Kupffer phase radiomics signature in Sonazoid contrast-enhanced ultrasound predicts immunohistochemistry marker expression in hepatocellular carcinoma

Kupffer phase radiomics signature in Sonazoid contrast-enhanced ultrasound predicts immunohistochemistry marker expression in hepatocellular carcinoma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Kupffer phase radiomics signature in Sonazoid contrast-enhanced ultrasound predicts immunohistochemistry marker expression in hepatocellular carcinoma Chen Li, Yuan Liu, Mingxiao Wu, Weide Dai, Jinghai Song, Hanzhang Wang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5362429/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose: Few studies have explored the value of radiomics signatures in predicting immunohistochemical (IHC) staining markers. This study aimed to investigate and validate radiomics models based on the Kupffer phase of Sonazoid contrast-enhanced intraoperative ultrasonography (S-CEUS) images for predicting IHC marker expression in hepatocellular carcinoma (HCC). Method: Overall, 113 consecutive patients diagnosed with HCC between November 2019 and May 2023 were retrospectively analyzed. Histopathological assessment included IHC staining for GS, CD10, GPC3, and HSP70. Radiomic features extracted from S-CEUS images were selected and analyzed. A Naïve Bayes classifier was employed to predict IHC marker expression in HCC, using selected clinical biomarkers and radiomic features. Results: For GPC3, the radiomics classifier achieved a macro-average area under the receiver operating characteristic curve (AUC) of 0.700, indicating strong performance. For GS, both radiomics and combined clinical-radiomics classifiers exhibited strong discrimination (AUCs: 0.870 and 0.882, respectively). The radiomics classifier outperformed clinical biomarkers (total and direct bilirubin) in predicting CD10, with a macro-average AUC of 0.834. However, its accuracy decreased for higher HSP70 marker expression levels (AUC: 0.694). These findings underscore the consistent effectiveness of radiomics across different IHC markers when compared to traditional clinical approaches. Conclusions: The Kupffer phase in the S-CEUS-based radiomics signature is an excellent biomarker for predicting IHC marker expression in patients with HCC. hepatocellular carcinoma contrast-enhanced ultrasound radiomics immunohistochemical markers Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Hepatocellular carcinoma (HCC), one of the most common primary hepatic malignant tumors, is the third leading cause of cancer-related deaths worldwide, with a 5-year survival rate of < 20% [ 1 , 2 ]. Several therapies, such as hepatic surgical resection, transplantation, and transcatheter arterial chemoembolization, are effective and commonly used to treat HCC. Among these, hepatectomy is the preferred therapeutic method for most patients with HCC. The prognosis of HCC has improved with the advances in hepatectomy and imaging technologies. However, its high potential for vascular invasion, metastasis, and recurrence post-resection leads to a poor prognosis [ 3 , 4 ]. Glypican-3 (GPC3), glutamine synthetase (GS), and heat shock protein 70 (HSP70) are biomarkers currently used to discriminate the nature of hepatocellular lesions smaller than 2 cm detected in patients with liver cirrhosis, which lack the radiological features of HCC 5. CD10 is a tissue marker used to confirm the diagnosis of HCC through imaging in histology [ 5 ]. Among the prognostic factors for HCC, GPC3 has been closely associated with postoperative metastasis/recurrence in patients with HCC [ 6 ]. The expression of GPC3 in HCC is an important independent factor in predicting a patient’s poor prognosis. International guidelines recommend the combined use of these immunohistochemical (IHC) markers for a more accurate diagnosis. Compared with early HCC, assessable using this panel of markers, our clinical challenge involved distinguishing dysplastic nodules [ 7 ]. Radiomics analysis is an emerging medical imaging technique that has attracted considerable attention in recent years. Radiomics can be used to extract high-dimensional information that is invisible to humans [ 8 ]. This feature could be used in the characterization of tumor heterogeneity, as well as reflecting the tumor tissue microenvironment and cancer phenotype [ 9 ]. Consequently, the radiomics signature has become a prognostic biomarker that can augment available clinical data, aid in lesion detection, improve the accuracy of diagnosis, predict the risk of disease, and determine treatment strategies [ 10 ]. Ultrasound, particularly Sonazoid contrast-enhanced ultrasound (S-CEUS), has been successfully applied to predict the pathological grading of tumors, evaluate malignancy, and assess treatment response with liver-specific contrast agents such as Sonazoid. This agent enables prolonged Kupffer phase imaging, more effectively characterizing focal liver lesions [ 11 , 12 ]. Several studies have reported that radiomics signatures derived from S-CEUS show potential in predicting microvascular invasion, histopathological grade, and Ki-67 protein expression levels in HCC [ 12 – 14 ]. However, to our knowledge, few studies have attempted to identify the potential of radiomics signatures in predicting other IHC markers, such as GPC3, GS, HSP70, and CD10. This study aimed to investigate and validate the performance of radiomics models based on the Kupffer phase of S-CEUS images in predicting the expression of IHC markers in HCC. Materials and methods Patients Ethical approval was obtained from the Institutional Review Board (2021BJYYEC-190-02), which waived the requirement for informed consent. All procedures were performed in accordance with the Declaration of Helsinki. A total of 113 consecutive patients with HCC were retrospectively considered between November 2019 and May 2023. The inclusion criteria were as follows: 1) patients with a suspected diagnosis of HCC; 2) possession of complete medical information; 3) having undergone preoperative grayscale ultrasound and Kupffer phase imaging of S-CEUS; 4) no previous treatments, such as radiofrequency ablation, microwave ablation, and chemotherapy treatment; and 5) a pathologic report confirming HCC, with routine immunochemical staining for GS, CD10, and GPC3 conducted. The exclusion criteria were: 1) absence of S-CEUS; 2) incomplete clinical information; 3) lack of an available IHC report; 4) contraindication for the use of the Sonazoid contrast agent; and 5) poor B-mode or S-CEUS image quality of focal liver lesions. Potential clinical biomarkers were identified through patient interviews and a thorough review of medical records. The clinical characteristics of the patients included age, sex, maximum tumor diameter, alpha-fetoprotein levels, total bilirubin (TBil), direct bilirubin (DBil), carbohydrate antigen 12-5 (CA125), carbohydrate antigen 19-9 (CA199), carbohydrate antigen 15-3 (CA153), aspartate transaminase (AST), and alanine transaminase (ALT). Data on risk factors, including hepatitis and alcohol consumption, were collected and recorded. Histopathological result and immunohistochemical staining procedure Histopathological results, including the Edmondson–Steiner (E-S) grade and IHC markers, were evaluated by a pathologist with >10 years’ experience. The resected specimens were fixed in 10% paraformaldehyde, embedded in paraffin, and cut into 4-μm-thick sections for hematoxylin–eosin staining or IHC identification. The histological grade of HCC was determined based on the E-S grade [15]. Grades I and II were classified as low grade, whereas grades III and IV were classified as high grade. For the IHC analysis, the expression levels of GS, CD10, and GPC3 were categorized. Negative expression (no immunoreactivity, -), mildly positive expression (50% immunoreactivity, ++) were identified based on the staining patterns [15,16]. Sonazoid contrast-enhanced ultrasound imaging All patients underwent routine grayscale ultrasonography and S-CEUS. Two sonographers conducted the S-CEUS examinations using an Aplio 500 (Canon), equipped with convex (6C1, 1-6 MHz) and linear (11L4, 4-11 MHz) probes, and an Aplio i800 (Canon), equipped with convex (PVI-475BX, 1-8 MHz) and linear (11L4, 4-11 MHz) probes. According to the size of the lesion, the mechanical index of the acoustic output was set to 0.19-0.22 with a dynamic range of 65-70 dB. The patients were injected with 0.5 mL of Sonazoid through a peripheral venous line, followed by 5 mL of saline. The enhancement features were recorded and analyzed according to the latest WFUMB guidelines [11]. The Kupffer phase was obtained by scanning for 15 min. The best frames were identified from the grayscale ultrasound image and the Kupffer phase image was sampled for radiomic analysis. Radiomics data acquisition In the radiomics analysis, a single cine depicting the largest cross-sectional view of the HCC lesion was selected from both B-mode and Kupffer phase ultrasound images by an ultrasound radiologist. This was stored in the Digital Imaging and Communications in Medicine format. The region of interest, which defined the tumor area, was delineated by a senior ultrasound radiologist using ITK-SNAP (an open-source software, http://www.itksnap.org). Radiomic features were extracted using NovoUltrasound Kit (NUK V1.5.0, GE HealthCare). The features considered, including shape-related, pixel intensity-based histogram, and texture