{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,7]],"date-time":"2026-05-07T16:02:21Z","timestamp":1778169741470,"version":"3.51.4"},"reference-count":49,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T00:00:00Z","timestamp":1760486400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:p>Ovarian cancer remains the deadliest gynecologic malignancy, and transvaginal ultrasound (TVS), the first-line test, still suffers from limited specificity and operator dependence. We introduce a learned early-fusion (joint projection) hybrid that couples EfficientNet-B7 (local descriptors) with a Swin Transformer (hierarchical global context) to classify eight ovarian tumor categories from 2D TVS. Using the public, de-identified OTU-2D dataset (<jats:italic>n<\/jats:italic>\u202f=\u202f1,469 images across eight histopathologic classes), we conducted patient-level, stratified 5-fold cross-validation repeated 10\u00d7. To address class imbalance while preventing leakage, training used train-only oversampling, ultrasound-aware augmentations, and strong regularization; validation\/test folds were never resampled. The hybrid achieved AUC 0.9904, accuracy 92.13%, sensitivity 92.38%, and specificity 98.90%, outperforming single CNN or ViT baselines. A soft ensemble of the top hybrids further improved performance to AUC 0.991, accuracy 93.3%, sensitivity 93.6%, and specificity 99.0%. Beyond discrimination, we provide deployment-oriented evaluation: isotonic calibration yielded reliable probabilities, decision-curve analysis showed net clinical benefit across 5\u201320% risk thresholds, entropy-based uncertainty supported confidence-based triage, and Grad-CAM highlighted clinically salient regions. All metrics are reported with 95% bootstrap confidence intervals, and the evaluation protocol preserves real-world data distributions. Taken together, this work advances ovarian ultrasound AI from accuracy-only reporting to calibrated, explainable, and uncertainty-aware decision support, offering a reproducible reference framework for multiclass ovarian ultrasound and a clear path toward clinical integration and prospective validation.<\/jats:p>","DOI":"10.3389\/frai.2025.1679310","type":"journal-article","created":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T05:43:29Z","timestamp":1760507009000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":3,"title":["Early-fusion hybrid CNN-transformer models for multiclass ovarian tumor ultrasound classification"],"prefix":"10.3389","volume":"8","author":[{"given":"Igor","family":"Garcia-Atutxa","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jos\u00e9","family":"Mart\u00ednez-M\u00e1s","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Andr\u00e9s","family":"Bueno-Crespo","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Francisca","family":"Villanueva-Flores","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2025,10,15]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"1574","DOI":"10.1002\/jcu.24048","article-title":"Comparison of the diagnostic performance of ovarian adnexal reporting data system (O-RADS) with IOTA simple rules and ADNEX model for classifying adnexal masses: a head-to-head meta-analysis","volume":"53","author":"Almeida","year":"2025","journal-title":"J. Clin. Ultrasound"},{"key":"ref2","doi-asserted-by":"publisher","first-page":"168","DOI":"10.1148\/radiol.2019191150","article-title":"O-RADS US risk stratification and management system: a consensus guideline from the ACR ovarian-adnexal reporting and data system committee","volume":"294","author":"Andreotti","year":"2020","journal-title":"Radiology"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1038\/s41698-024-00527-8","article-title":"Machine learning and radiomics for segmentation and classification of adnexal masses on ultrasound","volume":"8","author":"Barcroft","year":"2024","journal-title":"NPJ Precis. Oncol."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"3820","DOI":"10.3390\/cancers16223820","article-title":"Accuracy of O-RADS system in differentiating between benign and malignant adnexal masses assessed via external validation by inexperienced gynecologists","volume":"16","author":"Buranaworathitikul","year":"2024","journal-title":"Cancer"},{"key":"ref5","volume-title":"TransUNet: transformers make strong encoders for medical image segmentation","author":"Chen","year":"2021"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"155","DOI":"10.1002\/uog.23530","article-title":"Ultrasound image analysis using deep neural networks for discriminating between benign and malignant ovarian tumors: comparison with expert subjective assessment","volume":"57","author":"Christiansen","year":"2021","journal-title":"Ultrasound Obstet. Gynecol."},{"key":"ref7","doi-asserted-by":"publisher","first-page":"189","DOI":"10.1038\/s41591-024-03329-4","article-title":"International multicenter validation of AI-driven ultrasound detection of ovarian cancer","volume":"31","author":"Christiansen","year":"2025","journal-title":"Nat. Med."