{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T22:33:58Z","timestamp":1790807638249,"version":"4.1.0"},"reference-count":30,"publisher":"Springer Science and Business Media LLC","issue":"6","license":[{"start":{"date-parts":[[2026,1,28]],"date-time":"2026-01-28T00:00:00Z","timestamp":1769558400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,1,28]],"date-time":"2026-01-28T00:00:00Z","timestamp":1769558400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Eur Radiol"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>To evaluate the diagnostic performance of semi-supervised learning models for aggressive prostate cancer detection on MRI compared to fully supervised models trained with additional expert annotations.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Materials and methods<\/jats:title>\n                    <jats:p>\n                      We used 1500 MRI scans from the PI-CAI challenge training subset. Positive scans had 220 human and 205 AI-generated annotations. The mtU-Net (proposed teacher-student semi-supervised approach) was compared to supervised (trained using only 220 human annotations) and semi-supervised (trained on human and AI-generated annotations) nnU-Net. The 205 AI-annotated scans were manually annotated, and a fully supervised model was trained. External validation was performed on a newly annotated dataset from the PROMIS study (\n                      <jats:italic>n<\/jats:italic>\n                      \u2009=\u2009574,\u00a0403 lesions) and the Prostate158 dataset (\n                      <jats:italic>n<\/jats:italic>\n                      \u2009=\u2009158,\u00a0126 lesions). Patient-level performance was evaluated using the area under the curve (AUC) and lesion-level detection (overlap &gt;\u20090.10) using average precision (AP), along with 95% confidence Intervals (in brackets), and the DeLong test to compare AUCs against the supervised and fully supervised models.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      The fully supervised nnU-Net showed the highest performance on the internal PI-CAI test set (AUC\u2009=\u20090.89 [0.87\u20130.91], AP\u2009=\u20090.65 [0.60\u20130.70]) and external validation datasets PROMIS (AUC\u2009=\u20090.68 [0.64\u20130.72], AP\u2009=\u20090.24 [0.20\u20130.29]) and Prostate158 (AUC\u2009=\u20090.87 [0.82\u20130.92], AP\u2009=\u20090.64 [0.56\u20130.72]), significantly outperforming the supervised baseline (\n                      <jats:italic>p<\/jats:italic>\n                      \u2009&lt;\u20090.0\u00a05). The proposed semi-supervised mtU-Net demonstrated close external validation performance on PROMIS (AUC\u2009=\u20090.66 [0.62\u20130.71], AP\u2009=\u20090.20 [0.16\u20130.25]) and Prostate158 (AUC\u2009=\u20090.86 [0.81\u20130.92], AP\u2009=\u20090.58 [0.49\u20130.67]), significantly outperforming the supervised baseline on both datasets (\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.047 and\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.014, respectively), and showing no significant difference to the fully supervised model (\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.199 and\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.702, respectively).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>In prostate MRI tumor detection, fully supervised learning performed best. However, in external validation, the semi-supervised methods demonstrated performance that approached that of the fully supervised model, proving a valuable approach when expert annotations are limited.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Key Points<\/jats:title>\n                    <jats:p>\n                      <jats:bold>\n                        <jats:italic>Question<\/jats:italic>\n                      <\/jats:bold>\n                      <jats:italic>The need for extensive expert voxel-level annotations delays the development of AI-based prostate cancer diagnostic tools and their implementation in clinical practice<\/jats:italic>\n                      .\n                    <\/jats:p>\n                    <jats:p>\n                      <jats:bold>\n                        <jats:italic>Findings<\/jats:italic>\n                      <\/jats:bold>\n                      <jats:italic>The combination of pseudo-labeling with consistency regularization achieved performance comparable to that of fully supervised methods, demonstrating that data diversity matches the impact of expert annotation volume<\/jats:italic>\n                      .\n                    <\/jats:p>\n                    <jats:p>\n                      <jats:bold>\n                        <jats:italic>Clinical relevance<\/jats:italic>\n                      <\/jats:bold>\n                      <jats:italic>Semi-supervised learning reduces dependence on expert annotations while maintaining detection accuracy, enabling the development of scalable, automated diagnostic tools for prostate cancer amid growing clinical workflow demands<\/jats:italic>\n                      .\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.1007\/s00330-026-12324-x","type":"journal-article","created":{"date-parts":[[2026,1,28]],"date-time":"2026-01-28T18:43:11Z","timestamp":1769625791000},"page":"5011-5021","update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":4,"title":["Semi-supervised learning in prostate MRI tumor detection approaches fully supervised performance on external validation"],"prefix":"10.1007","volume":"36","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4661-9920","authenticated-orcid":false,"given":"Eduardo H. P.","family":"Pooch","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Georgios","family":"Agrotis","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lishan","family":"Cai","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mark","family":"Emberton","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Taimur T.","family":"Shah","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hashim U.","family":"Ahmed","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Regina