Assessment of Inter-observer Reproducibility of Radiomic Features in Multiparametric Magnetic Resonance Imaging of Endometrial Cancer

Assessment of Inter-observer Reproducibility of Radiomic Features in Multiparametric Magnetic Resonance Imaging of Endometrial Cancer | 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 Assessment of Inter-observer Reproducibility of Radiomic Features in Multiparametric Magnetic Resonance Imaging of Endometrial Cancer Fazılhan Altıntaş, Mehmet Ali Gültekin, Ahmet Akçay, Hüseyin Toprak, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8184695/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 Objective: The aim of our study is to evaluate the inter-observer agreement of radiomic features extracted from different sequences of multiparametric magnetic resonance imaging (mpMRI) in endometrial cancer, thereby contributing to the development of radiomic models with higher reproducibility and facilitating their integration into clinical practice. Materials and Methods: This single-center, retrospective study included 59 patients with pathologically confirmed, untreated endometrial cancer. The MRI scans of the patients were obtained from both in-house and external Picture Archiving and Communication System databases. Three radiologists segmented endometrial cancer tissue on three plane T2 weighted images, sagittal (T2WS), coronal (T2WC) and axial (T2WA), T1 weighted contrast enhanced sagittal (T1CE) images and diffusion weighted (DWI) images with the highest b-value. Apparent diffusion coefficient (ADC) masks were transferred from DWI segmentations. Results: The percentage of radiomic features with excellent agreement (ICC ≥ 0.90) was 58.9% for T2WS, 54.2% for T2WC, 44.9% for T2WA, 57.9% for T1CE, 54.2% for DWI, and 45.8% for ADC. NGTDM and GLCM performed more poorly than the other subclasses, 30% and 45.8% of features being reproducible, respectively. Shape and GLDM performed better than other subclasses, 60.7% and 64.3% of features being reproducible, respectively. Segmentations from T2WS had better reproducibility despite similar levels of Dice Correlation Similarity (DSC) compared to other planes. Conclusion: Among T2-weighted planes, the highest inter-observer agreement was observed in the sagittal plane (T2WS), while the lowest was found in the axial plane (T2WA). Among all sequences, T2WS and T1CE showed the highest agreement, whereas T2WA and ADC demonstrated the lowest. In the subclasses of radiomic features, the highest agreement is observed in Shape and GLDM, while the lowest agreement is seen in NGTDM and GLCM. Overall robustness of MRI derived radiomics features to inter-reader segmentation differences in EC was moderate. Endometrial Cancer MRI Radiomics Reproducibility Segmentation Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Endometrial cancer (EC) is the sixth most common cancer among women and the most frequently occurring cancer of the female reproductive system. Worldwide, in 2020 alone, 417,000 new cases were diagnosed [ 1 ]. The potential of radiomics has been investigated by several studies to assist in detecting prognostic factors such as myometrial invasion or lymph node metastasis [ 2 ]. Radiomics is the process of extracting high-throughput quantitative data from medical images using computational analysis, generating objective features beyond what the human eye can assess, such as tumor biology [ 3 ]. However despite an ever increasing number of radiomics models, clinical use of them has been very limited, due to poor reproducibility of the models [ 4 ]. The reproducibility problem arises due to use of features prone to variability. Since creation of radiomic models is a multi-step process, variations introduced at different stages—such as image acquisition, tumor annotation, feature extraction, exploratory analysis and model creation—can affect feature stability and limit the models performance when applied to new datasets, making reproducibility a major challenge in radiomics research [ 5 ]. Manual segmentation is the most used segmentation method in EC radiomics studies and inter-observer segmentation variations are one possible cause of reproducibility problems in EC radiomics models. Contrast-enhanced pelvic MRI is considered the gold standard imaging modality in pretreatment staging of EC, due to its high soft tissue resolution [ 6 ]. In addition to conventional sequences that provide detailed information on tumor location, extent, contours, depth, and lymph node involvement, the use of dynamic contrast-enhanced imaging and diffusion-weighted imaging—which measures the diffusivity of water molecules within the tumor—enables more comprehensive characterization of tumor tissue. Though a multitude of radiomics models on MRI derived features has been created [ 2 ], the reproducibility of MRI derived radiomics features following multi-observer manual segmentation of EC has not been assessed yet, as far as we know. To address this gap in the literature, the goal of this study is to evaluate robustness of MRI derived radiomics features of EC across sequences, imaging orientation and feature classes. 2. Material and Method 2.1. Patient Selection This retrospective study was approved by the Institutional Ethics Board (E-54022451-050.04-168547). Therefore informed consent requirements were waived. Pathology reports and patient information were available from medical reports. Between 2018 February − 2025 February, EC patients who underwent magnetic resonance imaging were selected according to the following criteria: (1) All three planes were acquired in T2 Weighted Imaging (2) Patients received no treatment before an MRI was acquired. If following criteria was met, patients were excluded: (1) Artefacts reducing image quality, (2) Tumors smaller than 1 cm or not visible at all, (3) Extensive hemorrhage impairing tumor evaluation. According to these criteria, 59 patients were deemed eligible for the study. The selection process was summarized in Fig. 1 . 2.2. Image Acquisition Pretreatment MRIs were collected from the picture archiving and communication system (PACS:FUJIPACS LTD). Because this study was conducted in a university hospital, some of the patients have undergone MRI in the referring hospitals and their images were already uploaded in our database. In total 7 different MRI models from 2 vendors have been used. Detailed information about scanners and acquisition parameters are available in Table 1. The ADC maps used in the study were derived directly from the original MRI scans; no standardized reconstruction of new ADC maps was performed, in order to reflect real-world clinical practice. 2.3. Segmentations The relevant volumes (VOI – Volumes of Interest) were manually segmented using the open-source software 3D Slicer (version 5.6.2)[ 7 ]. Two experienced radiologists specialized in abdominopelvic radiology (with 7 and 12 years of experience, respectively – AA and MAG) and a radiology resident with 4 years of experience (FA) independently segmented the EC regions. The radiology resident received assistance from another specialist with 15 years of experience in abdominal radiology (HT) when needed. The readers carefully delineated the EC regions on T2-weighted sagittal (T2WS), coronal (T2WC), axial (T2WA),T1-weighted contrast-enhanced sagittal (T1CE), and diffusion-weighted imaging (DWI) sequences, avoiding other possible surrounding structures such as fibroids or implants. For diffusion measurements, the highest b-value was used. The segmentation masks performed on the diffusion-weighted images (DWI) were transferred as-is to the ADC maps. Examples of the segmentations are displayed in Fig. 2 . 2.4. Radiomics For each segmented sequence, 107 radiomics features were extracted using Pyradiomics (v 3.0.1), reaching a sum of 642 features for each case [ 8 ]. No wavelet features were extracted. Features were extracted from seven distinct classes: First Order (n = 18), Gray Level Dependence Matrix (GLDM, n = 14), Gray Level Co-occurrence Matrix (GLCM, n = 24), Gray Level Run Length Matrix (GLRLM, n = 16), Gray Level Size Zone Matrix (GLSZM, n = 16), Neighbouring Gray Tone Difference Matrix (NGTDM, n = 5), and Shape features (n = 14). A bin width of 18 was used during the extraction process. To prevent any influence on inter-observer variability, no filters or normalization methods were applied. 2.4. Statistical Analysis For each case and reader pair a Dice Similarity Coefficient (DSC) was calculated, in order to compare similarity of delineation areas. The agreement between observers was assessed using the Intraclass Correlation Coefficient (ICC), applying a two-way random effects model with a single-measurement and absolute agreement approach. Absolute agreement was chosen over consistency because using the consistency definition might compromise the reliability of features, especially in scenarios that involve applying cutoff values. Following the interpretation guidelines by Koo et al., agreement levels were classified as poor for ICC ≤ 0.5, moderate for ICC between 0.50 and 0.75, good for ICC between 0.75 and 0.90, and excellent for ICC ≥ 0.9 [ 9 ]. In this study, features with ICC values of 0.9 or higher were regarded as robust across different readers. Radiomic features with an ICC of 0.9 or higher were considered resistant to inter-reader variability. ICC values were calculated for each sequence and imaging plane. Descriptive statistics were used to compare ICC differences across subgroups. Statistical analyses and relevant visualizations were performed using R programming language (R Core Team v4.4.2) on R Studio (R Studio Team v4.3.3) [ 10 ]. The “tidyverse” and “irr” packages were utilized. 3. Results 3.1. Patient Population and Segmentation Metrics In our study, preoperative imaging from 59 patients diagnosed with EC was used. The mean age was calculated as 62 years (range: 38–88). Among the 59 patients included in the study, 52 (88.1%) were diagnosed with endometrioid adenocarcinoma, 6 (10.2%) with serous adenocarcinoma, and 1 (1.7%) with clear cell carcinoma. According to the Federation of Gynecology and Obstetrics (FIGO) grading system, 34 cases (57.6%) were classified as high-grade tumors, while the remaining 25 cases (42.4%) were low-grade. Based on T2-weighted sagittal images, the mean lesion volume, averaged across all readers using the mesh volume feature, was calculated as 23,973 mm³. Using the same method, the mean short axis of the lesions was approximately 2 cm, and the mean long axis was approximately 4.5 cm. The overall mean DSC for all segmentations was 0.77. When analyzed by sequence, the mean DSCs were as follows: T2WS: 0.79 (range: 0.51–0.92, SD: 0.07), T2WC: 0.77 (range: 0.53–0.92, SD: 0.09), T2WA: 0.80 (range: 0.59–0.91, SD: 0.08), T1CE: 0.77 (range: 0.40–0.90, SD: 0.07), and DWI: 0.73 (range: 0.42–0.90, SD: 0.08). 