \narea under the receiver operating characteristic cu rve (AUC). The sensitivity, \nspecificity, positive predictive value (PPV), negat ive predictive value (NPV), \nand accuracy of each parameter was assessed as well . \nResults or Findings: A total of 106 patients were included, of whom 24 h ad \nSOMs (lesion, n = 48), while 82 had SOM-mimics (les ion, n = 202). SOMs \nassociated with significantly higher CT value, Dfat (calcium), and Dfat(HAP) \ncompared to SOM-mimics (P < .001). The AUCs were 0. 674, 0.879, and 0.887, \nrespectively. The sensitivity, specificity, PPV, NP V, and accuracy of \nfat(calcium) were 77.1%, 85.1%, 55.2%, 94.0%, 83.6% , respectively; while \nthose for Dfat(HAP) were 83.3%, 80.7%, 50.6%, 95.3% , 81.2%, respectively. \nThe optimal diagnostic cutoffs for Dfat(calcium) an d Dfat(HAP) were ≥ 1000.0 \nmg/cm3 and ≥ 966.9 mg/cm3, respectively, which achieved consist ent \ndiagnostic results among 89.6% lesions (n = 224). T he combined use of \nDfat(HAP) and Dfat(calcium) achieved significantly better diagnostic \nperformance, with AUC, sensitivity, specificity, PP V, NPV, and accuracy of \n0.910, 82.2%, 87.2%, 61.7%, 95.1%, and 86.2%, respe ctively. \nConclusion: Dfat(calcium) and Dfat(HAP) on DECT carry the poten tial as \nparameters for the discrimination of SOMs from SOM- mimics \nLimitations: This was a retrospective study with a relatively sm all sample size. \nIn addition, the focus on thoracolumbar lesions lim its the generalizability of our \nresults. \nFunding for this study: Natural Science Foundation of Fujian Province, Chin a \n(grant numbers: 2023J01181) \nEthics committee - additional information: Fujian Cancer Hospital Ethics \nCommittee (K2023-198-01) \nAuthor Disclosures:  \nJianfang Liu: Nothing to disclose \nJie Li: Nothing to disclose \n \n \n\n \n \nSunday \nAbstract-based Programme \n \n 293  \nMultimodal machine learning method for the identifi cation of prognostic \nand predictive biomarkers in Adolescent and Young A dults (AYA) \nsarcoma: a pilot study \n*S. Lusi*, R. Romanelli, A. Marzullo, A. Laffi, A. Santoro, L. Balzarini,  \nM. Francone, A. F. Bertuzzi; Milan/IT \n \nPurpose or Learning Objective: This study relies on the use of Artificial \nIntelligence (AI) to develop and validate a multimo dal machine learning method \nthat could provide a prognostic model in AYA patien ts affected by sarcomas, \nexploring the clinical, radiomic and pathological f eatures that may be predictive \nof disease outcome. \nMethods or Background: The study is a monocentric retrospective cohort \nstudy involving 245 patients with sarcomas. Our pre liminary and full results \nwere performed on a smaller cohort of 22 patients w ith soft tissue sarcoma of \nthe extremities (13 non-AYA and 9 AYA) for whom cli nical data (using Excel \nform), radiomic features (from a pre-treatment MRI)  and histopathological \nfeatures (extracted using a foundation model) were collected. All this data was \nthen used to match similar patient profiles in the two groups using logistic \nregression propensity scores. Disease-free survival  of matched patients was \ndescribed using a Kaplan-Meier curve. \nResults or Findings: Statistical analysis didn't identify any correlatio n \nbetween clinical and radiological features that cou ld explain the differences in \nprognosis between the two groups, probably due to t he small cohort size. \nHowever, AI analysis using a Kaplan-Meier curve sho wed that AYA patients \nhad a worse prognosis than non-AYA patients (p < 0. 