},{"key":"ref8","first-page":"54","article-title":"C\u00e1ncer De Ovario: Tamizaje Y Diagn\u00f3stico Imagenol\u00f3gico","volume":"37","author":"Cort\u00e9s Morera","year":"2020","journal-title":"Med. Leg. Costa Rica"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"102923","DOI":"10.1016\/j.eclinm.2024.102923","article-title":"Development and validation of a deep learning pipeline to diagnose ovarian masses using ultrasound screening: a retrospective multicenter study","volume":"78","author":"Dai","year":"2024","journal-title":"Eclinicalmedicine"},{"key":"ref10","first-page":"248","article-title":"Imagenet: a large-scale hierarchical image database","author":"Deng","year":"2009"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"101376","DOI":"10.1016\/j.gore.2024.101376","article-title":"Ovarian cancer think tank: an overview of the current status of ovarian cancer screening and recommendations for future directions","volume":"53","author":"Dexter","year":"2024","journal-title":"Gynecol. Oncol. Rep."},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2504.08481","article-title":"A hybrid fully convolutional CNN-transformer model for inherently interpretable medical image classification","author":"Djoumessi","year":"2025","journal-title":"arXiv"},{"key":"ref13","author":"Dosovitskiy","year":"2021"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"89","DOI":"10.1186\/s12880-024-01251-2","article-title":"Ultrasound-based deep learning radiomics model for differentiating benign, borderline, and malignant ovarian tumours: a multi-class classification exploratory study","volume":"24","author":"Du","year":"2024","journal-title":"BMC Med. Imaging"},{"key":"ref15","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1186\/s12938-024-01234-y","article-title":"Development and validation of an ultrasound-based deep learning radiomics nomogram for predicting the malignant risk of ovarian tumours","volume":"23","author":"Du","year":"2024","journal-title":"Biomed. Eng. Online"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"607","DOI":"10.1186\/s12885-024-12331-5","article-title":"Reference-free inferring of transcriptomic events in cancer cells on single-cell data","volume":"24","author":"Eralp","year":"2024","journal-title":"BMC Cancer"},{"key":"ref17","volume-title":"Global cancer observatory: cancer today","author":"Ferlay","year":"2024"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"E179","DOI":"10.1016\/S2589-7500(21)00278-8","article-title":"Deep learning-enabled pelvic ultrasound images for accurate diagnosis of ovarian cancer in China: a retrospective, multicentre, diagnostic study","volume":"4","author":"Gao","year":"2022","journal-title":"Lancet Digit. Health"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"1474930","DOI":"10.3389\/fonc.2024.1474930","article-title":"The diagnostic performance of international ovarian tumor analysis: simple rules for diagnosing ovarian tumors\u2014a systematic review and meta-analysis","volume":"14","author":"Gareeballah","year":"2025","journal-title":"Front. Oncol."},{"key":"ref20","doi-asserted-by":"publisher","first-page":"TC06","DOI":"10.7860\/JCDR\/2017\/26790.10353","article-title":"Evaluation of IOTA simple ultrasound rules to distinguish benign and malignant ovarian tumours","volume":"11","author":"Garg","year":"2017","journal-title":"J. Clin. Diagn. Res."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"4123","DOI":"10.3390\/jcm13144123","article-title":"Enhancing ovarian tumor diagnosis: performance of convolutional neural networks in classifying ovarian masses using ultrasound images","volume":"13","author":"Giourga","year":"2024","journal-title":"J. Clin. Med."},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1512.03385","article-title":"Deep residual learning for image recognition","author":"He","year":"2015","journal-title":"arXiv"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"406","DOI":"10.3390\/diagnostics15040406","article-title":"Early diagnosis of ovarian cancer: a comprehensive review of the advances, challenges, and future directions","volume":"15","author":"Hong","year":"2025","journal-title":"Diagnostics (Basel)"},{"key":"ref24","first-page":"2261","article-title":"Densely connected convolutional networks","author":"Huang","year":"2017"},{"key":"ref25","year":"2025"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"8747","DOI":"10.21037\/qims-23-542","article-title":"Transformers in medical image segmentation: a narrative review","volume":"13","author":"Khan","year":"2023","journal-title":"Quant. Imaging Med. Surg."},{"key":"ref27","doi-asserted-by":"publisher","first-page":"2024.06.21.24309265","DOI":"10.1101\/2024.06.21.24309265","article-title":"Systematic review of hybrid vision transformer architectures for radiological image analysis","author":"Kim","year":"2024","journal-title":"medRxiv"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"280","DOI":"10.3322\/caac.21559","article-title":"Epithelial ovarian cancer: evolution of management in the era of precision medicine","volume":"69","author":"Lheureux","year":"2019","journal-title":"CA Cancer J. Clin."},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2103.14030","article-title":"Swin transformer: hierarchical vision transformer using shifted windows","author":"Liu","year":"2021","journal-title":"arXiv"},{"key":"ref30","volume-title":"Guidance on qualification and classification of software in regulation (EU) 2017\/745\u2014MDR and regulation (EU) 2017\/746\u2014IVDR (MDCG 2019\u201311 rev. 1)","year":"2025"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"909","DOI":"10.1097\/AOG.0000000000002580","article-title":"Ovarian cancer prevention and screening","volume":"131","author":"Menon","year":"2018","journal-title":"Obstet. Gynecol."