3.2. Sequences A total of 338 out of 642 radiomic features (53%) demonstrated excellent reproducibility (ICC > 0.9) despite inter-reader segmentation variability. When the lower bound of the 95% confidence interval was considered, this proportion decreased to 240 out of 642 features (37%). Radiomic features derived from T2WS and T1CE sequences showed the highest reproducibility (58.9% and 57.9%, respectively), whereas those derived from T2WAand ADC maps showed the lowest reproducibility (44.9% and 45.8%, respectively). The median ICC values by sequence were calculated as follows: T2WS: 0.91 (range: 0.61–0.99), T2WC: 0.91 (0.60–0.99), T2WA: 0.87 (0.45–0.99), T1CE: 0.92 (0.65–0.99), DWI: 0.91 (0.33–0.99), and ADC: 0.89 (0.28–0.99). The agreement of radiomic features was visualized graphically using ICC values and their 95% confidence intervals in Fig. 3 . 3.3. Radiomics Subclasses In order to compare feature subclasses among sequences, T2WS was chosen as a representative of T2W images due to its better performance. Examining the reproducibility of radiomic feature classes, shape and GLDM features showed the highest reproducibility (60.7% and 64.3%, respectively), whereas NGTDM and GLCM features exhibited the lowest reproducibility (30% and 45.8%, respectively) (Fig. 4 ). Among the radiomic features, 19 demonstrated excellent agreement across all sequences, while 23 features showed poor agreement in all sequences (Table 2). When evaluated by feature class, none of the features from the NGTDM and First Order classes exhibited excellent agreement across all sequences (Table 2). There were 9 “HighGray” and 9 “LowGray” features for each sequence. The distribution of 'HighGray' features with excellent reproducibility across sequences was: T2WS 5/9, T2WC 9/9, T2WA 3/9, T1CE 0/9, DWI 9/9, ADC 0/9. The distribution of 'LowGray' features with excellent reproducibility across sequences was:T2WS 2/9, T2WC 0/9, T2WA 1/9, T1CE 1/9, DWI 0/9, ADC 1/9 All ICC values for each feature-sequence pair can be found in Supplementary Material A. 4. Discussion Considering the Dice Similarity Coefficients (DSC) and the agreement of shape features between reader pairs across sequences, a difference in segmentation contours among readers was observed — being the smallest in T2WA (DSC: 0.8) and the largest in DWI (DSC: 0.73). Except for DWI/ADC, no major differences were observed in the mean Dice Similarity Coefficients (DSC), which reflect the overlapping areas of the segmentations, across different sequences. We attributed the relatively lower DSC in these segmentations to the fact that most of the images in our dataset were acquired using 1.5 Tesla MRI scanners, which typically yield diffusion-weighted images with lower signal-to-noise ratios, making segmentation more challenging. This explains the lower mean DSC observed in the diffusion-weighted images. Since our study aimed to reflect a real-world radiomic modeling scenario, we chose to include cases with low DSC across sequences rather than exclude them. Our findings suggest that inter-reader segmentation differences have a significant impact on radiomic features. When evaluating the agreement of radiomic features calculated across sequences, discrepancies were observed that could not be fully explained by the mean or standard deviation of the DSC. In the study by Reijd et al., where 30 colorectal liver metastases were segmented by three different readers to evaluate the impact of manual segmentation contour differences on the agreement of radiomic features across MRI sequences, it was suggested that the variation in segmentation contours between readers had only a minor effect on the consistency of MRI-based radiomic features [ 11 ]. However, in lesions like EC, as in our study—where segmentation is more challenging compared to lesions such as lung nodules or liver metastases—contour differences appear to have a much greater impact on the agreement of radiomic features. Despite the contouring difficulty, particularly due to myometrial invasion, there are still radiomic features that demonstrate high reproducibility. In an analysis conducted by Xue et al. in 2021, which reviewed over 100 radiomic studies across various organs, it was suggested that inter-observer segmentation variability has a rather limited impact on the reproducibility of radiomic features [ 12 ]. In the same analysis, studies using the MRI modality that assessed the agreement of segmentation differences using ICC were found to have, on average, 80% (ranging from 20% to 100%) of their features deemed suitable for use after agreement analysis—referred to as the Satisfactory Feature Rate. However, although the satisfactory feature rate in this analysis was evaluated based on imaging modalities, it was not assessed on a lesion-specific basis. Moreover, only one EC radiomics study was included in this part of the analysis. Therefore, in studies developing MRI-based radiomic models for EC [ 13 – 15 ], the impact of segmentation variability on feature agreement remains a critical area of investigation. Particularly in EC radiomic modeling studies that use ICC for agreement assessment, the radiomic features will be filtered based on different reproducibility thresholds, such as 0.75 [ 14 ] or 0.8 [ 16 ]. Therefore, the selection of this threshold is of critical importance. Since there is no standardization for the ICC threshold selected to define agreement, reporting the proportion of radiomic features that are consistent at different ICC levels, and explicitly stating the ICC values of the features used in the constructed radiomic signature, can help create more reproducible models in such studies. When comparing the radiomic features derived from the specified segmentations across different sequences, features extracted from T1CE and T2WS images showed the highest reproducibility (57.9% and 59.9%, respectively), while those from T2WA and ADC images demonstrated the lowest reproducibility (44.9% and 45.8%, respectively). This suggests that different sequences have different inter-reader radiomics feature reproducibility in EC. By evaluating the agreement of shape features and DSC from segmentations performed on T2-weighted images, the three radiologists showed comparable results in terms of the similarity of segmented regions across all planes. However, despite this similarity in the delineated areas, the agreement of radiomic features derived from T2WA lagged behind those obtained from T2WS across all feature classes. This suggests that, when developing MRI-based radiomic models for EC, not only the sequence but also the imaging plane should be taken into consideration. Therefore our results do not align with what Reijd et al. has suggested, radiomic feature reproducibility being independent of sequence plane orientation [ 11 ]. Although most studies use the sagittal plane of T2-weighted sequences for segmentation, some have shown that axial [ 17 ] or axial-oblique [ 18 ] planes are also used. Such variations in plane selection may influence the reproducibility of the extracted features. In our study, the segmentation mask was placed on the DWI images, and the ADC features were extracted in parallel based on this mask. However, the agreement of radiomic features derived from the ADC maps—particularly first-order features—was lower compared to those extracted from DWI. This is most likely due to differences in the MRI scanners and the software used to generate the ADC images [ 19 – 21 ]. In order to better reflect routine clinical practice, we deliberately utilized the original ADC maps without applying any standardized reconstruction methods, preserving them in their native form. Had we applied different normalization and quantization approaches prior to feature extraction, we might have observed improved agreement between radiomic features obtained from different readers [ 22 ]. In our study, we performed segmentation on the DWI images; however, had the segmentation been carried out directly on the ADC maps, an increase in reproducibility might have been achieved. Shape features demonstrated the highest level of agreement among all radiomic feature groups. In both MRI- and CT-based studies, shape features are consistently identified as one of the most robust classes against test-retest variability and inter-reader differences, and our findings are in line with the existing literature [ 22 – 24 ]. Among the shape features, although “Elongation”, “Sphericity” and “Flatness” showed low reproducibility across all sequences, features such as “Least Axis Length,” “Major Axis Length,” “Maximum 2D Diameter Slice,” “Maximum 3D Diameter Slice,” and “Minor Axis Length” demonstrated high reproducibility in all sequences. Since the first three features, referred to as compactness descriptors, primarily reflect the roundness of the lesion, we suggest that they are significantly influenced by factors such as slice thickness, resolution, and segmentation variability. On the other hand, the consistent agreement of features measuring the lesion’s widest and narrowest dimensions indicates that, despite uncertainties at lesion borders, the readers shared a similar understanding of lesion size. When comparing radiomic feature subclasses, the NGTDM class showed the lowest reproducibility across all sequences. NGTDM is especially sensitive to contouring differences because it calculates its features by using a pixel and its neighbouring pixels. Since these neighboring pixels change significantly when contouring, NGTDM values become sensitive to segmentation differences. The low repeatability and high sensitivity to segmentation differences among readers of NGTDM features have been demonstrated in various disease contexts, and our findings are consistent with the existing literature [ 11 , 25 , 26 ]. GLCM, the second least reproducible radiomics subclass, quantifies how often specific combinations of gray-level intensities occur between pairs of pixels or voxels at a defined spatial relationship and orientation. We hypothesize since the pixel pair would be affected even at the slightest change in contouring, this subclass proved to be less producible than the other subclasses. There is a clear predilection of “HighGray” features being more reproducible among sequences, namely T2W and DWI, in which cancer tissue is more hyperintense compared to surrounding healthy tissue. In these segmentations regions with low signal intensities consist of only a small number of voxels, rendering them more susceptible to variability. However in T1CE where cancer tissue is supposed to be lower intensity than the highly enhancing myometrium, a relatively low number of “LowGray’ features showed excellent responsibility, which suggests tumor to healthy tissue contrast is more dominant in T2W images compared to T1CE. In the future, the use of fully automated segmentation