05), confirming for the \nfirst time, to our knowledge, by deep machine learn ing what is observed in \nclinical practice. \nConclusion: Despite the limitations of these preliminary result s based on a \nsmall cohort of patients, our findings provide valu able insights into the \ndifferential prognosis that characterises these two  groups. AI holds promise for \nuncovering hidden characteristics, with future rese arch potentially incorporating \nbiological markers to further explore therapeutic t argets. \nLimitations: The small cohort size of the study limits its stati stical power. The \nretrospective design may introduce selection bias. Future studies in larger \npopulations are needed. \nFunding for this study: No funding was received for this study. \nEthics committee - additional information: The study is retrospective. \nAuthor Disclosures:  \nAlexia Francesca Bertuzzi: Nothing to disclose \nArmando Santoro: Nothing to disclose \nAldo Marzullo: Nothing to disclose \nAlice Laffi: Nothing to disclose \nMarco Francone: Nothing to disclose \nRoberta Romanelli: Nothing to disclose \nStefano Lusi: Nothing to disclose \nLuca Balzarini: Nothing to disclose \n \n \nRadiomics in MRI to improve the characterization of  cartilaginous bone \ntumours \n*Q. Bui*, M. Lacroix, L. S. Fournier, F. Larousseri e, A. Feydy; Paris/FR \n \nPurpose or Learning Objective: In long bones, distinguishing between \nenchondromas and chondrosarcomas before surgery is often challenging and \nmay require invasive biopsy for accurate diagnosis.  The purpose of this work \nwas to assess the performance of MRI radiomics-base d machine learning in \nclassifying enchondromas and chondrosarcomas in lon g bones. \nMethods or Background: Ninety-eight patients with pathology-proven \ncartilaginous tumours of long bones were retrospect ively included from a \ntertiary bone tumour centre. The training set consi sted of 81 MRI (n = 33 \nenchondromas; n = 48 chondrosarcomas). The internal  test set consisted of 17 \nMRI (n = 7 enchondromas; n = 10 chondrosarcomas). 3 D segmentation was \nperformed on T1-weighted and fat-suppressed T2-weig hted MRI images and \nradiomics features were extracted. Dimensionality r eduction was performed \nbased on reproducibility, redundancy and feature im portance. Different models \nwere tested, including multiparametric, single sequ ence and sequential. A \nRandom Forest classifier was tuned on the training set using five-fold cross-\nvalidation and tested on the internal test set. \nResults or Findings: The Random Forest classifier with the T2 then T1 \nsequential model, which was the best-performing mod el, achieved an AUC of \n0.943 [0.832 – 1.000] on the internal test set. Its  accuracy in correctly \nclassifying enchondromas and chondrosarcomas was 71 % (5/7) and 100% \n(10/10), respectively. \nConclusion: This work shows that MRI radiomics can accurately d ifferentiate \nbetween benign and malignant cartilaginous tumours.  Although comparisons \namong various models did not achieve statistical si gnificance, the data suggest \nthat a sequential approach using single sequence mo dels might outperform a \nmultiparametric model. Further studies with larger sample sizes are needed to \nintegrate these findings into clinical practice and  improve preoperative \ndiagnosis of cartilaginous tumours within the conte xt of personalised medicine. \nLimitations: Limitations of the study were the small sample size  and the lack \nof an external test set. \nFunding for this study: The author received a research grant from Societé \nFrançaise de Radiologie and Assistance Publique - H ôpitaux de Paris. \nEthics committee - additional information: This study involved a \nretrospective analysis of anonymised data collected  as part of routine care. \nAuthor Disclosures:  \nLaure S. Fournier: Nothing to disclose \nAntoine Feydy: Nothing to disclose \nQuentin Bui: Research/Grant Support: Société França ise de Radiologie \nResearch/Grant Support: Assistance Publique - Hôpit aux de Paris \nFrederique Larousserie: Nothing to disclose \nMaxime