features (GLCM, GLDM, GLRLM, GLSZM, and NGTDM), were defined according to the IBSI standard. Prior to feature extraction, B-mode ultrasound and S-CEUS cines were transformed to a grayscale format. Numerous filters (such as Laplacian of Gaussian, wavelet, and local binary patterns) were applied to the original images for better characterization across different spectra. The feature values were normalized based on the Z-score. Radiomic features from B-mode and Kupffer phase ultrasound images were fused into a single feature vector for analysis (Figure 1). Multi-class prediction of IHC markers Given the high dimensionality and inter-feature redundancy of radiomics, a feature selection process utilizing analysis of variance (ANOVA) and Spearman correlation (ANOVA–Spearman) was conducted to eliminate features with insignificant variance among IHC groups and those showing significant inter-feature correlation (Spearman’s ρ > 0.75). We employed the commonly used Naïve Bayes classifier for multi-classification tasks. Two models were developed: a Naïve Bayes classifier using ANOVA-selected clinical biomarkers, and a combined model using both the selected clinical biomarkers and radiomic features (Figure 1). Statistical methods Descriptive statistics for continuous clinical biomarkers are presented either as mean ± standard deviation or median (inter-quartile range), depending on their statistical distribution determined by the Shapiro–Wilk test ( P value > 0.05 indicates normal distribution). The 95% confidence intervals (CIs) were determined via bootstrapping (1,000 resamples). Accuracy was used to provide a straightforward measure of the proportion of correct predictions. The receiver operating characteristic (ROC) and the area under the ROC (AUC) curve were utilized to assess the discriminative power of individual prediction classes. The F1 score denoted the harmonic mean of precision and recall. The macro-averaged AUC offered an aggregate perspective on discrimination performance across all classes. Cohen’s kappa measures chance correlations on a scale from 0 to 1, where 1 indicates optimal reliability, and values below 0 indicate performance inferior to randomness. It is categorized as slight, fair, moderate, substantial, or perfect [17]. The net reclassification improvement (NRI) metric was used to quantitatively evaluate the combined model’s proficiency in accurately reclassifying subjects compared with a benchmark model, specifically the clinical biomarker model in this context. An NRI exceeding 0 indicates enhanced classification. P values <0.05 denoted statistical significance. Results A total of 86 patients with 106 histologically confirmed HCC nodules (89.6% of which are male, average age 62.02 ± 10.21 years) were consecutively enrolled in this study. The variance among different classes of the three IHC markers was measured. Of the 106 nodules, 7 (6.6%) were GS negative, 40 (37.7%) were GS mildly positive, and 59 (55.7%) were GS strongly positive. With regard to CD10 expression, 31 nodules (29.2%) displayed negative expression, 67 (63.2%) displayed mild positive expression, and 8 (7.5%) displayed strongly positive expression. Significant differences in pathological grade were observed across different classes of the IHC markers GS and GPC3 among the examined clinical biomarkers. TBil and DBil were identified as significant IHC markers for CD10 (Table 1). Prediction result for the IHC marker GS Eleven radiomic features from B-mode ultrasound and 12 features from the Kupffer phase of S-CEUS were identified as relevant to GS expression by using the ANOVA–Spearman selection process (Table 2). Figure 2 presents the ROC curves for the classifiers based on clinical data (pathological grade), radiomics, and a combined clinical-radiomics approach. The combined Naïve Bayes classifier, integrating pathological grades with radiomic features, demonstrated a superior AUC compared with the classifiers using clinical or radiomics data alone. Both radiomics and combined classifiers exhibited a higher per-class AUC for discriminating between GS-negative and GS-positive (mildly positive, strongly positive) expression in patients, with AUCs of 0.870 and 0.882, respectively. The macro-averaged AUC, which represents the overall performance of both the radiomics and combined classifiers, was 0.706 (Figure 2). Thus, in terms of AUC values, the combination method was more effective in classifying GS expression than the clinical method alone. However, adding pathological grades to the radiomics analysis did not improve these results. Regarding diagnostic performance, the combined classifier achieved a higher accuracy, F1 score, and Cohen’s kappa than the clinical and radiomics classifiers (Table 3). Both the radiomics and combined methods demonstrated fair accordance and yielded a positive NRI compared with the clinical method, with NRI values of 0.367 and 0.454, respectively. Additionally, a positive NRI value of 0.048 was observed when comparing the combined method with the radiomics classifier. These positive NRI values indicate that incorporating radiomics-based classifications improved the classification results of the clinical biomarkers alone. Prediction result for the IHC marker CD10 For the IHC marker CD10, the ANOVA–Spearman selection process identified 14 radiomic features from B-mode images and four from Kupffer phase images (Table 4). The classifier based on the clinical biomarkers TBil and DBil was less effective according to the ROC analysis. In contrast, the radiomics classifier demonstrated superior discrimination performance, evidenced by the highest macro-averaged AUC value of 0.834 and the highest per-class AUC value for each class of CD10 expression level. Compared with TBil/DBil, the combined classifier showed increased performance across all measured metrics (Figure 3, Table 5). The NRI for the addition of radiomics to TBil/DBil was 0.137, indicating a positive improvement in classification results. Prediction result for the IHC marker GPC3 Eight B-mode radiomic features and six Kupffer phase features were selected for the GPC3 expression classification (Table 6). ROC analysis showed that the radiomics classifier was the most adequate, with the highest macro-average AUC value of 0.700. The radiomics classifier achieved similar results to the clinical classifier in discriminating GPC3 positive/negative patients (per-class AUC 0.637 versus 0.640). However, it was more effective in classifying positive patients into mild or strong expression categories (Figure 4). While combining clinical and radiomic features did not improve the AUC results compared with radiomics alone, this approach was noticeably better than using the clinical classifier alone. The clinical classifier demonstrated an inferior diagnostic value, as its classification results showed only marginal concordance with the actual expression levels. By contrast, both the radiomics classifier and combined methods displayed fair agreement with the actual values, with the radiomics classifier maintaining the highest metric values (Table 7). While the combined classifier improved the classification results of the clinical classifier (NRI = 0.134), it did not improve the classification results of the radiomics classifier (NRI = -0.007). Prediction result for the IHC marker HSP70 An ANOVA conducted on the clinicopathological variables revealed no significant differences between the HSP70 expression groups. Fourteen radiomic features (seven from each ultrasound mode) were related to HSP70 (Table 8). A Naïve Bayes classifier, utilizing these radiomic features, was constructed based on the diagnostic performances listed below (Table 9, Figure 5). The Naïve Bayes classifier, based on ultrasound radiomic features, demonstrated negative (-) and positive (+, ++, or +++) features with excellent performance (per-class AUC 0.880). However, its per-class discrimination ability diminished as the expression levels increased, with an AUC of only 0.694 in distinguishing (+++) features from those in the rest of the specimens. Discussion With the development of precision medicine, identifying new quantitative and radiomics-based noninvasive imaging biomarkers aimed at improving the predictive performance of medical images has become an area of interest in radiological research [ 18 , 19 ]. In recent years, with the development of new HCC treatment methods, such as immune checkpoint inhibitors [ 20 ] and regorafenib [ 21 ], radiomics methods, including texture analysis, shape, and intensity features, have shown potential relevance between medical imaging and personalized medicine [ 22 ]. The aim of our study was to propose and validate radiomics models based on the Kupffer phase of S-CEUS images for the potential prediction of IHC markers. The data showed that the radiomics model originating from S-CEUS images achieved promising results, suggesting its feasibility and applicability in clinical practice. In this retrospective study, we extracted radiomic features from the B-mode and Kupffer phases associated with IHC biomarkers, which is also the core principle of radiomics, and selected these data for radiomics score construction [ 23 ]. The developed radiomics model showed a mathematical correlation between ultrasound radiomics characteristics and IHC biomarkers. Utilizing the unique tissue features of the Kupffer phase, the CE-mode was constructed using quantitative radiomic characteristics from the S-CEUS Kupffer