techniques is anticipated to minimize inter-reader variability. In one study, when radiomic features derived from the segmentations of a deep learning model trained to segment EC on MRI were compared with those obtained from manual segmentations by an experienced radiologist, first-order and shape feature classes showed higher agreement than texture-based feature classes [ 27 ]. The same study reported that the segmentation model achieved a mean Dice Similarity Coefficient (DSC) of 0.806 when compared with the radiologist who segmented the training data and was considered the gold standard. This DSC is comparable to the average inter-reader DSC across all sequences observed in our study. Although radiomic features derived from both manual and automated segmentations were found to be consistent in that study, it should be noted that the segmentation area produced by the model inherently reflects the approach of the specific gold standard reader used during training. The lower agreements observed between different readers in our study suggest that, if such an automated segmentation model is to be developed, it should ideally be trained using segmentations performed by multiple experienced radiologists, preferably at different times. Otherwise, radiomic features extracted from automated segmentations may reflect the individual biases of the original reader. Our study has several limitations. First, it was designed retrospectively. Second, only 59 patients were included, and the reliability of our results could be improved by increasing the sample size. Third, only original radiomic features were analyzed; wavelet-derived features were not included. Fourth, 58 out of the 59 scans were acquired using MRI machines with a magnetic field strength of 1.5 Tesla, which may have lowered segmentation quality, particularly in diffusion-weighted images, due to reduced signal-to-noise ratio. Lastly, external scans accounted for approximately 14% of the total, and the vast majority of all scans (57 out of 59) were acquired using Siemens MRI systems. Including a more diverse set of scanner brands and models could have yielded a more heterogeneous dataset and potentially more generalizable results. Our study included 59 pre-treatment EC patients whose images were acquired from 9 different MRI scanners. We believe that this heterogeneous population enhances the generalizability of our findings. To develop better radiomic models, training on images from multiple centers may help compensate for acquisition differences, thus improving model generalizability [ 28 ]. In conclusion, with this study, we present the first MRI-based reproducibility analysis for EC and encourage other researchers to compare their results against ours. In EC, sagittal scan orientation yielded the best reproducibility among other orientations. T2 weighted and T1 weighted contrast enhanced images proved to be comparably reproducible while ADC features showed poor reproduciblity. Shape and GLDM features are the most reproducible radiomic feature subclasses. “HighGray” features are highly reproducible in T2 and diffusion weighted images. We hope that our findings will contribute to the generalizability and reproducibility of future radiomic models for EC. Abbreviations ADC, Apparent diffusion coefficient; DSC, Dice Similarity Coefficient; DWI, Diffusion weighted images; EC, Endometrial Cancer; FIGO, Federation of Gynecology and Obstetrics; GLCM, Gray Level Co-occurrence Matrix; GLDM, Gray Level Dependence Matrix; GLRLM, Gray Level Run Length Matrix; GLSZM, Gray Level Size Zone Matrix; ICC: Intraclass correlation coefficient; MRI, Magnetic Resonance Imaging; T1CE, T1 Weighted Contrast Enhanced Sagittal; NGTDM, Neighbouring Gray Tone Difference Matrix; T2WA, T2 Weighted Axial; T2WC, T2 Weighted Coronal; T2WS, T2 Weighted Sagittal; VOI, Volumes of Interest. Declarations Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work the author(s) used ChatGPT and Scite.ai in order to create code and find references. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. Author Contribution F.A : Conceptualization, Visualization, Investigation, Writing – original draft. M.A.G: Conceptualization, Writing – review and editing, Formal analysis, Software. A.A: Writing – review and editing, Formal analysis, Software. H.T: Supervision, Investigation. M.T: Data curation, Investigation. T.F.Y: Project administration, Methodology. References Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin . 2021;71(3):209–249. doi: 10.3322/caac.21660 He J, Liu Y, Li J, Liu S. Accuracy of radiomics in the diagnosis and preoperative high-risk assessment of endometrial cancer: a systematic review and meta-analysis. Front Oncol . 2024;14:1334546. Published 2024 Jan 25. doi: 10.3389/fonc.2024.1334546 Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer . 2012;48(4):441–446. doi: 10.1016/j.ejca.2011.11.036 Klontzas ME. Radiomics feature reproducibility: The elephant in the room. Eur J Radiol . 2024;175:111430. doi: 10.1016/j.ejrad.2024.111430 Yip SS, Aerts HJ. Applications and limitations of radiomics. Phys Med Biol . 2016;61(13):R150-R166. doi: 10.1088/0031-9155/61/13/R150 Nougaret S, Horta M, Sala E, et al. Endometrial Cancer MRI staging: Updated Guidelines of the European Society of Urogenital Radiology. Eur Radiol . 2019;29(2):792–805. doi: 10.1007/s00330-018-5515-y Fedorov A, Beichel R, Kalpathy-Cramer J, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. Magn Reson Imaging . 2012;30(9):1323–1341. doi: 10.1016/j.mri.2012.05.001 van Griethuysen JJM, Fedorov A, Parmar C, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res . 2017;77(21):e104-e107. doi: 10.1158/0008-5472.CAN-17-0339 Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research [published correction appears in J Chiropr Med. 2017;16(4):346. doi: 10.1016/j.jcm.2017 .10.001.]. J Chiropr Med . 2016;15(2):155–163. doi:10.1016/j.jcm.2016.02.012 M.A. Hadley Wickham, Jennifer Bryan, Winston Chang, Lucy D’Agostino McGowan, Romain François, Garrett Grolemund, Alex Hayes, Lionel Henry, Jim Hester, Max Kuhn, Thomas Lin Pedersen, Evan Miller, Stephan Milton Bache, Kirill Müller, Jeroen Ooms, David Robinson, Dana Paige Seidel, Vitalie Spinu, Kohske Takahashi, Davis Vaughan, Claus Wilke, Kara Woo, and Hiroaki Yutani, Welcome to the Tidyverse, The Journal of Open Source Software (2019). van der Reijd DJ, Chupetlovska K, van Dijk E, et al. Multi-sequence MRI radiomics of colorectal liver metastases: Which features are reproducible across readers?. Eur J Radiol . 2024;172:111346. doi: 10.1016/j.ejrad.2024.111346 Xue C, Yuan J, Lo GG, et al. Radiomics feature reliability assessed by intraclass correlation coefficient: a systematic review. Quant Imaging Med Surg . 2021;11(10):4431–4460. doi: 10.21037/qims-21-86 Xu X, Li H, Wang S, et al. Multiplanar MRI-Based Predictive Model for Preoperative Assessment of Lymph Node Metastasis in Endometrial Cancer. Front Oncol . 2019;9:1007. Published 2019 Oct 9. doi: 10.3389/fonc.2019.01007 Yan BC, Ma XL, Li Y, Duan SF, Zhang GF, Qiang JW. MRI-Based Radiomics Nomogram for Selecting Ovarian Preservation Treatment in Patients With Early-Stage Endometrial Cancer. Front Oncol . 2021;11:730281. Published 2021 Sep 9. doi: 10.3389/fonc.2021.730281 Liu XF, Yan BC, Li Y, Ma FH, Qiang JW. Radiomics Nomogram in Assisting Lymphadenectomy Decisions by Predicting Lymph Node Metastasis in Early-Stage Endometrial Cancer. Front Oncol . 2022;12:894918. Published 2022 May 31. doi: 10.3389/fonc.2022.894918 Jia Y, Hou L, Zhao J, et al. Radiomics analysis of multiparametric MRI for preoperative prediction of microsatellite instability status in endometrial cancer: a dual-center study. Front Oncol . 2024;14:1333020. Published 2024 Jan 29. doi: 10.3389/fonc.2024.1333020 Yan R, Qin S, Xu J, et al. A comparison of 2D and 3D magnetic resonance imaging-based intratumoral and peritumoral radiomics models for the prognostic prediction of endometrial cancer: a pilot study. Cancer Imaging . 2024;24(1):100. Published 2024 Jul 31. doi: 10.1186/s40644-024-00743-2 Liu D, Huang J, Zhang Y, et al. Multimodal MRI-based radiomics models for the preoperative prediction of lymphovascular space invasion of endometrial carcinoma. BMC Med Imaging . 2024;24(1):252. Published 2024 Sep 20. doi: 10.1186/s12880-024-01430-1 Ghosh A, Singh T, Singla V, Bagga R, Khandelwal N. Comparison of Absolute Apparent Diffusion Coefficient (ADC) Values in ADC Maps Generated Across Different Postprocessing Software: Reproducibility in Endometrial Carcinoma. AJR Am J Roentgenol . 2017;209(6):1312–1320. doi: 10.2214/AJR.17.18002 Traverso A, Kazmierski M, Shi Z, et al. Stability of radiomic features of apparent diffusion coefficient (ADC) maps for locally advanced rectal cancer in response to image pre-processing. Phys Med . 2019;61:44–51. doi: 10.1016/j.ejmp.2019.04.009 Michoux NF, Ceranka JW, Vandemeulebroucke J, et al. Repeatability and reproducibility of ADC measurements: a prospective multicenter whole-body-MRI study. Eur Radiol . 2021;31(7):4514–4527. doi: 10.1007/s00330-020-07522-0 Fiset S, Welch ML, Weiss J, et al. Repeatability and reproducibility of MRI-based radiomic features in cervical cancer. Radiother Oncol . 2019;135:107–114. doi: 10.1016/j.radonc.2019.03.001 Hu P, Wang J, Zhong H, et al. Reproducibility with repeat CT in radiomics study for rectal cancer. Oncotarget . 2016;7(44):71440–71446. doi: 10.18632/oncotarget.12199 van Timmeren JE, Leijenaar RTH, van Elmpt W, et al. Test-Retest Data for Radiomics Feature Stability Analysis: Generalizable or Study-Specific?. Tomography . 2016;2(4):361–365. doi: 10.18383/j.tom.2016.00208 Gourtsoyianni S, Doumou G, Prezzi D, et al. Primary Rectal Cancer: Repeatability of Global and Local-Regional MR Imaging Texture Features. Radiology. 2017;284(2):552–561. doi: 10.1148/radiol.2017161375 Sadeghi M, Abdalvand N, Mahdavi SR, et al. Magnetic Resonance Image Radiomic Reproducibility: The Impact of Preprocessing on Extracted Features from Gross and High-Risk Clinical Tumor Volumes in Cervical Cancer Patients before Brachytherapy. J Med Signals Sens. 2024;14:23. Published 2024 Aug 6. doi: 10.4103/jmss.jmss_57_22 Kurata Y, Nishio M, Moribata Y, et al. Automatic segmentation of uterine endometrial cancer on multi-sequence MRI using a convolutional neural network. Sci Rep . 2021;11(1):14440. Published 2021 Jul 14. doi: 10.1038/s41598-021-93792-7 Shahzadi I, Zwanenburg A, Lattermann A, et al. Analysis of MRI and CT-based radiomics features for personalized treatment in locally advanced rectal cancer and external validation of published radiomics models. Sci Rep . 2022;12(1):10192. Published 2022 Jun 17. doi: 10.1038/s41598-022-13967-8 Tables Tables 1 and 2 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table1.xlsx Table2.xlsx SupplementaryMaterialA.