Lacroix: Nothing to disclose \n \n \nOptimizing Cryoablation Outcomes in Desmoid Tumors:  A Machine \nLearning-Driven Radiomic Analysis \n*M. E. Chevasco Hanze*, L. Ponsa Cobas, J. A. Narvá ez García,  \nD. A. Sandoval Díaz, J. Hernández Gañan, J. C. Sard iñas Barrero; \nBarcelona/ES \n(Miguelemilioch27@gmail.com) \n \nPurpose or Learning Objective: Desmoid tumors (DT) are locally aggressive, \ninfiltrative neoplasms with a high risk of local re currence. Recently, \npercutaneous cryoablation has emerged as an alterna tive therapy, though its \nrole as a salvage treatment remains unclear. This s tudy aimed to evaluate \ndisease progression after cryoablation at a 1-year follow-up and develop a \npredictive model using clinical and radiomic variab les. \nMethods or Background: We performed a retrospective analysis of patients \ntreated with cryoablation for extra-abdominal DT fr om January 2018 to \nSeptember 2023. Pre- and post-cryoablation contrast -enhanced MRIs were \nreviewed, and disease progression was defined as le ss than 90% necrosis or \nnon-enhancement at follow-up. Radiomics features we re extracted from T2-\nweighted pre-cryoablation MRIs, and data were filte red based on correlation \nmatrices and statistical tests (T-Student, Mann-Whi tney U, Chi-square). \nPrediction models, including LASSO, Random Forest, XGBoost, SVM, and \nKNN, were evaluated using ROC analysis and 5-fold c ross-validation to \ndetermine the optimal approach. \nResults or Findings: Twenty-eight patients were included (median age 43;  \n67% women), with a no disease progression rate of 6 0.71%, significantly \nassociated with partial response on mRECIST criteri a (p = 0.022). The \nRandom Forest model showed the best performance (AU C = 0.77). Key \npredictive features included tumor diameter, spheri city, major axis length, \nminimum intensity, kurtosis, interquartile range, a nd tumor location. Tumors \n>61 mm, with ellipsoid shape (major axis length >79  mm) and regular form \n(sphericity <0.7), predicted no disease progression . Similarly, tumors <61 mm \nwith a regular shape (sphericity <0.5) and fibrous matrix (minimum <175) \npredicted favorable outcomes. \nConclusion: Cryoablation therapy has demonstrated a good therap y for DT \ntreatment. Radiomics shape and first order features  have shown their \nrelevance in cryoablation therapy planning as it se rve as a patient selection \ntool. \nLimitations: Small sample size. \nNo split train-test approach. \nFunding for this study: No funding was received \nEthics committee - additional information: No intervention, just an \nobservational study \nAuthor Disclosures:  \nDaniel Alejandro Sandoval Díaz: Nothing to disclose  \nLaia Ponsa Cobas: Nothing to disclose \nMiguel Emilio Chevasco Hanze: Nothing to disclose \nJavier Hernández Gañan: Nothing to disclose  \nJosé Antonio Narváez García: Nothing to disclose \nJuan Carlos Sardiñas Barrero: Nothing to disclose \n \n \n“Pseudo-CT” MRI sequences and detection of lytic le sions in multiple \nmyeloma \n*C. Chabot*; Brussels/BE \n(caroline.chabot@hotmail.com) \n \nPurpose or Learning Objective: To assess the diagnostic accuracy, \nrepeatability, and reproducibility of pseudo-CT MRI  sequences (ZTE, BB) in \ndetecting osteolytic lesions in MM using WB-CT as t he reference standard. \nMethods or Background: In this prospective study, consecutive patients wer e \nenrolled in our academic hospital. Inclusion criter ia were newly diagnosed MM, \nmonoclonal gammopathy of undetermined significance at high risk for MM, or \nsuspicion of progressive MM. Participants underwent  ZTE and BB sequences \ncovering the lumbar spine, pelvis, and proximal fem urs as part of 3T WB-MRI \nexaminations, as well as clinically indicated 18F-F DG PET/CT examination that \nincluded