phase images to distinguish the stages of HCC in IHC biomarkers [ 13 ]. In our study, the classifier that combined pathological grades of clinical risk factors with the radiomics signature was more effective at classifying GS and GPC3 expression than either the clinical or radiomics classifier alone. While the combination of clinical and radiomic features did not improve the AUC results compared with radiomics alone, it was noticeably better than the clinical classifier alone. The combined classifier improved the classification results of the clinical classifier, but it did not improve those of the radiomics classifier. The radiomics classifier demonstrated superior discrimination performance for each class of CD10 expression level. When compared with the TBil/DBil of clinical risk factors, the combined classifier showed more competitive performance across all measured metrics. The radiomics classifier identified negative and positive results with excellent performance for HSP70 expression levels (per-class AUC 0.880). The combined classifier achieved optimal predictive performance. This is attributed to the comprehensive overview provided by both qualitative and quantitative imaging features in the final model, elucidating the relationship between radiomic characteristics and the pathological status of HCC [ 18 ]. Radiomic features, including shape, intensity, and texture information, can present the complexity of the target tissue properties. Typically, the heterogeneity exhibited by malignant lesions is meticulously captured and quantified through subjective interpretations by experienced radiologists. This heterogeneity spans multiple domains (e.g., gross, cellular, genetic, and phenotypic levels) and dimensions (e.g., cellular density, angiogenesis, hemorrhage, and necrosis) [ 24 ]. Our tumor surrounding dilation radiomics provides a comprehensive and microscopic high-throughput prediction just like “virtual tissue pathology [ 25 ]”. Previous studies have shown that the imaging features constructed from the Kupffer phase of S-CEUS images, including textural features, may predict Ki-67 expression in HCC [ 13 ]. An effective tool is provided for the non-invasive and individualized prediction of GPC3-positive HCC through a predictive nomographic chart combining alpha-fetoprotein and radiomic features [ 26 ]. However, to the best of our knowledge, no other studies have investigated the potential value of integrating the quantitative analysis of clinical factors in predicting IHC biomarkers, such as GPC3, HSP70, GS, and CD10, in HCC. Our study investigated the prediction of IHC markers using radiomic feature models based on both the B-mode and Kupffer phases, which can only be achieved through invasive biopsy or surgery. These data indicate that radiomics models derived from Kupffer phase S-CEUS images achieved good results and are deemed feasible for application in clinical practice. The complicated correlation between bioprocesses and radiomic signatures poses a notable challenge, aligning with the current trend in precision and individualized medicine [ 27 , 28 ]. This study has several limitations. First, as this was a single-center study, external validation through additional multicenter cohorts may be required to verify the reliability and assess the generalizability of our results. Second, the results of our study on radiomic features from the arterial mode were unsatisfactory, leading to the dismissal of these results. Further investigation into the diagnostic value of dynamic graphs, including the arterial, portal venous, late phase, and Kupffer phase of SCEUS, is warranted. Finally, the absence of multimodal radiological data in this study should be noted. The potential use of S-CEUS combined with gadoxetic acid-enhanced magnetic resonance imaging and computed tomography may provide additional properties to enhance prediction accuracy. In the future, the multimodality of ultrasound imaging, including color Doppler flow imaging, ultrasound elastography, and the vascular phase of CEUS imaging, combined with magnetic resonance imaging and computed tomography data, could be used to explore the predictive performance of IHC biomarkers. In conclusion, the Kupffer phase in the S-CEUS-based radiomics signature is an excellent biomarker. It has achieved desirable results in predicting the IHC expression of GPC3, GS, HSP70, and CD10 in patients with HCC. The combined clinical factors and radiomics signature may provide an effective tool for the noninvasive and individualized prediction of immunohistochemistry expression in HCC. Abbreviations HCC, hepatocellular carcinoma; GPC3, glypican-3; GS, glutamine synthetase; HSP70, heat shock protein 70; IHC, immunohistochemical; S-CEUS, Sonazoid contrast-enhanced ultrasound; TBil, total bilirubin; DBil, direct bilirubin; CA125, carbohydrate antigen 12-5; CA199, carbohydrate antigen 19-9; CA153, carbohydrate antigen 15-3; AST, aspartate transaminase; ALT, alanine transaminase; E-S, Edmondson–Steiner; ANOVA, analysis of variance; ANOVA-Spearman, Spearman correlation of ANOVA; ROC, receiver operating characteristic; AUC, area under the ROC; NRI, net reclassification improvement Declarations Funding information This work was supported by the National High Level Hospital Clinical Research Funding (No. BJ-2021-187). Author contributions Study conception and design: C Li, M Wu and Y Wang. Acquisition of data: C Li and Y Liu. Analysis and interpretation of data: C Li and HZ Wang. Drafting of the manuscript: C Li and HZ Wang. Critical revision of the manuscript: C Li and WD Dai. All authors have read and approved the manuscript. Ethics declarations Ethics approval and consent to participate The study was approved by The Ethics Committees of Beijing Hospital (ethics approval letter no. 2022BJYYEC-029-02). References H. Sung, J. Ferlay, R.L. Siegel, M. Laversanne, I. Soerjomataram, A. Jemal, F. Bray, Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries, CA Cancer J. Clin. 71 (2021) 209–249. https://doi.org/10.3322/caac.21660. P. Konyn, A. Ahmed, D. Kim, Current epidemiology in hepatocellular carcinoma, Expert Rev. Gastroenterol. Hepatol. 15 (2021) 1295–1307. https://doi.org/10.1080/17474124.2021.1991792. A. Forner, M. Gilabert, J. Bruix, J.L. Raoul, Treatment of intermediate-stage hepatocellular carcinoma, Nat. Rev. Clin. Oncol. 11 (2014) 525–535. https://doi.org/10.1038/nrclinonc.2014.122. A.W.H. Chan, J. Zhong, S. Berhane, H. Toyoda, A. Cucchetti, K. Shi, T. Tada, C.C.N. Chong, B.D. Xiang, L.Q. Li, P.B.S. Lai, V. Mazzaferro, M. García-Fiñana, M. Kudo, T. Kumada, S. Roayaie, P.J. Johnson, Development of pre and post-operative models to predict early recurrence of hepatocellular carcinoma after surgical resection, J. Hepatol. 69 (2018) 1284–1293. https://doi.org/10.1016/j.jhep.2018.08.027. L. Di Tommaso, M. Roncalli, Tissue biomarkers in hepatocellular tumors: which, when, and how, Front. Med. (Lausanne). 4 (2017) 10. https://doi.org/10.3389/fmed.2017.00010. S. Ning, C. Bin, H. Na, S. Peng, D. Yi, Y. Xiang-hua, Z. Fang-yin, Z. Da-yong, L. Rong-cheng, Glypican-3, a novel prognostic marker of hepatocellular cancer, is related with postoperative metastasis and recurrence in hepatocellular cancer patients, Mol. Biol. Rep. 39 (2012) 351–357. https://doi.org/10.1007/s11033-011-0745-y. European Association for the Study of the Liver. Electronic address: [email protected] , Corrigendum to “EASL Clinical Practice Guidelines: management of hepatocellular carcinoma” [J Hepatol 69 (2018) 182-236], J. Hepatol. 70 (2019) 817. https://doi.org/10.1016/j.jhep.2019.01.020. E. Scalco, G. Rizzo, Texture analysis of medical images for radiotherapy applications, Br. J. Radiol. 90 (2017) 20160642. https://doi.org/10.1259/bjr.20160642. H.J. Aerts, E.R. Velazquez, R.T. Leijenaar, C. Parmar, P. Grossmann, S. Carvalho, J. Bussink, R. Monshouwer, B. Haibe-Kains, D. Rietveld, F. Hoebers, M.M. Rietbergen, C.R. Leemans, A. Dekker, J. Quackenbush, R.J. Gillies, P. Lambin, Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach, Nat. Commun. 5 (2014) 4006. https://doi.org/10.1038/ncomms5006. B. van Ginneken, Fifty years of computer analysis in chest imaging: rule-based, machine learning, deep learning, Radiol. Phys. Technol. 10 (2017) 23–32. https://doi.org/10.1007/s12194-017-0394-5. C.F. Dietrich, C.P. Nolsøe, R.G. Barr, A. Berzigotti, P.N. Burns, V. Cantisani, M.C. Chammas, N. Chaubal, B.I. Choi, D.A. Clevert, X. Cui, Y. Dong, M. D’Onofrio, J.B. Fowlkes, O.H. Gilja, P. Huang, A. Ignee, C. Jenssen, Y. Kono, M. Kudo, N. Lassau, W.J. Lee, J.Y. Lee, P. Liang, A. Lim, A. Lyshchik, M.F. Meloni, J.M. Correas, Y. Minami, F. Moriyasu, C. Nicolau, F. Piscaglia, A. Saftoiu, P.S. Sidhu, I. Sporea, G. Torzilli, X. Xie, R. Zheng, Guidelines and good clinical practice recommendations for contrast-enhanced ultrasound (CEUS) in the liver-update 2020 WFUMB in cooperation with EFSUMB, AFSUMB, AIUM, and FLAUS, Ultrasound Med. Biol. 46 (2020) 2579–2604. https://doi.org/10.1016/j.ultrasmedbio.2020.04.030. C. Li, J. Xu, Y. Liu, M. Wu, W. Dai, J. Song, H. Wang, Kupffer phase radiomics signature in Sonazoid-enhanced ultrasound is an independent and effective predictor of the pathologic grade of hepatocellular carcinoma, J. Oncol. 2022 (2022) 6123242. https://doi.org/10.1155/2022/6123242. Y. Dong, D. Zuo, Y.J. Qiu, J.Y. Cao, H.Z. Wang, W.P. Wang, Prediction of histological grades and Ki-67 expression of hepatocellular carcinoma based on Sonazoid contrast enhanced ultrasound radiomics signatures, Diagnostics (Basel). 