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-8184695","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":555428190,"identity":"da6fbf90-e33b-4c98-91ae-8ce5de0aaf47","order_by":0,"name":"Fazılhan Altıntaş","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABF0lEQVRIie3QMUvEMBTA8ScZbolkk4jDfYXIgdxwXL/KexR0E8VFuOEqB93EtSLoB3CpFIpjJEMXi2vFxXLgJpy4dDS5w61XdBPMfyhpeT/SBMDn+4P1IgBdR25lF+6LcA88Wk+4HdPkCGOWIMC2e0HVTeCbgCNquVcX6RW1pvvxoZiJB3PcjIPB0y0tXhX0xZZuJxxR02N4Ig0Dk2BIefWeSftju5dX2EoCQK0pZhQ5wpHhXlWmjqB6aSdc1JElU7pZkWkwSMqs6SQytDcWG0pXxGyk4jzv3kW+OVLQnWHK8P2CkmozH6KSa8/CxcF88RFP6Lo4m3/y0SQQF2X23JyO+mKnnbQkl5Pyp+MuoX8z7fP5fP+gLyTsZletRHo5AAAAAElFTkSuQmCC","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Fazılhan","middleName":"","lastName":"Altıntaş","suffix":""},{"id":555428192,"identity":"91fe6dda-e50e-44db-818a-844433e7eb66","order_by":1,"name":"Mehmet Ali Gültekin","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Mehmet","middleName":"Ali","lastName":"Gültekin","suffix":""},{"id":555428194,"identity":"f794ab74-3689-4b93-8742-d60bc6ee5fc1","order_by":2,"name":"Ahmet Akçay","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ahmet","middleName":"","lastName":"Akçay","suffix":""},{"id":555428196,"identity":"56b31e65-dd66-4b48-b033-2d57387158d9","order_by":3,"name":"Hüseyin Toprak","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Hüseyin","middleName":"","lastName":"Toprak","suffix":""},{"id":555428197,"identity":"a5c0bc21-f584-4f00-a58f-31b946333f7c","order_by":4,"name":"Murat Tabar","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Murat","middleName":"","lastName":"Tabar","suffix":""},{"id":555428199,"identity":"94e4beb8-80a4-4c38-9442-0f3de3897663","order_by":5,"name":"Temel Fatih Yılmaz","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Temel","middleName":"Fatih","lastName":"Yılmaz","suffix":""}],"badges":[],"createdAt":"2025-11-23 10:08:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8184695/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8184695/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97895322,"identity":"7901921c-96a9-4a07-b96d-7069f5a55fb4","added_by":"auto","created_at":"2025-12-10 15:34:00","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13564461,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/2ad8a59af2bd3447ace4c227.docx"},{"id":97720664,"identity":"45415019-9f9f-41c2-88da-09fc14996cb0","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82543,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/e75bd13a04e3e428e133b9f0.png"},{"id":97720667,"identity":"81c3e48b-d18b-4a85-a565-53821f445499","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":520526,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/8cfddc329ca4e37c02dd3dd8.png"},{"id":97720673,"identity":"06ad0244-412d-489f-8acb-2f7c4209b705","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5737,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/a9ced906ed8b1e5e1da70d5f.xlsx"},{"id":97895528,"identity":"3d638cfe-d6bb-443c-8772-c55a574a13b7","added_by":"auto","created_at":"2025-12-10 15:34:24","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":186688,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/a632289a85748e6a6f893922.png"},{"id":97720672,"identity":"d03715ce-48a5-41df-8822-e9fb57c67b41","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5655,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/2de85bb7fe99630723bce13a.xlsx"},{"id":97720674,"identity":"39dc9b48-9b67-4845-9967-d38384a25f66","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":574609,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/5f5a5341df136afa11b498d6.png"},{"id":97894914,"identity":"0891579a-51b4-4aa3-b5be-7f4e3a0d9931","added_by":"auto","created_at":"2025-12-10 15:33:13","extension":"json","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7875,"visible":true,"origin":"","legend":"","description":"","filename":"20641a688e6445b2a53f2fbb38d5abeb.json","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/f0bb2051c5ab5519c90ce6ab.json"},{"id":97895387,"identity":"4b4b4930-8f4d-4578-95d6-7d60b3f6ab35","added_by":"auto","created_at":"2025-12-10 15:34:07","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50757,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialA.docx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/0a42aae42e9ab66cae2d0a73.docx"},{"id":97720680,"identity":"183b035c-1ed0-4304-9faa-e73730f93623","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78026,"visible":true,"origin":"","legend":"","description":"","filename":"20641a688e6445b2a53f2fbb38d5abeb1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/dd065b94afb3371b2d043b64.xml"},{"id":97720678,"identity":"5290cf27-47e0-4743-b890-6732a20d86e4","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82543,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/f82f9529d2aa76d56154a44e.png"},{"id":97896155,"identity":"9896b32e-5497-4dba-a880-7b10f40f7c93","added_by":"auto","created_at":"2025-12-10 15:35:58","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":520526,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/cae7c8216c1b9e513dac8756.png"},{"id":97894556,"identity":"a4df422a-e33a-41a6-a23a-a8373790f002","added_by":"auto","created_at":"2025-12-10 15:32:42","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":186688,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/c0503b8d3e3ecc7555553672.png"},{"id":97894862,"identity":"020bfa85-5ebf-4234-b2da-c60d018fd17d","added_by":"auto","created_at":"2025-12-10 15:33:10","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":574609,"visible":true,"origin":"","legend":"","description":"","filename":"FIGURE4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/586b0c78d4afa0252c8756df.png"},{"id":97894705,"identity":"25a9f786-76f0-47e5-b655-6b398a6ca40f","added_by":"auto","created_at":"2025-12-10 15:32:54","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65904,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/8206d9519b1f58a73d114727.png"},{"id":97720694,"identity":"79b141df-f09f-4403-a75f-86b67f363a7b","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":482717,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/b5fe384e7f663b9646b06da8.png"},{"id":97720676,"identity":"0a0bc077-1e57-44ed-bf6a-486f9e096164","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":184576,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/3666de483a8f2cf35a68c3dd.png"},{"id":97896144,"identity":"f4e2b116-4850-4ceb-bc08-b7a0cafc6811","added_by":"auto","created_at":"2025-12-10 15:35:57","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":238315,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/35c9089e918973703d211bb1.png"},{"id":97895949,"identity":"130442d9-9673-4186-9f4e-2e0bd16f784b","added_by":"auto","created_at":"2025-12-10 15:35:26","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":26400,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFIGURE1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/6b408d2e3941d88857998b7e.png"},{"id":97720696,"identity":"90ef167c-0386-4047-865e-9ad23b51401b","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51425,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFIGURE2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/771e92d5a68ec86434f487a2.png"},{"id":97895395,"identity":"662be825-1fdb-4a84-a215-e3eaca83343d","added_by":"auto","created_at":"2025-12-10 15:34:07","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41865,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFIGURE3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/df6d0e4d5895c790b8b99932.png"},{"id":97720690,"identity":"cc1f81b5-8960-41eb-bcf5-0dfe2ee2f8c2","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":146082,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFIGURE4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/08e7f7631838c92e52d8f53b.png"},{"id":97720695,"identity":"ca773e3c-b398-4730-9cc1-f0a9126ebea7","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":26400,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/57b0b5b02c4f4b5c25b0b57e.png"},{"id":97720691,"identity":"463ce0c4-3880-4b2a-8e1d-e8e248e1444e","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51404,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/c245583b11eca94d575c14a3.png"},{"id":97895453,"identity":"02a2710c-ea07-4b62-baed-2582f695f8aa","added_by":"auto","created_at":"2025-12-10 15:34:15","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39745,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/3e757d9547018da50c953265.png"},{"id":97720685,"identity":"654456d7-7fc2-46f1-bb35-476e18335bc9","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58313,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/9fdfe2fdf5c49bb364c3acbb.png"},{"id":97720687,"identity":"1c703d89-9b00-422f-9bcb-2449bc7f27bd","added_by":"auto","created_at":"2025-12-08 15:36:06","extension":"xml","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75788,"visible":true,"origin":"","legend":"","description":"","filename":"20641a688e6445b2a53f2fbb38d5abeb1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/e3e27e9c019b7804b96ef668.xml"},{"id":97895576,"identity":"236a369a-de92-49f6-8bf7-6bcf58e9e0fb","added_by":"auto","created_at":"2025-12-10 15:34:28","extension":"html","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":85222,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/affdcf30e5ba795ad52b1e51.html"},{"id":97720663,"identity":"6e9be648-b00d-46d0-97b2-ec474faad3d9","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":82543,"visible":true,"origin":"","legend":"\u003cp\u003eA graphical summary of patient selection process and inclusion-exclusion criteria\u003c/p\u003e","description":"","filename":"FIGURE1.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/f15c90ae65c268bca384afeb.png"},{"id":97720669,"identity":"1bf58cf0-2728-4c9d-99f2-57d6740ed853","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":520526,"visible":true,"origin":"","legend":"\u003cp\u003eA representative endometrial cancer case segmented by three different readers is shown across multiple MRI modalities: (A) contrast-enhanced T1-weighted sagittal, (B) T2-weighted axial, (C) T2-weighted sagittal, and (D) diffusion-weighted imaging. The inter-reader agreement for each modality was assessed using the Dice Similarity Coefficient (DSC) between reader pairs.\u003c/p\u003e","description":"","filename":"FIGURE2.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/bf2dbef643a8a8db88dff58a.png"},{"id":97895540,"identity":"1a15926b-6e11-49aa-a50f-18005a31cf2e","added_by":"auto","created_at":"2025-12-10 15:34:25","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":186688,"visible":true,"origin":"","legend":"\u003cp\u003eInter-reader reproducibility of radiomic features across MRI sequences, assessed using the Intraclass Correlation Coefficient (ICC) and the lower bound of the 95% confidence interval (LB 95% CI) for each radiomics feature.