optimized WB-CT. Ten bone regions and two scores (categorical \nscore/semiquantitative score) were assessed by thre e radiologists on the ZTE, \nBB, and WB-CT images. The accuracy, repeatability, and reproducibility of \ncategorical scores (Gwet agreement coefficients AC1  and AC2) and \n\n \n \nSunday \nAbstract-based Programme \n \n 294  \ndifferences in semiquantitative scores were assesse d at per-sequence, per-\nregion, and per-patient levels. \nResults or Findings: 47 participants were included. In experienced reade rs, \nBB and ZTE showed 98% accuracy per-patient, while B B accuracy ranged \nfrom 83%–100% and ZTE from 74%–94% per-region. Incr eased false-negative \nfindings in the spine ranging from 17%-23%, accordi ng to the lumbar vertebra, \nwas observed using ZTE(P<.013). Regardless of the r egion (except coxal \nbones), differences in the BB score minus the ZTE s core were positively \nskewed(P<.021). Repeatability was very good(AC1 ≥0.87), while reproducibility \nwas at least good(AC2≥0.63). \nConclusion: Both MRI-based ZTE and BB pseudo-CT sequences of th e \nlumbar spine, pelvis, and femurs demonstrated high diagnostic accuracy in \ndetecting osteolytic lesions in MM. Compared with B B, the ZTE sequence \nyielded more FN findings in the spine. \nLimitations: Pseudo-CT sequences were limited to the lumbar spin e, pelvis, \nand femurs; the reference CT required optimization from 18F-FDG PET/CT; \nfocus was on detecting osteolytic lesions, includin g nonactive ones that may \npersist post-treatment. \nFunding for this study: None \nEthics committee - additional information: This prospective study was \napproved by the institutional ethics committee (202 0/27JUL/380) and is \nregistered on ClinicalTrials.gov (no. NCT05381077).  Written informed consent \nwas obtained from all participants. \nAuthor Disclosures:  \nCaroline Chabot: Nothing to disclose \n \n \nImplementing tin prefiltration in routine clinical CT scans of the lower \nextremity: Impact on radiation dose \n*T. Marth*, C. Stern, R. Sutter; Zürich/CH \n(thomas-marth@hotmail.com) \n \nPurpose or Learning Objective: Several studies have demonstrated the \npotential of tin prefiltration to reduce the radiat ion dose while maintaining \ndiagnostic quality for musculoskeletal imaging. Sti ll, no study has reported data \non the impact of radiation dose reduction on clinic al routine scanning. \nMethods or Background: Retrospective inclusion of 300 patients who \nunderwent clinically indicated CT scans of the pelv is, knee, or ankle on a single \nCT scanner (SOMATOM Definition AS, Siemens Healthin eers) before January \n2020 (without tin filter) and after December 2020 ( with tin filter). For each joint, \n50 patients with tin prefiltration and 50 patients without tin prefiltration were \nselected. \nResults or Findings: The CTDIvol, DLP, and effective dose were reduced \nsignificantly in all tin-prefiltered examinations c ompared to the conventional, \nnon-tin-prefiltered examinations (p<.001). Tin-pref iltered CT scans had a \nsignificantly lower CTDIvol and mean effective dose  (all p<.001): CTDIvol was \n65% lower in the pelvis, 73% in the knee, and 54% i n the ankle. This resulted \nin a reduction of the effective dose of 61%, 71%, a nd 60%, respectively. In \nabsolute numbers, the reduction of the median effec tive dose delivered in a \nsingle scan of the pelvis was 2.29 mSv, 0.15 mSv fo r the knee, and 0.03 mSv \nin the ankle. \nConclusion: The implementation of tin prefiltration in clinical  routine scan \nprotocols significantly reduced the effective radia tion dose for unenhanced CT \nscans of the pelvis (61% reduction, 2.29 mSv), the knee (71% reduction, 0.15 \nmSv), and the ankle (60% reduction, 0.03 mSv). \nLimitations: It would be possible to reduce radiation dose even