12 (2022) 2175. https://doi.org/10.3390/diagnostics12092175. Y. Dong, D. Zuo, Y.J. Qiu, J.Y. Cao, H.Z. Wang, L.Y. Yu, W.P. Wang, Preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma based on Kupffer phase radiomics features of Sonazoid contrast-enhanced ultrasound (SCEUS): A prospective study, Clin. Hemorheol. Microcirc. 81 (2022) 97–107. https://doi.org/10.3233/CH-211363. L. Zhou, J.A. Rui, W.X. Zhou, S.B. Wang, S.G. Chen, Q. Qu, Edmondson-Steiner grade: A crucial predictor of recurrence and survival in hepatocellular carcinoma without microvascular invasion, Pathol. Res. Pract. 213 (2017) 824–830. https://doi.org/10.1016/j.prp.2017.03.002. H. Schmilovitz-Weiss, A. Tobar, M. Halpern, I. Levy, E. Shabtai, Z. Ben-Ari, Tissue expression of squamous cellular carcinoma antigen and Ki67 in hepatocellular carcinoma-correlation with prognosis: a historical prospective study, Diagn. Pathol. 6 (2011) 121. https://doi.org/10.1186/1746-1596-6-121. J.R. Landis, G.G. Koch, The measurement of observer agreement for categorical data, Biometrics. 33 (1977) 159–174. https://doi.org/10.2307/2529310. P. Lambin, E. Rios-Velazquez, R. Leijenaar, S. Carvalho, R.G. van Stiphout, P. Granton, C.M. Zegers, R. Gillies, R. Boellard, A. Dekker, H.J. Aerts, Radiomics: extracting more information from medical images using advanced feature analysis, Eur. J. Cancer. 48 (2012) 441–446. https://doi.org/10.1016/j.ejca.2011.11.036. R.J. Gillies, P.E. Kinahan, H.H. Hricak, Radiomics: images are more than pictures, they are data, Radiology. 278 (2016) 563–577. https://doi.org/10.1148/radiol.2015151169. M.A. Abd El Aziz, A. Facciorusso, T. Nayfeh, S. Saadi, M. Elnaggar, C. Cotsoglou, R. Sacco, Immune checkpoint inhibitors for unresectable hepatocellular carcinoma, Vaccines (Basel). 8 (2020) 616. https://doi.org/10.3390/vaccines8040616. A. Facciorusso, M.A. Abd El Aziz, R. Sacco, Efficacy of regorafenib in hepatocellular carcinoma patients: A systematic review and meta-analysis, Cancers (Basel). 12 (2019) 36. https://doi.org/10.3390/cancers12010036. P. Lambin, R.T.H. Leijenaar, T.M. Deist, J. Peerlings, E.E.C. de Jong, J. van Timmeren, S. Sanduleanu, R.T.H.M. Larue, A.J.G. Even, A. Jochems, Y. van Wijk, H. Woodruff, J. van Soest, T. Lustberg, E. Roelofs, W. van Elmpt, A. Dekker, F.M. Mottaghy, J.E. Wildberger, S. Walsh, Radiomics: the bridge between medical imaging and personalized medicine, Nat. Rev. Clin. Oncol. 14 (2017) 749–762. https://doi.org/10.1038/nrclinonc.2017.141. S.S. Yip, H.J. Aerts, Applications and limitations of radiomics, Phys. Med. Biol. 61 (2016) R150–R166. https://doi.org/10.1088/0031-9155/61/13/R150. M.G. Lubner, A.D. Smith, K. Sandrasegaran, D.V. Sahani, P.J. Pickhardt, CT texture analysis: definitions, applications, biologic correlates, and challenges, RadioGraphics. 37 (2017) 1483–1503. https://doi.org/10.1148/rg.2017170056. H. Chong, Y. Gong, X. Pan, A. Liu, L. Chen, C. Yang, M. Zeng, Peritumoral dilation radiomics of gadoxetate disodium-enhanced MRI excellently predicts early recurrence of hepatocellular carcinoma without macrovascular invasion after hepatectomy, J. Hepatocell Carcinoma. 8 (2021) 545–563. https://doi.org/10.2147/JHC.S309570. D. Gu, Y. Xie, J. Wei, W. Li, Z. Ye, Z. Zhu, J. Tian, X. Li, MRI-based radiomics signature: A potential biomarker for identifying glypican 3-positive hepatocellular carcinoma, J. Magn. Reson. Imaging. 52 (2020) 1679–1687. https://doi.org/10.1002/jmri.27199. N.Q.K. Le, Q.H. Kha, V.H. Nguyen, Y.C. Chen, S.J. Cheng, C.Y. Chen, Machine learning-based radiomics signatures for EGFR and KRAS mutations prediction in non-small-cell lung cancer, Int. J. Mol. Sci. 22 (2021) 9254. https://doi.org/10.3390/ijms22179254. N.Q.K. Le, Q.T. Ho, Deep transformers and convolutional neural network in identifying DNA N6-methyladenine sites in cross-species genomes, Methods. 204 (2022) 199–206. https://doi.org/10.1016/j.ymeth.2021.12.004. Tables Tables 1 to 9 are available in the Supplementary Files section Additional Declarations No competing interests reported. Supplementary Files Tables.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5362429","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":372654577,"identity":"85d74d54-ffda-47cd-959c-34dcb5624273","order_by":0,"name":"Chen Li","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Chen","middleName":"","lastName":"Li","suffix":""},{"id":372654578,"identity":"faa48660-3777-482d-85ca-52eedf5f5647","order_by":1,"name":"Yuan Liu","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Liu","suffix":""},{"id":372654579,"identity":"e773d9ce-3268-4175-b4e6-550432ef54cf","order_by":2,"name":"Mingxiao Wu","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Mingxiao","middleName":"","lastName":"Wu","suffix":""},{"id":372654580,"identity":"e5e7dba6-be56-4c9b-b827-8c3fd58820f0","order_by":3,"name":"Weide Dai","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Weide","middleName":"","lastName":"Dai","suffix":""},{"id":372654581,"identity":"769facc8-d2aa-43f5-b283-aceb52f4c441","order_by":4,"name":"Jinghai Song","email":"","orcid":"","institution":"Beijing Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jinghai","middleName":"","lastName":"Song","suffix":""},{"id":372654582,"identity":"6f766b93-f369-4cab-bf73-5423a1578190","order_by":5,"name":"Hanzhang Wang","email":"","orcid":"","institution":"GE healthcare","correspondingAuthor":false,"prefix":"","firstName":"Hanzhang","middleName":"","lastName":"Wang","suffix":""},{"id":372654583,"identity":"c469cbcc-5483-435a-9e66-0166e9b1784b","order_by":6,"name":"Yong Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYDADfgbGBhK1SDaQrMXgANEqj589/Jqn4o7d5uOHGz/8zKlj4G8/wPi5AJ+WM3lp1jxnniVvO5PYLNm77TCDxJkEZukZeLSYHcgxM+ZtO5xsdoOxjYF3G9B5NxjYmHnwaTn/BqLFeAZjG+PfbXUM8gS13MgxfgzUYmcgwdjGzLuNmcGAkBb7G2/MGOecOZwgAfSLtOy2wzyGIAY+LZL9OcYf3lQctudvP/7w49ttdXJyxw8f/IxPCxCwSQEVJDZAeUA24Thl/vgD6EBCqkbBKBgFo2AEAwBlPk0w1TiSxAAAAABJRU5ErkJggg==","orcid":"","institution":"National Cancer Center, Chinese Academy of Medical Sciences and Peking Union Medical College","correspondingAuthor":true,"prefix":"","firstName":"Yong","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2024-10-30 16:23:36","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5362429/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5362429/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":69447727,"identity":"70d599ba-528d-4095-8655-11aa128ce921","added_by":"auto","created_at":"2024-11-20 12:10:48","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":137989,"visible":true,"origin":"","legend":"\u003cp\u003eStudy design\u003c/p\u003e\n\u003cp\u003eStep 1: Collection and classification of immunohistochemistry.\u003c/p\u003e\n\u003cp\u003eStep 2: Workflow for radiomics analysis of grayscale and Kupffer S-CEUS images.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/29dac9b234f46b4f2a1fd0c7.jpg"},{"id":69446192,"identity":"958122d8-56d5-4bb6-851b-afc9fbd35412","added_by":"auto","created_at":"2024-11-20 12:02:48","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":106783,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for clinical, radiomics, and combined machine learning prediction of GS\u003c/p\u003e\n\u003cp\u003eROC curves for the clinical (pathological grade) classifier (a), radiomics classifier (b), and the combined clinical-radiomics classifier (c) for the multiclass prediction of GS\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/8e0b6aa487d5228ccb31b0c7.jpg"},{"id":69446191,"identity":"d1e6e32c-1819-4311-9b79-517b4e234bda","added_by":"auto","created_at":"2024-11-20 12:02:48","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":100792,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for clinical, radiomics, and combined machine learning prediction of CD10\u003c/p\u003e\n\u003cp\u003eROC curves for the clinical (clinical biomarkers TBil and DBil) classifier (a), radiomics classifier (b), and the combined clinical-radiomics classifier (c) for the multiclass prediction of CD10\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/c9679d2b81c8c126dae807bb.jpg"},{"id":69446195,"identity":"1779a337-6fab-467b-b527-da2699917aa8","added_by":"auto","created_at":"2024-11-20 12:02:48","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":96435,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for clinical, radiomics, and combined machine learning prediction of GPC3\u003c/p\u003e\n\u003cp\u003eROC curves for the clinical (pathological grade) classifier (a), radiomics classifier (b), and the combined clinical-radiomics classifier (c) for the multiclass prediction of GPC3\u003c/p\u003e","description":"","filename":"Figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/2c4bdf01be48800989e9b696.jpg"},{"id":69446193,"identity":"6a0fdd01-0d40-414a-9d14-604c66825dd6","added_by":"auto","created_at":"2024-11-20 12:02:48","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":160928,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for radiomic machine learning prediction of HSP70\u003c/p\u003e","description":"","filename":"Figure5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/18edf207ac5170d6ba1cfb81.jpg"},{"id":69447961,"identity":"1019fd8f-ba92-470b-b3bf-fc7db659a269","added_by":"auto","created_at":"2024-11-20 12:18:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1048478,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/1feba55b-8304-47c6-97c1-26e9e9b97a06.pdf"},{"id":69446190,"identity":"d597b14b-6d7e-43d0-b597-c2fbda9ccdb6","added_by":"auto","created_at":"2024-11-20 12:02:48","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":38908,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-5362429/v1/3a729eeaa5aa58f073ac4c84.