\u003c/p\u003e","description":"","filename":"FIGURE3.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/deee4578890587d61a547953.png"},{"id":97894730,"identity":"f8609b4c-eb1c-4892-9380-744f148814d9","added_by":"auto","created_at":"2025-12-10 15:32:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":574609,"visible":true,"origin":"","legend":"\u003cp\u003eReproducibility of radiomic feature subgroups for each annotated sequence.\u003c/p\u003e","description":"","filename":"FIGURE4.png","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/4c97b49b72b294456b49841f.png"},{"id":101239475,"identity":"e199d87e-f5a5-477a-8695-937c2277c51c","added_by":"auto","created_at":"2026-01-27 15:13:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1802248,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/c43ebdfd-dfff-4ceb-b41b-830a66ffec43.pdf"},{"id":97720661,"identity":"edb2034a-76c5-42b2-a944-4fe2f9433eed","added_by":"auto","created_at":"2025-12-08 15:36:05","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":5737,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/dbe5518896d1f34afc46b96e.xlsx"},{"id":97896055,"identity":"b9d017ca-f57e-4e4e-9593-0572322ff183","added_by":"auto","created_at":"2025-12-10 15:35:45","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5655,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/4c0cb58fff1651d1bd4e04f5.xlsx"},{"id":97894717,"identity":"c36700cb-5703-4f23-ba8c-3332fa299550","added_by":"auto","created_at":"2025-12-10 15:32:56","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":50757,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialA.docx","url":"https://assets-eu.researchsquare.com/files/rs-8184695/v1/47edff8fdd10cc8060635dda.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAssessment of Inter-observer Reproducibility of Radiomic Features in Multiparametric Magnetic Resonance Imaging of Endometrial Cancer\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eEndometrial cancer (EC) is the sixth most common cancer among women and the most frequently occurring cancer of the female reproductive system. Worldwide, in 2020 alone, 417,000 new cases were diagnosed [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The potential of radiomics has been investigated by several studies to assist in detecting prognostic factors such as myometrial invasion or lymph node metastasis [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Radiomics is the process of extracting high-throughput quantitative data from medical images using computational analysis, generating objective features beyond what the human eye can assess, such as tumor biology [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. However despite an ever increasing number of radiomics models, clinical use of them has been very limited, due to poor reproducibility of the models [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe reproducibility problem arises due to use of features prone to variability. Since creation of radiomic models is a multi-step process, variations introduced at different stages\u0026mdash;such as image acquisition, tumor annotation, feature extraction, exploratory analysis and model creation\u0026mdash;can affect feature stability and limit the models performance when applied to new datasets, making reproducibility a major challenge in radiomics research [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Manual segmentation is the most used segmentation method in EC radiomics studies and inter-observer segmentation variations are one possible cause of reproducibility problems in EC radiomics models.\u003c/p\u003e\u003cp\u003eContrast-enhanced pelvic MRI is considered the gold standard imaging modality in pretreatment staging of EC, due to its high soft tissue resolution [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. In addition to conventional sequences that provide detailed information on tumor location, extent, contours, depth, and lymph node involvement, the use of dynamic contrast-enhanced imaging and diffusion-weighted imaging\u0026mdash;which measures the diffusivity of water molecules within the tumor\u0026mdash;enables more comprehensive characterization of tumor tissue. Though a multitude of radiomics models on MRI derived features has been created [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], the reproducibility of MRI derived radiomics features following multi-observer manual segmentation of EC has not been assessed yet, as far as we know. To address this gap in the literature, the goal of this study is to evaluate robustness of MRI derived radiomics features of EC across sequences, imaging orientation and feature classes.\u003c/p\u003e"},{"header":"2. Material and Method","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1. Patient Selection\u003c/h2\u003e\u003cp\u003e This retrospective study was approved by the Institutional Ethics Board (E-54022451-050.04-168547). Therefore informed consent requirements were waived. Pathology reports and patient information were available from medical reports.\u003c/p\u003e\u003cp\u003eBetween 2018 February \u0026minus;\u0026thinsp;2025 February, EC patients who underwent magnetic resonance imaging were selected according to the following criteria: (1) All three planes were acquired in T2 Weighted Imaging (2) Patients received no treatment before an MRI was acquired. If following criteria was met, patients were excluded: (1) Artefacts reducing image quality, (2) Tumors smaller than 1 cm or not visible at all, (3) Extensive hemorrhage impairing tumor evaluation. According to these criteria, 59 patients were deemed eligible for the study. The selection process was summarized in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Image Acquisition\u003c/h2\u003e\u003cp\u003ePretreatment MRIs were collected from the picture archiving and communication system (PACS:FUJIPACS LTD). Because this study was conducted in a university\u003c/p\u003e\u003cp\u003ehospital, some of the patients have undergone MRI in the referring hospitals and their images were already uploaded in our database. In total 7 different MRI models from 2 vendors have been used. Detailed information about scanners and acquisition parameters are available in Table\u0026nbsp;1. The ADC maps used in the study were derived directly from the original MRI scans; no standardized reconstruction of new ADC maps was performed, in order to reflect real-world clinical practice.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Segmentations\u003c/h2\u003e\u003cp\u003eThe relevant volumes (VOI \u0026ndash; Volumes of Interest) were manually segmented using the open-source software 3D Slicer (version 5.6.2)[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Two experienced radiologists specialized in abdominopelvic radiology (with 7 and 12 years of experience, respectively \u0026ndash; AA and MAG) and a radiology resident with 4 years of experience (FA) independently segmented the EC regions. The radiology resident received assistance from another specialist with 15 years of experience in abdominal radiology (HT) when needed. The readers carefully delineated the EC regions on T2-weighted sagittal (T2WS), coronal (T2WC), axial (T2WA),T1-weighted contrast-enhanced sagittal (T1CE), and diffusion-weighted imaging (DWI) sequences, avoiding other possible surrounding structures such as fibroids or implants. For diffusion measurements, the highest b-value was used. The segmentation masks performed on the diffusion-weighted images (DWI) were transferred as-is to the ADC maps. Examples of the segmentations are displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Radiomics\u003c/h2\u003e\u003cp\u003eFor each segmented sequence, 107 radiomics features were extracted using Pyradiomics (v 3.0.1), reaching a sum of 642 features for each case [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. No wavelet features were extracted. Features were extracted from seven distinct classes: First Order (n\u0026thinsp;=\u0026thinsp;18), Gray Level Dependence Matrix (GLDM, n\u0026thinsp;=\u0026thinsp;14), Gray Level Co-occurrence Matrix (GLCM, n\u0026thinsp;=\u0026thinsp;24), Gray Level Run Length Matrix (GLRLM, n\u0026thinsp;=\u0026thinsp;16), Gray Level Size Zone Matrix (GLSZM, n\u0026thinsp;=\u0026thinsp;16), Neighbouring Gray Tone Difference Matrix (NGTDM, n\u0026thinsp;=\u0026thinsp;5), and Shape features (n\u0026thinsp;=\u0026thinsp;14). A bin width of 18 was used during the extraction process. To prevent any influence on inter-observer variability, no filters or normalization methods were applied.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Statistical Analysis\u003c/h2\u003e\u003cp\u003eFor each case and reader pair a Dice Similarity Coefficient (DSC) was calculated, in order to compare similarity of delineation areas. The agreement between observers was assessed using the Intraclass Correlation Coefficient (ICC), applying a two-way random effects model with a single-measurement and absolute agreement approach. Absolute agreement was chosen over consistency because using the consistency definition might compromise the reliability of features, especially in scenarios that involve applying cutoff values. Following the interpretation guidelines by Koo et al., agreement levels were classified as poor for ICC\u0026thinsp;\u0026le;\u0026thinsp;0.5, moderate for ICC between 0.50 and 0.75, good for ICC between 0.75 and 0.90, and excellent for ICC\u0026thinsp;\u0026ge;\u0026thinsp;0.9 [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. In this study, features with ICC values of 0.9 or higher were regarded as robust across different readers. Radiomic features with an ICC of 0.9 or higher were considered resistant to inter-reader variability. ICC values were calculated for each sequence and imaging plane. Descriptive statistics were used to compare ICC differences across subgroups. Statistical analyses and relevant visualizations were performed using R programming language (R Core Team v4.4.2) on R Studio (R Studio Team v4.3.3) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. The \u0026ldquo;tidyverse\u0026rdquo; and \u0026ldquo;irr\u0026rdquo; packages were utilized.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Patient Population and Segmentation Metrics\u003c/h2\u003e\u003cp\u003eIn our study, preoperative imaging from 59 patients diagnosed with EC was used. The mean age was calculated as 62 years (range: 38\u0026ndash;88). Among the 59 patients included in the study, 52 (88.1%) were diagnosed with endometrioid adenocarcinoma, 6 (10.2%) with serous adenocarcinoma, and 1 (1.7%) with clear cell carcinoma. According to the Federation of Gynecology and Obstetrics (FIGO) grading system, 34 cases (57.6%) were classified as high-grade tumors, while the remaining 25 cases (42.4%) were low-grade. Based on T2-weighted sagittal images, the mean lesion volume, averaged across all readers using the mesh volume feature, was calculated as 23,973 mm\u0026sup3;. Using the same method, the mean short axis of the lesions was approximately 2 cm, and the mean long axis was approximately 4.5 cm. The overall mean DSC for all segmentations was 0.77. When analyzed by sequence, the mean DSCs were as follows: T2WS: 0.79 (range: 0.51\u0026ndash;0.92, SD: 0.07), T2WC: 0.77 (range: 0.53\u0026ndash;0.92, SD: 0.09), T2WA: 0.80 (range: 0.59\u0026ndash;0.91, SD: 0.08), T1CE: 0.77 (range: 0.40\u0026ndash;0.90, SD: 0.07), and DWI: 0.73 (range: 0.42\u0026ndash;0.90, SD: 0.08).