more by \napplying deep learning-based denoising algorithms, however this was not yet \navailable in clinical routine during the data acqui sition. \nFunding for this study: No specific funding. \nEthics committee - additional information: BASEC-ID 2024-01094 \nKantonale Ethikkommission Zürich \nAuthor Disclosures:  \nThomas Marth: Other: Balgrist University Hospital a nd Balgrist Campus each \nhave an academic research collaboration with Siemen s Healthineers. Balgrist \nUniversity Hospital also has an academic research c ollaboration with Bayer. \nChristoph Stern: Other: Balgrist University Hospita l has an academic research \ncollaboration with Siemens Healthineers. Balgrist U niversity Hospital also has \nan academic research collaboration with Bayer. \nReto Sutter: Other: Balgrist University Hospital ha s an academic research \ncollaboration with Siemens Healthineers. Balgrist U niversity Hospital also has \nan academic research collaboration with Bayer. \n \n \n \n \n \n \n \n \n \n \nEvaluation of the diagnostic potential of a Tomosyn thesis system for \nMSK \nY. Beer¹, *N. Shabshin*², L. Copel¹, Y. Kimmel³, R.  Ophir¹,  \nY. S. Schiffenbauer³, S. Tal¹; ¹Zrifin/IL, ²Afula/I L, ³Petach Tikva/IL \n(nogah.shabshin@gmail.com) \n \nPurpose or Learning Objective: Digital tomosynthesis (DTS) is a well-\nestablished technology that has become the gold sta ndard for breast \nmammography. In recent years its benefits in muscul oskeletal (MSK) imaging \nhave been acknowledged, leading to a rapid increase  in its utilization. It \nimproves detection, localization and characterizati on of subtle fractures. In \nsome patients DTS can alleviate the need for CT wit h lower complexity. \nRecently a new technology based on Cold Cathode Xra y tube with a multi tube \nset-up has made this technology more affordable and  accessible. The purpose \nof this study is to evaluate the diagnostic potenti al of the cold-cathode multi-\ntube DTS \nMethods or Background: The study included 19 patients with suspected \nfractures who underwent CT and radiographs (XR). Pa tients were scanned \nusing the cold-cathode DTS . 15 patients had imagin g performed with Cast or \nmetal. Images were evaluated by an MSK radiologist and orthopedic surgeon \nin consensus with CT as the gold standard. Studies were evaluated for \npresence, location, intraarticular involvement, dis placed fragments and \nincidental lesions. The surgeon was asked whether D TS provided valuable \ninformation and whether it increased the confidence  of the final diagnosis. \nResults or Findings: In 17/19 studies DTS added value to the XR. In 7, D TS \nfound fractures occult in XR. In 3, DTS was able to  clear a suspected fracture. \nin 5, DTS was able to better localize the fracture.  In 1, DTS was able to \ndetermine fracture age and in 1 study DTS found a s clerotic lesion obscured in \nXR. In addition, on XR the cast limited evaluation of fine bony details, however \nthere was no such limitation with DTS \nConclusion: Cold cathode DTS provides high quality tomography o f \nanatomies enabling depiction of occult pathologies,  localization, \ncharacterization and resolution of questionable fin dings \nLimitations: Initial study results \nFunding for this study: Study was funded by Nanox-x Imaging Ltd. \nEthics committee - additional information: Study was approved by the local \nEthics committee in the institution and each patien t underwent informed \nconsent \nAuthor Disclosures:  \nRobenpour Ophir: Nothing to disclose \nLaurian Copel: Nothing to disclose \nYiftah Beer: Nothing to disclose \nYotam Kimmel: Nothing to disclose \nYael S Schiffenbauer: Nothing to disclose \nNogah Shabshin: Consultant: Nano-x imaging \nSigal Tal: Nothing to disclose","source_license":"CC-BY-4.0","license_restricted":false}