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Kupffer phase radiomics signature in Sonazoid contrast-enhanced ultrasound predicts immunohistochemistry marker expression in hepatocellular carcinoma","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHepatocellular carcinoma (HCC), one of the most common primary hepatic malignant tumors, is the third leading cause of cancer-related deaths worldwide, with a 5-year survival rate of \u0026lt;\u0026thinsp;20% [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Several therapies, such as hepatic surgical resection, transplantation, and transcatheter arterial chemoembolization, are effective and commonly used to treat HCC. Among these, hepatectomy is the preferred therapeutic method for most patients with HCC. The prognosis of HCC has improved with the advances in hepatectomy and imaging technologies. However, its high potential for vascular invasion, metastasis, and recurrence post-resection leads to a poor prognosis [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eGlypican-3 (GPC3), glutamine synthetase (GS), and heat shock protein 70 (HSP70) are biomarkers currently used to discriminate the nature of hepatocellular lesions smaller than 2 cm detected in patients with liver cirrhosis, which lack the radiological features of HCC 5. CD10 is a tissue marker used to confirm the diagnosis of HCC through imaging in histology [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Among the prognostic factors for HCC, GPC3 has been closely associated with postoperative metastasis/recurrence in patients with HCC [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The expression of GPC3 in HCC is an important independent factor in predicting a patient\u0026rsquo;s poor prognosis. International guidelines recommend the combined use of these immunohistochemical (IHC) markers for a more accurate diagnosis. Compared with early HCC, assessable using this panel of markers, our clinical challenge involved distinguishing dysplastic nodules [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRadiomics analysis is an emerging medical imaging technique that has attracted considerable attention in recent years. Radiomics can be used to extract high-dimensional information that is invisible to humans [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. This feature could be used in the characterization of tumor heterogeneity, as well as reflecting the tumor tissue microenvironment and cancer phenotype [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Consequently, the radiomics signature has become a prognostic biomarker that can augment available clinical data, aid in lesion detection, improve the accuracy of diagnosis, predict the risk of disease, and determine treatment strategies [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Ultrasound, particularly Sonazoid contrast-enhanced ultrasound (S-CEUS), has been successfully applied to predict the pathological grading of tumors, evaluate malignancy, and assess treatment response with liver-specific contrast agents such as Sonazoid. This agent enables prolonged Kupffer phase imaging, more effectively characterizing focal liver lesions [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Several studies have reported that radiomics signatures derived from S-CEUS show potential in predicting microvascular invasion, histopathological grade, and Ki-67 protein expression levels in HCC [\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. However, to our knowledge, few studies have attempted to identify the potential of radiomics signatures in predicting other IHC markers, such as GPC3, GS, HSP70, and CD10. This study aimed to investigate and validate the performance of radiomics models based on the Kupffer phase of S-CEUS images in predicting the expression of IHC markers in HCC.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e\u003cstrong\u003ePatients\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthical approval was obtained from the Institutional Review Board (2021BJYYEC-190-02), which waived the requirement for informed consent. All procedures were performed in accordance with the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003eA total of 113 consecutive patients with HCC were retrospectively considered between November 2019 and May 2023. The inclusion criteria were as follows: 1) patients with a suspected diagnosis of HCC; 2) possession of complete medical information; 3) having undergone preoperative grayscale ultrasound and Kupffer phase imaging of S-CEUS; 4) no previous treatments, such as radiofrequency ablation, microwave ablation, and chemotherapy treatment; and 5) a pathologic report confirming HCC, with routine immunochemical staining for GS, CD10, and GPC3 conducted. The exclusion criteria were: 1) absence of S-CEUS; 2) incomplete clinical information; 3) lack of an available IHC report; 4) contraindication for the use of the Sonazoid contrast agent; and 5) poor B-mode or S-CEUS image quality of focal liver lesions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePotential clinical biomarkers were identified through patient interviews and a thorough review of medical records. The clinical characteristics of the patients included age, sex, maximum tumor diameter, alpha-fetoprotein levels, total bilirubin (TBil), direct bilirubin (DBil), carbohydrate antigen 12-5 (CA125), carbohydrate antigen 19-9 (CA199), carbohydrate antigen 15-3 (CA153), aspartate transaminase (AST), and alanine transaminase (ALT). Data on risk factors, including hepatitis and alcohol consumption, were collected and recorded.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHistopathological result and immunohistochemical staining procedure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHistopathological results, including the Edmondson–Steiner (E-S) grade and IHC markers, were evaluated by a pathologist with \u0026gt;10 years’ experience. The resected specimens were fixed in 10% paraformaldehyde, embedded in paraffin, and cut into 4-μm-thick sections for hematoxylin–eosin staining or IHC identification. The histological grade of HCC was determined based on the E-S grade [15]. Grades I and II were classified as low grade, whereas grades III and IV were classified as high grade. For the IHC analysis, the expression levels of GS, CD10, and GPC3 were categorized. Negative expression (no immunoreactivity, -), mildly positive expression (\u0026lt;50% immunoreactivity, +), and strongly positive expression (\u0026gt;50% immunoreactivity, ++) were identified based on the staining patterns [15,16].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSonazoid contrast-enhanced ultrasound imaging\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll patients underwent routine grayscale ultrasonography and S-CEUS. Two sonographers conducted the S-CEUS examinations using an Aplio 500 (Canon), equipped with convex (6C1, 1-6 MHz) and linear (11L4, 4-11 MHz) probes, and an Aplio i800 (Canon), equipped with convex (PVI-475BX, 1-8 MHz) and linear (11L4, 4-11 MHz) probes. According to the size of the lesion, the mechanical index of the acoustic output was set to 0.19-0.22 with a dynamic range of 65-70 dB. The patients were injected with 0.5 mL of Sonazoid through a peripheral venous line, followed by 5 mL of saline. The enhancement features were recorded and analyzed according to the latest WFUMB guidelines [11]. The Kupffer phase was obtained by scanning for 15 min. The best frames were identified from the grayscale ultrasound image and the Kupffer phase image was sampled for radiomic analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRadiomics data acquisition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn the radiomics analysis, a single cine depicting the largest cross-sectional view of the HCC lesion was selected from both B-mode and Kupffer phase ultrasound images by an ultrasound radiologist. This was stored in the Digital Imaging and Communications in Medicine format. The region of interest, which defined the tumor area, was delineated by a senior ultrasound radiologist using ITK-SNAP (an open-source software, http://www.itksnap.org). Radiomic features were extracted using NovoUltrasound Kit (NUK V1.5.0, GE HealthCare). The features considered, including shape-related, pixel intensity-based histogram, and texture features (GLCM, GLDM, GLRLM, GLSZM, and NGTDM), were defined according to the IBSI standard. Prior to feature extraction, B-mode ultrasound and S-CEUS cines were transformed to a grayscale format. Numerous filters (such as Laplacian of Gaussian, wavelet, and local binary patterns) were applied to the original images for better characterization across different spectra. The feature values were normalized based on the Z-score. Radiomic features from B-mode and Kupffer phase ultrasound images were fused into a single feature vector for analysis (Figure 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMulti-class prediction of IHC markers\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGiven the high dimensionality and inter-feature redundancy of radiomics, a feature selection process utilizing analysis of variance (ANOVA) and Spearman correlation (ANOVA–Spearman) was conducted to eliminate