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Sequences\u003c/h2\u003e\u003cp\u003eA total of 338 out of 642 radiomic features (53%) demonstrated excellent reproducibility (ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.9) despite inter-reader segmentation variability. When the lower bound of the 95% confidence interval was considered, this proportion decreased to 240 out of 642 features (37%). Radiomic features derived from T2WS and T1CE sequences showed the highest reproducibility (58.9% and 57.9%, respectively), whereas those derived from T2WAand ADC maps showed the lowest reproducibility (44.9% and 45.8%, respectively). The median ICC values by sequence were calculated as follows: T2WS: 0.91 (range: 0.61\u0026ndash;0.99), T2WC: 0.91 (0.60\u0026ndash;0.99), T2WA: 0.87 (0.45\u0026ndash;0.99), T1CE: 0.92 (0.65\u0026ndash;0.99), DWI: 0.91 (0.33\u0026ndash;0.99), and ADC: 0.89 (0.28\u0026ndash;0.99). The agreement of radiomic features was visualized graphically using ICC values and their 95% confidence intervals in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Radiomics Subclasses\u003c/h2\u003e\u003cp\u003eIn order to compare feature subclasses among sequences, T2WS was chosen as a representative of T2W images due to its better performance. Examining the reproducibility of radiomic feature classes, shape and GLDM features showed the highest reproducibility (60.7% and 64.3%, respectively), whereas NGTDM and GLCM features exhibited the lowest reproducibility (30% and 45.8%, respectively) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Among the radiomic features, 19 demonstrated excellent agreement across all sequences, while 23 features showed poor agreement in all sequences (Table\u0026nbsp;2). When evaluated by feature class, none of the features from the NGTDM and First Order classes exhibited excellent agreement across all sequences (Table\u0026nbsp;2). There were 9 \u0026ldquo;HighGray\u0026rdquo; and 9 \u0026ldquo;LowGray\u0026rdquo; features for each sequence. The distribution of 'HighGray' features with excellent reproducibility across sequences was: T2WS 5/9, T2WC 9/9, T2WA 3/9, T1CE 0/9, DWI 9/9, ADC 0/9. The distribution of 'LowGray' features with excellent reproducibility across sequences was:T2WS 2/9, T2WC 0/9, T2WA 1/9, T1CE 1/9, DWI 0/9, ADC 1/9\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAll ICC values for each feature-sequence pair can be found in Supplementary Material A.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eConsidering the Dice Similarity Coefficients (DSC) and the agreement of shape features between reader pairs across sequences, a difference in segmentation contours among readers was observed \u0026mdash; being the smallest in T2WA (DSC: 0.8) and the largest in DWI (DSC: 0.73). Except for DWI/ADC, no major differences were observed in the mean Dice Similarity Coefficients (DSC), which reflect the overlapping areas of the segmentations, across different sequences. We attributed the relatively lower DSC in these segmentations to the fact that most of the images in our dataset were acquired using 1.5 Tesla MRI scanners, which typically yield diffusion-weighted images with lower signal-to-noise ratios, making segmentation more challenging. This explains the lower mean DSC observed in the diffusion-weighted images. Since our study aimed to reflect a real-world radiomic modeling scenario, we chose to include cases with low DSC across sequences rather than exclude them.\u003c/p\u003e\u003cp\u003eOur findings suggest that inter-reader segmentation differences have a significant impact on radiomic features. When evaluating the agreement of radiomic features calculated across sequences, discrepancies were observed that could not be fully explained by the mean or standard deviation of the DSC. In the study by Reijd et al., where 30 colorectal liver metastases were segmented by three different readers to evaluate the impact of manual segmentation contour differences on the agreement of radiomic features across MRI sequences, it was suggested that the variation in segmentation contours between readers had only a minor effect on the consistency of MRI-based radiomic features [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, in lesions like EC, as in our study\u0026mdash;where segmentation is more challenging compared to lesions such as lung nodules or liver metastases\u0026mdash;contour differences appear to have a much greater impact on the agreement of radiomic features. Despite the contouring difficulty, particularly due to myometrial invasion, there are still radiomic features that demonstrate high reproducibility. In an analysis conducted by Xue et al. in 2021, which reviewed over 100 radiomic studies across various organs, it was suggested that inter-observer segmentation variability has a rather limited impact on the reproducibility of radiomic features [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. In the same analysis, studies using the MRI modality that assessed the agreement of segmentation differences using ICC were found to have, on average, 80% (ranging from 20% to 100%) of their features deemed suitable for use after agreement analysis\u0026mdash;referred to as the Satisfactory Feature Rate. However, although the satisfactory feature rate in this analysis was evaluated based on imaging modalities, it was not assessed on a lesion-specific basis. Moreover, only one EC radiomics study was included in this part of the analysis. Therefore, in studies developing MRI-based radiomic models for EC [\u003cspan additionalcitationids=\"CR14\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], the impact of segmentation variability on feature agreement remains a critical area of investigation. Particularly in EC radiomic modeling studies that use ICC for agreement assessment, the radiomic features will be filtered based on different reproducibility thresholds, such as 0.75 [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] or 0.8 [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Therefore, the selection of this threshold is of critical importance. Since there is no standardization for the ICC threshold selected to define agreement, reporting the proportion of radiomic features that are consistent at different ICC levels, and explicitly stating the ICC values of the features used in the constructed radiomic signature, can help create more reproducible models in such studies.\u003c/p\u003e\u003cp\u003eWhen comparing the radiomic features derived from the specified segmentations across different sequences, features extracted from T1CE and T2WS images showed the highest reproducibility (57.9% and 59.9%, respectively), while those from T2WA and ADC images demonstrated the lowest reproducibility (44.9% and 45.8%, respectively). This suggests that different sequences have different inter-reader radiomics feature reproducibility in EC.\u003c/p\u003e\u003cp\u003eBy evaluating the agreement of shape features and DSC from segmentations performed on T2-weighted images, the three radiologists showed comparable results in terms of the similarity of segmented regions across all planes. However, despite this similarity in the delineated areas, the agreement of radiomic features derived from T2WA lagged behind those obtained from T2WS across all feature classes. This suggests that, when developing MRI-based radiomic models for EC, not only the sequence but also the imaging plane should be taken into consideration. Therefore our results do not align with what Reijd et al. has suggested, radiomic feature reproducibility being independent of sequence plane orientation [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Although most studies use the sagittal plane of T2-weighted sequences for segmentation, some have shown that axial [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] or axial-oblique [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] planes are also used. Such variations in plane selection may influence the reproducibility of the extracted features.\u003c/p\u003e\u003cp\u003eIn our study, the segmentation mask was placed on the DWI images, and the ADC features were extracted in parallel based on this mask. However, the agreement of radiomic features derived from the ADC maps\u0026mdash;particularly first-order features\u0026mdash;was lower compared to those extracted from DWI. This is most likely due to differences in the MRI scanners and the software used to generate the ADC images [\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. In order to better reflect routine clinical practice, we deliberately utilized the original ADC maps without applying any standardized reconstruction methods, preserving them in their native form. Had we applied different normalization and quantization approaches prior to feature extraction, we might have observed improved agreement between radiomic features obtained from different readers [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. In our study, we performed segmentation on the DWI images; however, had the segmentation been carried out directly on the ADC maps, an increase in reproducibility might have been achieved. Shape features demonstrated the highest level of agreement among all radiomic feature groups. In both MRI- and CT-based studies, shape features are consistently identified as one of the most robust classes against test-retest variability and inter-reader differences, and our findings are in line with the existing literature [\u003cspan additionalcitationids=\"CR23\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Among the shape features, although \u0026ldquo;Elongation\u0026rdquo;, \u0026ldquo;Sphericity\u0026rdquo; and \u0026ldquo;Flatness\u0026rdquo; showed low reproducibility across all sequences, features such as \u0026ldquo;Least Axis Length,\u0026rdquo; \u0026ldquo;Major Axis Length,\u0026rdquo; \u0026ldquo;Maximum 2D Diameter Slice,\u0026rdquo; \u0026ldquo;Maximum 3D Diameter Slice,\u0026rdquo; and \u0026ldquo;Minor Axis Length\u0026rdquo; demonstrated high reproducibility in all sequences. Since the first three features, referred to as compactness descriptors, primarily reflect the roundness of the lesion, we suggest that they are significantly influenced by factors such as slice thickness, resolution, and segmentation variability. On the other hand, the consistent agreement of features measuring the lesion\u0026rsquo;s widest and narrowest dimensions indicates that, despite uncertainties at lesion borders, the readers shared a similar understanding of lesion size.