},{"key":"ref32","doi-asserted-by":"publisher","first-page":"195","DOI":"10.3390\/info16030195","article-title":"Deep convolutional neural networks in medical image analysis: a review","volume":"16","author":"Mienye","year":"2025","journal-title":"Information"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"E15734056326685","DOI":"10.2174\/0115734056326685250101113959","article-title":"Enhanced pneumonia detection in chest X-rays using hybrid convolutional and vision transformer networks","volume":"21","author":"Mustapha","year":"2025","journal-title":"Curr. Med. Imaging"},{"key":"ref34","doi-asserted-by":"publisher","first-page":"330","DOI":"10.1089\/cmb.2024.0807","article-title":"DRGAT: predicting drug responses via diffusion-based graph attention network","volume":"32","author":"Sefer","year":"2025","journal-title":"J. Comput. Biol."},{"key":"ref35","first-page":"618","article-title":"Grad-CAM: visual explanations from deep networks via gradient-based localization","author":"Selvaraju","year":"2017"},{"key":"ref36","doi-asserted-by":"publisher","first-page":"264","DOI":"10.5694\/mja2.52227","article-title":"Screening and prevention of ovarian cancer","volume":"220","author":"Sideris","year":"2024","journal-title":"Med. J. Aust."},{"key":"ref37","doi-asserted-by":"publisher","first-page":"7","DOI":"10.3322\/caac.21654","article-title":"Cancer statistics, 2021","volume":"71","author":"Siegel","year":"2021","journal-title":"CA Cancer J. Clin."},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1905.11946","article-title":"EfficientNet: rethinking model scaling for convolutional neural networks","author":"Tan","year":"2020","journal-title":"arXiv"},{"key":"ref39","doi-asserted-by":"publisher","first-page":"1581157","DOI":"10.3389\/fonc.2025.1581157","article-title":"Advancements in artificial intelligence for ultrasound diagnosis of ovarian cancer: a comprehensive review","volume":"15","author":"Tang","year":"2025","journal-title":"Front. Oncol."},{"key":"ref40","doi-asserted-by":"publisher","first-page":"1467","DOI":"10.3390\/cancers16081467","article-title":"Imaging of peritoneal metastases in ovarian cancer using MDCT, MRI, and FDG PET\/CT: a systematic review and meta-analysis","volume":"16","author":"Tsili","year":"2024","journal-title":"Cancer"},{"key":"ref41","year":"2025"},{"key":"ref42","doi-asserted-by":"publisher","first-page":"25","DOI":"10.2147\/IJWH.S38347","article-title":"Transvaginal ultrasonography in ovarian cancer screening: current perspectives","volume":"6","author":"Van Nagell","year":"2013","journal-title":"Int. J. Women\u2019s Health"},{"key":"ref43","doi-asserted-by":"publisher","first-page":"673","DOI":"10.3390\/diagnostics13040673","article-title":"O-RADS classification for ultrasound assessment of adnexal masses: agreement between IOTA lexicon and ADNEX model for assigning risk group","volume":"13","author":"Vara","year":"2023","journal-title":"Diagnostics (Basel)"},{"key":"ref44","doi-asserted-by":"publisher","first-page":"890809","DOI":"10.3389\/fcvm.2022.890809","article-title":"Real-world and regulatory perspectives of artificial intelligence in cardiovascular imaging","volume":"9","author":"Wellnhofer","year":"2022","journal-title":"Front. Cardiovasc. Med."},{"key":"ref9008","doi-asserted-by":"publisher","DOI":"10.3389\/fonc.2023.1154200","article-title":"Deep convolutional neural networks for multiple histologic types of ovarian tumors classification in ultrasound images","volume":"13","author":"Wu","year":"2023"},{"key":"ref45","first-page":"395","article-title":"Deep learning for ovarian tumor classification with ultrasound images","author":"Wu","year":"2018"},{"key":"ref46","doi-asserted-by":"publisher","first-page":"346","DOI":"10.1007\/s00432-024-05872-6","article-title":"Developing a deep learning model for predicting ovarian Cancer in ovarian-adnexal reporting and data system ultrasound (O-RADS US) category 4 lesions: a multicenter study","volume":"150","author":"Xie","year":"2024","journal-title":"J. Cancer Res. Clin. Oncol."},{"key":"ref47","volume-title":"Transfuse: fusing transformers and CNNS for medical image segmentation","author":"Zhang","year":"2021"},{"key":"ref48","doi-asserted-by":"publisher","first-page":"1432987","DOI":"10.3389\/fphys.2024.1432987","article-title":"A dual-branch and dual attention transformer and CNN hybrid network for ultrasound image segmentation","volume":"15","author":"Zhang","year":"2024","journal-title":"Front. Physiol."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1679310\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T05:43:31Z","timestamp":1760507011000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1679310\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,15]]},"references-count":49,"alternative-id":["10.3389\/frai.2025.1679310"],"URL":"https:\/\/doi.org\/10.3389\/frai.2025.1679310","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,15]]},"article-number":"1679310"}}