features with insignificant variance among IHC groups and those showing significant inter-feature correlation (Spearman’s ρ \u0026gt; 0.75). We employed the commonly used Naïve Bayes classifier for multi-classification tasks. Two models were developed: a Naïve Bayes classifier using ANOVA-selected clinical biomarkers, and a combined model using both the selected clinical biomarkers and radiomic features (Figure 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical methods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDescriptive statistics for continuous clinical biomarkers are presented either as mean ± standard deviation or median (inter-quartile range), depending on their statistical distribution determined by the Shapiro–Wilk test (\u003cem\u003eP\u003c/em\u003e value \u0026gt; 0.05 indicates normal distribution). The 95% confidence intervals (CIs) were determined via bootstrapping (1,000 resamples). Accuracy was used to provide a straightforward measure of the proportion of correct predictions. The receiver operating characteristic (ROC) and the area under the ROC (AUC) curve were utilized to assess the discriminative power of individual prediction classes. The F1 score denoted the harmonic mean of precision and recall. The macro-averaged AUC offered an aggregate perspective on discrimination performance across all classes. Cohen’s kappa measures chance correlations on a scale from 0 to 1, where 1 indicates optimal reliability, and values below 0 indicate performance inferior to randomness. It is categorized as slight, fair, moderate, substantial, or perfect [17]. The net reclassification improvement (NRI) metric was used to quantitatively evaluate the combined model’s proficiency in accurately reclassifying subjects compared with a benchmark model, specifically the clinical biomarker model in this context. An NRI exceeding 0 indicates enhanced classification. \u003cem\u003eP\u003c/em\u003e values \u0026lt;0.05 denoted statistical significance.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eA total of 86 patients with 106 histologically confirmed HCC nodules (89.6% of which are male, average age 62.02 ± 10.21 years) were consecutively enrolled in this study. The variance among different classes of the three IHC markers was measured. Of the 106 nodules, 7 (6.6%) were GS negative, 40 (37.7%) were GS mildly positive, and 59 (55.7%) were GS strongly positive. With regard to CD10 expression, 31 nodules (29.2%) displayed negative expression, 67 (63.2%) displayed mild positive expression, and 8 (7.5%) displayed strongly positive expression. Significant differences in pathological grade were observed across different classes of the IHC markers GS and GPC3 among the examined clinical biomarkers. TBil and DBil were identified as significant IHC markers for CD10 (Table 1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrediction result for the IHC marker GS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEleven radiomic features from B-mode ultrasound and 12 features from the Kupffer phase of S-CEUS were identified as relevant to GS expression by using the ANOVA–Spearman selection process (Table 2). Figure 2 presents the ROC curves for the classifiers based on clinical data (pathological grade), radiomics, and a combined clinical-radiomics approach. The combined Naïve Bayes classifier, integrating pathological grades with radiomic features, demonstrated a superior AUC compared with the classifiers using clinical or radiomics data alone. Both radiomics and combined classifiers exhibited a higher per-class AUC for discriminating between GS-negative and GS-positive (mildly positive, strongly positive) expression in patients, with AUCs of 0.870 and 0.882, respectively. The macro-averaged AUC, which represents the overall performance of both the radiomics and combined classifiers, was 0.706 (Figure 2). Thus, in terms of AUC values, the combination method was more effective in classifying GS expression than the clinical method alone. However, adding pathological grades to the radiomics analysis did not improve these results.\u003c/p\u003e\n\u003cp\u003eRegarding diagnostic performance, the combined classifier achieved a higher accuracy, F1 score, and Cohen’s kappa than the clinical and radiomics classifiers (Table 3). Both the radiomics and combined methods demonstrated fair accordance and yielded a positive NRI compared with the clinical method, with NRI values of 0.367 and 0.454, respectively. Additionally, a positive NRI value of 0.048 was observed when comparing the combined method with the radiomics classifier. These positive NRI values indicate that incorporating radiomics-based classifications improved the classification results of the clinical biomarkers alone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrediction result for the IHC marker CD10\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the IHC marker CD10, the ANOVA–Spearman selection process identified 14 radiomic features from B-mode images and four from Kupffer phase images (Table 4). The classifier based on the clinical biomarkers TBil and DBil was less effective according to the ROC analysis. In contrast, the radiomics classifier demonstrated superior discrimination performance, evidenced by the highest macro-averaged AUC value of 0.834 and the highest per-class AUC value for each class of CD10 expression level. Compared with TBil/DBil, the combined classifier showed increased performance across all measured metrics (Figure 3, Table 5). The NRI for the addition of radiomics to TBil/DBil was 0.137, indicating a positive improvement in classification results.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrediction result for the IHC marker GPC3\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEight B-mode radiomic features and six Kupffer phase features were selected for the GPC3 expression classification (Table 6). ROC analysis showed that the radiomics classifier was the most adequate, with the highest macro-average AUC value of 0.700. The radiomics classifier achieved similar results to the clinical classifier in discriminating GPC3 positive/negative patients (per-class AUC 0.637 versus 0.640). However, it was more effective in classifying positive patients into mild or strong expression categories (Figure 4). While combining clinical and radiomic features did not improve the AUC results compared with radiomics alone, this approach was noticeably better than using the clinical classifier alone.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe clinical classifier demonstrated an inferior diagnostic value, as its classification results showed only marginal concordance with the actual expression levels. By contrast, both the radiomics classifier and combined methods displayed fair agreement with the actual values, with the radiomics classifier maintaining the highest metric values (Table 7). While the combined classifier improved the classification results of the clinical classifier (NRI = 0.134), it did not improve the classification results of the radiomics classifier (NRI = -0.007).\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003ePrediction result for the IHC marker HSP70\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eAn ANOVA conducted on the clinicopathological variables revealed no significant differences between the HSP70 expression groups. Fourteen radiomic features (seven from each ultrasound mode) were related to HSP70 (Table 8). A Naïve Bayes classifier, utilizing these radiomic features, was constructed based on the diagnostic performances listed below (Table 9, Figure 5).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe Naïve Bayes classifier, based on ultrasound radiomic features, demonstrated negative (-) and positive (+, ++, or +++) features with excellent performance (per-class AUC 0.880). However, its per-class discrimination ability diminished as the expression levels increased, with an AUC of only 0.694 in distinguishing (+++) features from those in the rest of the specimens.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWith the development of precision medicine, identifying new quantitative and radiomics-based noninvasive imaging biomarkers aimed at improving the predictive performance of medical images has become an area of interest in radiological research [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. In recent years, with the development of new HCC treatment methods, such as immune checkpoint inhibitors [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] and regorafenib [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], radiomics methods, including texture analysis, shape, and intensity features, have shown potential relevance between medical imaging and personalized medicine [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The aim of our study was to propose and validate radiomics models based on the Kupffer phase of S-CEUS images for the potential prediction of IHC markers. The data showed that the radiomics model originating from S-CEUS images achieved promising results, suggesting its feasibility and applicability in clinical practice.