\u003c/p\u003e\u003cp\u003eWhen comparing radiomic feature subclasses, the NGTDM class showed the lowest reproducibility across all sequences. NGTDM is especially sensitive to contouring differences because it calculates its features by using a pixel and its neighbouring pixels. Since these neighboring pixels change significantly when contouring, NGTDM values become sensitive to segmentation differences. The low repeatability and high sensitivity to segmentation differences among readers of NGTDM features have been demonstrated in various disease contexts, and our findings are consistent with the existing literature [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. GLCM, the second least reproducible radiomics subclass, quantifies how often specific combinations of gray-level intensities occur between pairs of pixels or voxels at a defined spatial relationship and orientation. We hypothesize since the pixel pair would be affected even at the slightest change in contouring, this subclass proved to be less producible than the other subclasses.\u003c/p\u003e\u003cp\u003eThere is a clear predilection of \u0026ldquo;HighGray\u0026rdquo; features being more reproducible among sequences, namely T2W and DWI, in which cancer tissue is more hyperintense compared to surrounding healthy tissue. In these segmentations regions with low signal intensities consist of only a small number of voxels, rendering them more susceptible to variability. However in T1CE where cancer tissue is supposed to be lower intensity than the highly enhancing myometrium, a relatively low number of \u0026ldquo;LowGray\u0026rsquo; features showed excellent responsibility, which suggests tumor to healthy tissue contrast is more dominant in T2W images compared to T1CE.\u003c/p\u003e\u003cp\u003eIn the future, the use of fully automated segmentation techniques is anticipated to minimize inter-reader variability. In one study, when radiomic features derived from the segmentations of a deep learning model trained to segment EC on MRI were compared with those obtained from manual segmentations by an experienced radiologist, first-order and shape feature classes showed higher agreement than texture-based feature classes [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. The same study reported that the segmentation model achieved a mean Dice Similarity Coefficient (DSC) of 0.806 when compared with the radiologist who segmented the training data and was considered the gold standard. This DSC is comparable to the average inter-reader DSC across all sequences observed in our study. Although radiomic features derived from both manual and automated segmentations were found to be consistent in that study, it should be noted that the segmentation area produced by the model inherently reflects the approach of the specific gold standard reader used during training. The lower agreements observed between different readers in our study suggest that, if such an automated segmentation model is to be developed, it should ideally be trained using segmentations performed by multiple experienced radiologists, preferably at different times. Otherwise, radiomic features extracted from automated segmentations may reflect the individual biases of the original reader.\u003c/p\u003e\u003cp\u003eOur study has several limitations. First, it was designed retrospectively. Second, only 59 patients were included, and the reliability of our results could be improved by increasing the sample size. Third, only original radiomic features were analyzed; wavelet-derived features were not included. Fourth, 58 out of the 59 scans were acquired using MRI machines with a magnetic field strength of 1.5 Tesla, which may have lowered segmentation quality, particularly in diffusion-weighted images, due to reduced signal-to-noise ratio. Lastly, external scans accounted for approximately 14% of the total, and the vast majority of all scans (57 out of 59) were acquired using Siemens MRI systems. Including a more diverse set of scanner brands and models could have yielded a more heterogeneous dataset and potentially more generalizable results.\u003c/p\u003e\u003cp\u003eOur study included 59 pre-treatment EC patients whose images were acquired from 9 different MRI scanners. We believe that this heterogeneous population enhances the generalizability of our findings. To develop better radiomic models, training on images from multiple centers may help compensate for acquisition differences, thus improving model generalizability [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn conclusion, with this study, we present the first MRI-based reproducibility analysis for EC and encourage other researchers to compare their results against ours. In EC, sagittal scan orientation yielded the best reproducibility among other orientations. T2 weighted and T1 weighted contrast enhanced images proved to be comparably reproducible while ADC features showed poor reproduciblity. Shape and GLDM features are the most reproducible radiomic feature subclasses. \u0026ldquo;HighGray\u0026rdquo; features are highly reproducible in T2 and diffusion weighted images. We hope that our findings will contribute to the generalizability and reproducibility of future radiomic models for EC.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eADC, Apparent diffusion coefficient; DSC, Dice Similarity Coefficient; DWI, Diffusion weighted images; EC, Endometrial Cancer; FIGO, Federation of Gynecology and Obstetrics; GLCM, Gray Level Co-occurrence Matrix; GLDM, Gray Level Dependence Matrix; GLRLM, Gray Level Run Length Matrix; GLSZM, Gray Level Size Zone Matrix; ICC: Intraclass correlation coefficient; MRI, Magnetic Resonance Imaging; T1CE, T1 Weighted Contrast Enhanced Sagittal; NGTDM, Neighbouring Gray Tone Difference Matrix; T2WA, T2 Weighted Axial; T2WC, T2 Weighted Coronal; T2WS, T2 Weighted Sagittal; VOI, Volumes of Interest.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eDeclaration of generative AI and AI-assisted technologies in the writing process\u003c/h2\u003e\n\u003cp\u003eDuring the preparation of this work the author(s) used ChatGPT and Scite.ai in order to create code and find references. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eF.A : Conceptualization, Visualization, Investigation, Writing \u0026ndash; original draft. M.A.G: Conceptualization, Writing \u0026ndash; review and editing, Formal analysis, Software. A.A: Writing \u0026ndash; review and editing, Formal analysis, Software. H.T: Supervision, Investigation. M.T: Data curation, Investigation. T.F.Y: Project administration, Methodology.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. \u003cem\u003eCA Cancer J Clin\u003c/em\u003e. 2021;71(3):209\u0026ndash;249. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3322/caac.21660\u003c/span\u003e\u003cspan address=\"10.3322/caac.21660\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHe J, Liu Y, Li J, Liu S. Accuracy of radiomics in the diagnosis and preoperative high-risk assessment of endometrial cancer: a systematic review and meta-analysis. \u003cem\u003eFront Oncol\u003c/em\u003e. 2024;14:1334546. Published 2024 Jan 25. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2024.1334546\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2024.1334546\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: extracting more information from medical images using advanced feature analysis. \u003cem\u003eEur J Cancer\u003c/em\u003e. 2012;48(4):441\u0026ndash;446. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ejca.2011.11.036\u003c/span\u003e\u003cspan address=\"10.1016/j.ejca.2011.11.036\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKlontzas ME. Radiomics feature reproducibility: The elephant in the room. \u003cem\u003eEur J Radiol\u003c/em\u003e. 2024;175:111430. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ejrad.2024.111430\u003c/span\u003e\u003cspan address=\"10.1016/j.ejrad.2024.111430\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYip SS, Aerts HJ. Applications and limitations of radiomics. \u003cem\u003ePhys Med Biol\u003c/em\u003e. 2016;61(13):R150-R166. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1088/0031-9155/61/13/R150\u003c/span\u003e\u003cspan address=\"10.1088/0031-9155/61/13/R150\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNougaret S, Horta M, Sala E, et al. Endometrial Cancer MRI staging: Updated Guidelines of the European Society of Urogenital Radiology. \u003cem\u003eEur Radiol\u003c/em\u003e. 2019;29(2):792\u0026ndash;805. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00330-018-5515-y\u003c/span\u003e\u003cspan address=\"10.1007/s00330-018-5515-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFedorov A, Beichel R, Kalpathy-Cramer J, et al. 3D Slicer as an image computing platform for the Quantitative Imaging Network. \u003cem\u003eMagn Reson Imaging\u003c/em\u003e. 2012;30(9):1323\u0026ndash;1341. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.mri.2012.05.001\u003c/span\u003e\u003cspan address=\"10.1016/j.mri.2012.05.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003evan Griethuysen JJM, Fedorov A, Parmar C, et al. Computational Radiomics System to Decode the Radiographic Phenotype. \u003cem\u003eCancer Res\u003c/em\u003e. 2017;77(21):e104-e107. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1158/0008-5472.CAN-17-0339\u003c/span\u003e\u003cspan address=\"10.1158/0008-5472.CAN-17-0339\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKoo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research [published correction appears in J Chiropr Med. 2017;16(4):346. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jcm.2017\u003c/span\u003e\u003cspan address=\"10.1016/j.jcm.2017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.10.001.]. \u003cem\u003eJ Chiropr Med\u003c/em\u003e. 2016;15(2):155\u0026ndash;163. doi:10.1016/j.jcm.2016.02.012\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eM.A. Hadley Wickham, Jennifer Bryan, Winston Chang, Lucy D\u0026rsquo;Agostino McGowan, Romain Fran\u0026ccedil;ois, Garrett Grolemund, Alex Hayes, Lionel Henry, Jim Hester, Max Kuhn, Thomas Lin Pedersen, Evan Miller, Stephan Milton Bache, Kirill M\u0026uuml;ller, Jeroen Ooms, David Robinson, Dana Paige Seidel, Vitalie Spinu, Kohske Takahashi, Davis Vaughan, Claus Wilke, Kara Woo, and Hiroaki Yutani, Welcome to the Tidyverse, The Journal of Open Source Software (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003evan der Reijd DJ, Chupetlovska K, van Dijk E, et al. Multi-sequence MRI radiomics of colorectal liver metastases: Which features are reproducible across readers?. \u003cem\u003eEur J Radiol\u003c/em\u003e. 2024;172:111346. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ejrad.2024.111346\u003c/span\u003e\u003cspan address=\"10.1016/j.ejrad.2024.111346\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXue C, Yuan J, Lo GG, et al. Radiomics feature reliability assessed by intraclass correlation coefficient: a systematic review. \u003cem\u003eQuant Imaging Med Surg\u003c/em\u003e. 2021;11(10):4431\u0026ndash;4460. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/qims-21-86\u003c/span\u003e\u003cspan address=\"10.21037/qims-21-86\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXu X, Li H, Wang S, et al. Multiplanar MRI-Based Predictive Model for Preoperative Assessment of Lymph Node Metastasis in Endometrial Cancer. \u003cem\u003eFront Oncol\u003c/em\u003e. 2019;9:1007. Published 2019 Oct 9. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2019.01007\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2019.01007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan BC, Ma XL, Li Y, Duan SF, Zhang GF, Qiang JW. MRI-Based Radiomics Nomogram for Selecting Ovarian Preservation Treatment in Patients With Early-Stage Endometrial Cancer. \u003cem\u003eFront Oncol\u003c/em\u003e. 2021;11:730281. Published 2021 Sep 9. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2021.730281\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2021.730281\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu XF, Yan BC, Li Y, Ma FH, Qiang JW. Radiomics Nomogram in Assisting Lymphadenectomy Decisions by Predicting Lymph Node Metastasis in Early-Stage Endometrial Cancer. \u003cem\u003eFront Oncol\u003c/em\u003e. 2022;12:894918. Published 2022 May 31. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2022.894918\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2022.894918\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJia Y, Hou L, Zhao J, et al. Radiomics analysis of multiparametric MRI for preoperative prediction of microsatellite instability status in endometrial cancer: a dual-center study. \u003cem\u003eFront Oncol\u003c/em\u003e. 2024;14:1333020. Published 2024 Jan 29. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2024.1333020\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2024.1333020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan R, Qin S, Xu J, et al. A comparison of 2D and 3D magnetic resonance imaging-based intratumoral and peritumoral radiomics models for the prognostic prediction of endometrial cancer: a pilot study. \u003cem\u003eCancer Imaging\u003c/em\u003e. 2024;24(1):100. Published 2024 Jul 31. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s40644-024-00743-2\u003c/span\u003e\u003cspan address=\"10.1186/s40644-024-00743-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu D, Huang J, Zhang Y, et al. Multimodal MRI-based radiomics models for the preoperative prediction of lymphovascular space invasion of endometrial carcinoma. \u003cem\u003eBMC Med Imaging\u003c/em\u003e. 2024;24(1):252. Published 2024 Sep 20. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12880-024-01430-1\u003c/span\u003e\u003cspan address=\"10.1186/s12880-024-01430-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGhosh A, Singh T, Singla V, Bagga R, Khandelwal N. Comparison of Absolute Apparent Diffusion Coefficient (ADC) Values in ADC Maps Generated Across Different Postprocessing Software: Reproducibility in Endometrial Carcinoma. \u003cem\u003eAJR Am J Roentgenol\u003c/em\u003e. 2017;209(6):1312\u0026ndash;1320. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2214/AJR.17.18002\u003c/span\u003e\u003cspan address=\"10.2214/AJR.17.18002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTraverso A, Kazmierski M, Shi Z, et al. Stability of radiomic features of apparent diffusion coefficient (ADC) maps for locally advanced rectal cancer in response to image pre-processing. \u003cem\u003ePhys Med\u003c/em\u003e. 2019;61:44\u0026ndash;51. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ejmp.2019.04.009\u003c/span\u003e\u003cspan address=\"10.1016/j.ejmp.2019.04.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichoux NF, Ceranka JW, Vandemeulebroucke J, et al. Repeatability and reproducibility of ADC measurements: a prospective multicenter whole-body-MRI study. \u003cem\u003eEur Radiol\u003c/em\u003e. 2021;31(7):4514\u0026ndash;4527. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00330-020-07522-0\u003c/span\u003e\u003cspan address=\"10.1007/s00330-020-07522-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFiset S, Welch ML, Weiss J, et al. Repeatability and reproducibility of MRI-based radiomic features in cervical cancer. \u003cem\u003eRadiother Oncol\u003c/em\u003e. 2019;135:107\u0026ndash;114. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.radonc.2019.03.001\u003c/span\u003e\u003cspan address=\"10.1016/j.radonc.2019.03.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu P, Wang J, Zhong H, et al. Reproducibility with repeat CT in radiomics study for rectal cancer. \u003cem\u003eOncotarget\u003c/em\u003e. 2016;7(44):71440\u0026ndash;71446. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18632/oncotarget.12199\u003c/span\u003e\u003cspan address=\"10.18632/oncotarget.12199\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003evan Timmeren JE, Leijenaar RTH, van Elmpt W, et al. Test-Retest Data for Radiomics Feature Stability Analysis: Generalizable or Study-Specific?. \u003cem\u003eTomography\u003c/em\u003e. 2016;2(4):361\u0026ndash;365. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18383/j.tom.2016.00208\u003c/span\u003e\u003cspan address=\"10.18383/j.tom.2016.00208\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGourtsoyianni S, Doumou G, Prezzi D, et al. Primary Rectal Cancer: Repeatability of Global and Local-Regional MR Imaging Texture Features. Radiology. 2017;284(2):552\u0026ndash;561. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1148/radiol.2017161375\u003c/span\u003e\u003cspan address=\"10.1148/radiol.2017161375\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSadeghi M, Abdalvand N, Mahdavi SR, et al. Magnetic Resonance Image Radiomic Reproducibility: The Impact of Preprocessing on Extracted Features from Gross and High-Risk Clinical Tumor Volumes in Cervical Cancer Patients before Brachytherapy. J Med Signals Sens. 2024;14:23. Published 2024 Aug 6. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4103/jmss.jmss_57_22\u003c/span\u003e\u003cspan address=\"10.4103/jmss.jmss_57_22\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKurata Y, Nishio M, Moribata Y, et al. Automatic segmentation of uterine endometrial cancer on multi-sequence MRI using a convolutional neural network. \u003cem\u003eSci Rep\u003c/em\u003e. 2021;11(1):14440. Published 2021 Jul 14. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-021-93792-7\u003c/span\u003e\u003cspan address=\"10.1038/s41598-021-93792-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShahzadi I, Zwanenburg A, Lattermann A, et al. Analysis of MRI and CT-based radiomics features for personalized treatment in locally advanced rectal cancer and external validation of published radiomics models. \u003cem\u003eSci Rep\u003c/em\u003e. 2022;12(1):10192. Published 2022 Jun 17. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-022-13967-8\u003c/span\u003e\u003cspan address=\"10.1038/s41598-022-13967-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 and 2 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":"Endometrial Cancer, MRI, Radiomics, Reproducibility, Segmentation","lastPublishedDoi":"10.21203/rs.3.rs-8184695/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8184695/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective:\u003c/strong\u003e The aim of our study is to evaluate the inter-observer agreement of radiomic features extracted from different sequences of multiparametric magnetic resonance imaging (mpMRI) in endometrial cancer, thereby contributing to the development of radiomic models with higher reproducibility and facilitating their integration into clinical practice.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMaterials and Methods:\u003c/strong\u003eThis single-center, retrospective study included 59 patients with pathologically confirmed, untreated endometrial cancer. The MRI scans of the patients were obtained from both in-house and external Picture Archiving and Communication System databases. Three radiologists segmented endometrial cancer tissue on three plane T2 weighted images, sagittal (T2WS), coronal (T2WC) and axial (T2WA), T1 weighted contrast enhanced sagittal (T1CE) images and diffusion weighted (DWI) images with the highest b-value. Apparent diffusion coefficient (ADC) masks were transferred from DWI segmentations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e The percentage of radiomic features with excellent agreement (ICC ≥ 0.90) was \u0026nbsp;58.9% for T2WS, 54.2% for T2WC, 44.9% for T2WA, 57.9% for T1CE, 54.2% for DWI, \u0026nbsp;and 45.8% for ADC. NGTDM and GLCM performed more poorly than the other \u0026nbsp;subclasses, 30% and 45.8% of features being reproducible, respectively. Shape \u0026nbsp;and GLDM performed better than other subclasses, 60.7% and 64.3% of features \u0026nbsp;being reproducible, respectively. Segmentations from T2WS had better \u0026nbsp;reproducibility despite similar levels of Dice Correlation Similarity (DSC) \u0026nbsp;compared to other planes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e Among T2-weighted planes, the highest inter-observer agreement was observed in the sagittal plane (T2WS), while the lowest was found in the axial plane (T2WA). Among all sequences, T2WS and T1CE showed the highest agreement, whereas T2WA and ADC demonstrated the lowest. In the subclasses of radiomic features, the highest agreement is observed in Shape and GLDM, while the lowest agreement is seen in NGTDM and GLCM. Overall robustness of MRI derived radiomics features to inter-reader segmentation differences in EC was moderate.\u003c/p\u003e","manuscriptTitle":"Assessment of Inter-observer Reproducibility of Radiomic Features in Multiparametric Magnetic Resonance Imaging of Endometrial Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-08 15:36:00","doi":"10.21203/rs.3.rs-8184695/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":"78883310-bfd9-4c2e-9987-d20452f827e3","owner":[],"postedDate":"December 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-27T15:12:11+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-08 15:36:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8184695","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8184695","identity":"rs-8184695","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}