\u003c/p\u003e \u003cp\u003eIn this retrospective study, we extracted radiomic features from the B-mode and Kupffer phases associated with IHC biomarkers, which is also the core principle of radiomics, and selected these data for radiomics score construction [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. The developed radiomics model showed a mathematical correlation between ultrasound radiomics characteristics and IHC biomarkers. Utilizing the unique tissue features of the Kupffer phase, the CE-mode was constructed using quantitative radiomic characteristics from the S-CEUS Kupffer phase images to distinguish the stages of HCC in IHC biomarkers [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. In our study, the classifier that combined pathological grades of clinical risk factors with the radiomics signature was more effective at classifying GS and GPC3 expression than either the clinical or radiomics classifier alone. While the combination of clinical and radiomic features did not improve the AUC results compared with radiomics alone, it was noticeably better than the clinical classifier alone. The combined classifier improved the classification results of the clinical classifier, but it did not improve those of the radiomics classifier. The radiomics classifier demonstrated superior discrimination performance for each class of CD10 expression level. When compared with the TBil/DBil of clinical risk factors, the combined classifier showed more competitive performance across all measured metrics. The radiomics classifier identified negative and positive results with excellent performance for HSP70 expression levels (per-class AUC 0.880). The combined classifier achieved optimal predictive performance. This is attributed to the comprehensive overview provided by both qualitative and quantitative imaging features in the final model, elucidating the relationship between radiomic characteristics and the pathological status of HCC [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Radiomic features, including shape, intensity, and texture information, can present the complexity of the target tissue properties. Typically, the heterogeneity exhibited by malignant lesions is meticulously captured and quantified through subjective interpretations by experienced radiologists. This heterogeneity spans multiple domains (e.g., gross, cellular, genetic, and phenotypic levels) and dimensions (e.g., cellular density, angiogenesis, hemorrhage, and necrosis) [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Our tumor surrounding dilation radiomics provides a comprehensive and microscopic high-throughput prediction just like \u0026ldquo;virtual tissue pathology [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u0026rdquo;.\u003c/p\u003e \u003cp\u003ePrevious studies have shown that the imaging features constructed from the Kupffer phase of S-CEUS images, including textural features, may predict Ki-67 expression in HCC [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. An effective tool is provided for the non-invasive and individualized prediction of GPC3-positive HCC through a predictive nomographic chart combining alpha-fetoprotein and radiomic features [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. However, to the best of our knowledge, no other studies have investigated the potential value of integrating the quantitative analysis of clinical factors in predicting IHC biomarkers, such as GPC3, HSP70, GS, and CD10, in HCC. Our study investigated the prediction of IHC markers using radiomic feature models based on both the B-mode and Kupffer phases, which can only be achieved through invasive biopsy or surgery. These data indicate that radiomics models derived from Kupffer phase S-CEUS images achieved good results and are deemed feasible for application in clinical practice. The complicated correlation between bioprocesses and radiomic signatures poses a notable challenge, aligning with the current trend in precision and individualized medicine [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study has several limitations. First, as this was a single-center study, external validation through additional multicenter cohorts may be required to verify the reliability and assess the generalizability of our results. Second, the results of our study on radiomic features from the arterial mode were unsatisfactory, leading to the dismissal of these results. Further investigation into the diagnostic value of dynamic graphs, including the arterial, portal venous, late phase, and Kupffer phase of SCEUS, is warranted. Finally, the absence of multimodal radiological data in this study should be noted. The potential use of S-CEUS combined with gadoxetic acid-enhanced magnetic resonance imaging and computed tomography may provide additional properties to enhance prediction accuracy. In the future, the multimodality of ultrasound imaging, including color Doppler flow imaging, ultrasound elastography, and the vascular phase of CEUS imaging, combined with magnetic resonance imaging and computed tomography data, could be used to explore the predictive performance of IHC biomarkers.\u003c/p\u003e \u003cp\u003eIn conclusion, the Kupffer phase in the S-CEUS-based radiomics signature is an excellent biomarker. It has achieved desirable results in predicting the IHC expression of GPC3, GS, HSP70, and CD10 in patients with HCC. The combined clinical factors and radiomics signature may provide an effective tool for the noninvasive and individualized prediction of immunohistochemistry expression in HCC.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e\u003cstrong\u003eHCC,\u0026nbsp;\u003c/strong\u003ehepatocellular carcinoma; \u003cstrong\u003eGPC3,\u0026nbsp;\u003c/strong\u003eglypican-3; \u003cstrong\u003eGS,\u003c/strong\u003e glutamine synthetase; \u003cstrong\u003eHSP70,\u003c/strong\u003e heat shock protein 70; \u003cstrong\u003eIHC,\u003c/strong\u003e immunohistochemical; \u003cstrong\u003eS-CEUS,\u003c/strong\u003e Sonazoid contrast-enhanced ultrasound; \u003cstrong\u003eTBil,\u003c/strong\u003e total bilirubin; \u003cstrong\u003eDBil,\u003c/strong\u003e direct bilirubin; \u003cstrong\u003eCA125,\u003c/strong\u003e carbohydrate antigen 12-5; \u003cstrong\u003eCA199,\u003c/strong\u003e carbohydrate antigen 19-9; \u003cstrong\u003eCA153,\u003c/strong\u003e carbohydrate antigen 15-3;\u003cstrong\u003e\u0026nbsp;AST,\u003c/strong\u003e aspartate transaminase; \u003cstrong\u003eALT,\u003c/strong\u003e alanine transaminase; \u003cstrong\u003eE-S,\u0026nbsp;\u003c/strong\u003eEdmondson–Steiner; \u003cstrong\u003eANOVA,\u003c/strong\u003e analysis of variance; \u003cstrong\u003eANOVA-Spearman,\u003c/strong\u003e Spearman correlation of ANOVA; \u003cstrong\u003eROC,\u003c/strong\u003e receiver operating characteristic; \u003cstrong\u003eAUC,\u003c/strong\u003e area under the ROC; \u003cstrong\u003eNRI,\u003c/strong\u003e net reclassification improvement\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National High Level Hospital Clinical Research Funding (No. BJ-2021-187).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStudy conception and design: C Li,\u0026nbsp;M Wu\u0026nbsp;and Y Wang. Acquisition of data: C Li and Y Liu. Analysis and interpretation of data: C Li and\u0026nbsp;HZ Wang. Drafting of the manuscript: C Li and\u0026nbsp;HZ Wang. Critical revision of the manuscript: C Li and\u0026nbsp;WD Dai. All authors have read and approved the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was approved by The Ethics Committees of Beijing Hospital (ethics approval letter no. 2022BJYYEC-029-02).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eH. Sung, J. Ferlay, R.L. Siegel, M. Laversanne, I. Soerjomataram, A. Jemal, F. Bray, Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries, CA Cancer J. Clin. 71 (2021) 209\u0026ndash;249. https://doi.org/10.3322/caac.21660.\u003c/li\u003e\n \u003cli\u003eP. Konyn, A. Ahmed, D. Kim, Current epidemiology in hepatocellular carcinoma, Expert Rev. Gastroenterol. Hepatol. 15 (2021) 1295\u0026ndash;1307. https://doi.org/10.1080/17474124.2021.1991792.\u003c/li\u003e\n \u003cli\u003eA. Forner, M. Gilabert, J. Bruix, J.L. Raoul, Treatment of intermediate-stage hepatocellular carcinoma, Nat. Rev. Clin. Oncol. 11 (2014) 525\u0026ndash;535. https://doi.org/10.1038/nrclinonc.2014.122.\u003c/li\u003e\n \u003cli\u003eA.W.H. Chan, J. Zhong, S. Berhane, H. Toyoda, A. Cucchetti, K. Shi, T. Tada, C.C.N. Chong, B.D. Xiang, L.Q. Li, P.B.S. Lai, V. Mazzaferro, M. Garc\u0026iacute;a-Fi\u0026ntilde;ana, M. Kudo, T. Kumada, S. Roayaie, P.J. Johnson, Development of pre and post-operative models to predict early recurrence of hepatocellular carcinoma after surgical resection, J. Hepatol. 69 (2018) 1284\u0026ndash;1293. https://doi.org/10.1016/j.jhep.2018.08.027.\u003c/li\u003e\n \u003cli\u003eL. Di Tommaso, M. Roncalli, Tissue biomarkers in hepatocellular tumors: which, when, and how, Front. Med. (Lausanne). 4 (2017) 10. https://doi.org/10.3389/fmed.2017.00010.\u003c/li\u003e\n \u003cli\u003eS. Ning, C. Bin, H. Na, S. Peng, D. Yi, Y. Xiang-hua, Z. Fang-yin, Z. Da-yong, L. Rong-cheng, Glypican-3, a novel prognostic marker of hepatocellular cancer, is related with postoperative metastasis and recurrence in hepatocellular cancer patients, Mol. Biol. Rep. 39 (2012) 351\u0026ndash;357. https://doi.org/10.1007/s11033-011-0745-y.\u003c/li\u003e\n \u003cli\u003eEuropean Association for the Study of the Liver. Electronic address: [email protected], Corrigendum to \u0026ldquo;EASL Clinical Practice Guidelines: management of hepatocellular carcinoma\u0026rdquo; [J Hepatol 69 (2018) 182-236], J. Hepatol. 70 (2019) 817. https://doi.org/10.1016/j.jhep.2019.01.020.\u003c/li\u003e\n \u003cli\u003eE. Scalco, G. Rizzo, Texture analysis of medical images for radiotherapy applications, Br. J. Radiol. 90 (2017) 20160642. https://doi.org/10.1259/bjr.20160642.\u003c/li\u003e\n \u003cli\u003eH.J. Aerts, E.R. Velazquez, R.T. Leijenaar, C. Parmar, P. Grossmann, S. Carvalho, J. Bussink, R. Monshouwer, B. Haibe-Kains, D. Rietveld, F. Hoebers, M.M. Rietbergen, C.R. Leemans, A. Dekker, J. Quackenbush, R.J. Gillies, P. Lambin, Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach, Nat. Commun. 5 (2014) 4006. https://doi.org/10.1038/ncomms5006.\u003c/li\u003e\n \u003cli\u003eB. van Ginneken, Fifty years of computer analysis in chest imaging: rule-based, machine learning, deep learning, Radiol. Phys. Technol. 10 (2017) 23\u0026ndash;32. https://doi.org/10.1007/s12194-017-0394-5.\u003c/li\u003e\n \u003cli\u003eC.F. Dietrich, C.P. Nols\u0026oslash;e, R.G. Barr, A. Berzigotti, P.N. Burns, V. Cantisani, M.C. Chammas, N. Chaubal, B.I. Choi, D.A. Clevert, X. Cui, Y. Dong, M. D\u0026rsquo;Onofrio, J.B. Fowlkes, O.H. Gilja, P. Huang, A. Ignee, C. Jenssen, Y. Kono, M. Kudo, N. Lassau, W.J. Lee, J.Y. Lee, P. Liang, A. Lim, A. Lyshchik, M.F. Meloni, J.M. Correas, Y. Minami, F. Moriyasu, C. Nicolau, F. Piscaglia, A. Saftoiu, P.S. Sidhu, I. Sporea, G. Torzilli, X. Xie, R. Zheng, Guidelines and good clinical practice recommendations for contrast-enhanced ultrasound (CEUS) in the liver-update 2020 WFUMB in cooperation with EFSUMB, AFSUMB, AIUM, and FLAUS, Ultrasound Med. Biol. 46 (2020) 2579\u0026ndash;2604. https://doi.org/10.1016/j.ultrasmedbio.2020.04.030.\u003c/li\u003e\n \u003cli\u003eC. Li, J. Xu, Y. Liu, M. Wu, W. Dai, J. Song, H. Wang, Kupffer phase radiomics signature in Sonazoid-enhanced ultrasound is an independent and effective predictor of the pathologic grade of hepatocellular carcinoma, J. Oncol. 2022 (2022) 6123242. https://doi.org/10.1155/2022/6123242.\u003c/li\u003e\n \u003cli\u003eY. Dong, D. Zuo, Y.J. Qiu, J.Y. Cao, H.Z. Wang, W.P. Wang, Prediction of histological grades and Ki-67 expression of hepatocellular carcinoma based on Sonazoid contrast enhanced ultrasound radiomics signatures, Diagnostics (Basel). 12 (2022) 2175. https://doi.org/10.3390/diagnostics12092175.\u003c/li\u003e\n \u003cli\u003eY. Dong, D. Zuo, Y.J. Qiu, J.Y. Cao, H.Z. Wang, L.Y. Yu, W.P. Wang, Preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma based on Kupffer phase radiomics features of Sonazoid contrast-enhanced ultrasound (SCEUS): A prospective study, Clin. Hemorheol. Microcirc. 81 (2022) 97\u0026ndash;107. https://doi.org/10.3233/CH-211363.\u003c/li\u003e\n \u003cli\u003eL. Zhou, J.A. Rui, W.X. Zhou, S.B. Wang, S.G. Chen, Q. Qu, Edmondson-Steiner grade: A crucial predictor of recurrence and survival in hepatocellular carcinoma without microvascular invasion, Pathol. Res. Pract. 213 (2017) 824\u0026ndash;830. https://doi.org/10.1016/j.prp.2017.03.002.\u003c/li\u003e\n \u003cli\u003eH. Schmilovitz-Weiss, A. Tobar, M. Halpern, I. Levy, E. Shabtai, Z. Ben-Ari, Tissue expression of squamous cellular carcinoma antigen and Ki67 in hepatocellular carcinoma-correlation with prognosis: a historical prospective study, Diagn. Pathol. 6 (2011) 121. https://doi.org/10.1186/1746-1596-6-121.\u003c/li\u003e\n \u003cli\u003eJ.R. Landis, G.G. Koch, The measurement of observer agreement for categorical data, Biometrics. 33 (1977) 159\u0026ndash;174. https://doi.org/10.2307/2529310.\u003c/li\u003e\n \u003cli\u003eP. Lambin, E. Rios-Velazquez, R. Leijenaar, S. Carvalho, R.G. van Stiphout, P. Granton, C.M. Zegers, R. Gillies, R. Boellard, A. Dekker, H.J. Aerts, Radiomics: extracting more information from medical images using advanced feature analysis, Eur. J. Cancer. 48 (2012) 441\u0026ndash;446. https://doi.org/10.1016/j.ejca.2011.11.036.\u003c/li\u003e\n \u003cli\u003eR.J. Gillies, P.E. Kinahan, H.H. Hricak, Radiomics: images are more than pictures, they are data, Radiology. 278 (2016) 563\u0026ndash;577. https://doi.org/10.1148/radiol.2015151169.\u003c/li\u003e\n \u003cli\u003eM.A. Abd El Aziz, A. Facciorusso, T. Nayfeh, S. Saadi, M. Elnaggar, C. Cotsoglou, R. Sacco, Immune checkpoint inhibitors for unresectable hepatocellular carcinoma, Vaccines (Basel). 8 (2020) 616. https://doi.org/10.3390/vaccines8040616.\u003c/li\u003e\n \u003cli\u003eA. Facciorusso, M.A. Abd El Aziz, R. Sacco, Efficacy of regorafenib in hepatocellular carcinoma patients: A systematic review and meta-analysis, Cancers (Basel). 12 (2019) 36. https://doi.org/10.3390/cancers12010036.\u003c/li\u003e\n \u003cli\u003eP. Lambin, R.T.H. Leijenaar, T.M. Deist, J. Peerlings, E.E.C. de Jong, J. van Timmeren, S. Sanduleanu, R.T.H.M. Larue, A.J.G. Even, A. Jochems, Y. van Wijk, H. Woodruff, J. van Soest, T. Lustberg, E. Roelofs, W. van Elmpt, A. Dekker, F.M. Mottaghy, J.E. Wildberger, S. Walsh, Radiomics: the bridge between medical imaging and personalized medicine, Nat. Rev. Clin. Oncol. 14 (2017) 749\u0026ndash;762. https://doi.org/10.1038/nrclinonc.2017.141.\u003c/li\u003e\n \u003cli\u003eS.S. Yip, H.J. Aerts, Applications and limitations of radiomics, Phys. Med. Biol. 61 (2016) R150\u0026ndash;R166. https://doi.org/10.1088/0031-9155/61/13/R150.\u003c/li\u003e\n \u003cli\u003eM.G. Lubner, A.D. Smith, K. Sandrasegaran, D.V. Sahani, P.J. Pickhardt, CT texture analysis: definitions, applications, biologic correlates, and challenges, RadioGraphics. 37 (2017) 1483\u0026ndash;1503. https://doi.org/10.1148/rg.2017170056.\u003c/li\u003e\n \u003cli\u003eH. Chong, Y. Gong, X. Pan, A. Liu, L. Chen, C. Yang, M. Zeng, Peritumoral dilation radiomics of gadoxetate disodium-enhanced MRI excellently predicts early recurrence of hepatocellular carcinoma without macrovascular invasion after hepatectomy, J. Hepatocell Carcinoma. 8 (2021) 545\u0026ndash;563. https://doi.org/10.2147/JHC.S309570.\u003c/li\u003e\n \u003cli\u003eD. Gu, Y. Xie, J. Wei, W. Li, Z. Ye, Z. Zhu, J. Tian, X. Li, MRI-based radiomics signature: A potential biomarker for identifying glypican 3-positive hepatocellular carcinoma, J. Magn. Reson. Imaging. 52 (2020) 1679\u0026ndash;1687. https://doi.org/10.1002/jmri.27199.\u003c/li\u003e\n \u003cli\u003eN.Q.K. Le, Q.H. Kha, V.H. Nguyen, Y.C. Chen, S.J. Cheng, C.Y. Chen, Machine learning-based radiomics signatures for EGFR and KRAS mutations prediction in non-small-cell lung cancer, Int. J. Mol. Sci. 22 (2021) 9254. https://doi.org/10.3390/ijms22179254.\u003c/li\u003e\n \u003cli\u003eN.Q.K. Le, Q.T. Ho, Deep transformers and convolutional neural network in identifying DNA N6-methyladenine sites in cross-species genomes, Methods. 204 (2022) 199\u0026ndash;206. https://doi.org/10.1016/j.ymeth.2021.12.004.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 9 are available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"hepatocellular carcinoma, contrast-enhanced ultrasound, radiomics, immunohistochemical markers","lastPublishedDoi":"10.21203/rs.3.rs-5362429/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5362429/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose: \u003c/strong\u003eFew studies have explored the value of radiomics signatures in predicting immunohistochemical (IHC) staining markers. This study aimed to investigate and validate radiomics models based on the Kupffer phase of Sonazoid contrast-enhanced intraoperative ultrasonography (S-CEUS) images for predicting IHC marker expression in hepatocellular carcinoma (HCC).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethod: \u003c/strong\u003eOverall, 113 consecutive patients diagnosed with HCC between November 2019 and May 2023 were retrospectively analyzed. Histopathological assessment included IHC staining for GS, CD10, GPC3, and HSP70. Radiomic features extracted from S-CEUS images were selected and analyzed. A Naïve Bayes classifier was employed to predict IHC marker expression in HCC, using selected clinical biomarkers and radiomic features.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eFor GPC3, the radiomics classifier achieved a macro-average area under the receiver operating characteristic curve (AUC) of 0.700, indicating strong performance. For GS, both radiomics and combined clinical-radiomics classifiers exhibited strong discrimination (AUCs: 0.870 and 0.882, respectively). The radiomics classifier outperformed clinical biomarkers (total and direct bilirubin) in predicting CD10, with a macro-average AUC of 0.834. However, its accuracy decreased for higher HSP70 marker expression levels (AUC: 0.694). These findings underscore the consistent effectiveness of radiomics across different IHC markers when compared to traditional clinical approaches.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions: \u003c/strong\u003eThe Kupffer phase in the S-CEUS-based radiomics signature is an excellent biomarker for predicting IHC marker expression in patients with HCC.\u003c/p\u003e","manuscriptTitle":"Kupffer phase radiomics signature in Sonazoid contrast-enhanced ultrasound predicts immunohistochemistry marker expression in hepatocellular carcinoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-20 12:02:43","doi":"10.21203/rs.3.rs-5362429/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8cd6adc7-788f-423f-a1d6-f90d3ef6a7f9","owner":[],"postedDate":"November 20th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-12-07T10:23:21+00:00","versionOfRecord":[],"versionCreatedAt":"2024-11-20 12:02:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5362429","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5362429","identity":"rs-5362429","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}