G. H.","family":"Beets-Tan","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sean","family":"Benson","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tomas","family":"Janssen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ivo G.","family":"Schoots","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,1,28]]},"reference":[{"key":"12324_CR1","first-page":"394","volume":"68","author":"F Bray","year":"2018","unstructured":"Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A (2018) Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 68:394\u2013424","journal-title":"CA Cancer J Clin"},{"key":"12324_CR2","doi-asserted-by":"publisher","first-page":"384","DOI":"10.2214\/AJR.18.20796","volume":"213","author":"F Giganti","year":"2019","unstructured":"Giganti F, Rosenkrantz AB, Villeirs G et al (2019) The evolution of MRI of the prostate: the past, the present, and the future. Am J Roentgenol 213:384\u2013396","journal-title":"Am J Roentgenol"},{"key":"12324_CR3","doi-asserted-by":"crossref","unstructured":"Bass E, Pantovic A, Connor M et al (2021) A systematic review and meta-analysis of the diagnostic accuracy of biparametric prostate MRI for prostate cancer in men at risk. Prostate Cancer Prostatic Dis 24:596\u2013611","DOI":"10.1038\/s41391-020-00298-w"},{"key":"12324_CR4","doi-asserted-by":"publisher","first-page":"76","DOI":"10.1148\/radiol.2020190646","volume":"296","author":"AC Westphalen","year":"2020","unstructured":"Westphalen AC, McCulloch CE, Anaokar JM et al (2020) Variability of the positive predictive value of PI-RADS for prostate MRI across 26 centers: experience of the Society of Abdominal Radiology prostate cancer disease-focused panel. Radiology 296:76\u201384","journal-title":"Radiology"},{"key":"12324_CR5","doi-asserted-by":"crossref","unstructured":"Soerensen SJC, Fan RE, Seetharaman A et al (2021) Sonn, Deep learning improves speed and accuracy of prostate gland segmentations on magnetic resonance imaging for targeted biopsy. J Urol 206:604\u2013612","DOI":"10.1097\/JU.0000000000001783"},{"key":"12324_CR6","unstructured":"Lee D-H (2013) Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks. In: Workshop on Challenges in Representation Learning, ICML 3:896"},{"key":"12324_CR7","unstructured":"Luo X, Hu M, Song T, Wang G, Zhang S (2022) Semi-supervised medical image segmentation via cross teaching between CNN and transformer. In: International Conference on Medical Imaging with Deep Learning. PMLR, pp. 820\u2013833"},{"key":"12324_CR8","doi-asserted-by":"crossref","unstructured":"Luo X, Chen J, Song T, Wang G (2021) Semi-supervised medical image segmentation through dual-task consistency. pp. 8801\u20138809. https:\/\/arxiv.org\/abs\/2009.04448","DOI":"10.1609\/aaai.v35i10.17066"},{"key":"12324_CR9","doi-asserted-by":"publisher","first-page":"102517","DOI":"10.1016\/j.media.2022.102517","volume":"80","author":"X Luo","year":"2022","unstructured":"Luo X, Wang G, Liao W et al (2022) Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency. Med Image Anal 80:102517","journal-title":"Med Image Anal"},{"key":"12324_CR10","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1038\/s41592-020-01008-z","volume":"18","author":"F Isensee","year":"2020","unstructured":"Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH (2020) nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18:203\u2013211","journal-title":"Nat Methods"},{"key":"12324_CR11","doi-asserted-by":"crossref","unstructured":"Saha A, Bosma JS, Twilt JJ et al (2024) Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study. Lancet Oncol 25:879\u2013887","DOI":"10.1016\/S1470-2045(24)00220-1"},{"key":"12324_CR12","doi-asserted-by":"crossref","unstructured":"Bosma JS, Saha A, Hosseinzadeh M et al (2023) Semisupervised learning with report-guided pseudo labels for deep learning\u2013based prostate cancer detection using biparametric MRI. Radiol Artif Intell 5:e230031","DOI":"10.1148\/ryai.230031"},{"key":"12324_CR13","doi-asserted-by":"crossref","unstructured":"Ronneberger O, Fischer P, Brox T (2015) U-Net: Convolutional networks for biomedical image segmentation. Med Image Comput Comput Assist Interv 9351:234\u2013241","DOI":"10.1007\/978-3-319-24574-4_28"},{"key":"12324_CR14","doi-asserted-by":"publisher","first-page":"815","DOI":"10.1016\/S0140-6736(16)32401-1","volume":"389","author":"HU Ahmed","year":"2017","unstructured":"Ahmed HU, Bosaily AE-S, Brown LC et al (2017) Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet 389:815\u2013822","journal-title":"Lancet"},{"key":"12324_CR15","doi-asserted-by":"publisher","first-page":"105817","DOI":"10.1016\/j.compbiomed.2022.105817","volume":"148","author":"LC Adams","year":"2022","unstructured":"Adams LC, Makowski MR, Engel G et al (2022) Prostate158\u2014an expert-annotated 3T MRI dataset and algorithm for prostate cancer detection. Comput Biol Med 148:105817","journal-title":"Comput Biol Med"},{"key":"12324_CR16","first-page":"530","volume":"2021","author":"M Baumgartner","year":"2021","unstructured":"Baumgartner M, J\u00e4ger PF, Isensee F, Maier-Hein KH (2021) nnDetection: a self-configuring method for medical object detection. Med Image Comput Comput Assist Intervent 2021:530\u2013539","journal-title":"Med Image Comput Comput Assist Intervent"},{"key":"12324_CR17","unstructured":"Jaeger P, Kohl SAA, Bickelhaupt S et al (2018) Retina U-Net: embarrassingly simple exploitation of segmentation supervision for medical object detection, ML4H@NeurIPS. https:\/\/arxiv.org\/abs\/1811.08661"},{"key":"12324_CR18","doi-asserted-by":"crossref","unstructured":"Lin T-Y, Goyal P, Girshick R, He K, Doll\u00e1r P (2017) Focal loss for dense object detection. In: Proceedings of the IEEE international conference on computer vision. IEEE. pp. 2980\u20132988","DOI":"10.1109\/ICCV.2017.324"},{"key":"12324_CR19","unstructured":"Tarvainen A, Valpola H (2017) Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. In: Advances in Neural Information Processing Systems. Curran Associates, Inc. 30"},{"key":"12324_CR20","unstructured":"Cai L, Li Y, Fan X et al (2024) Low-contrast-enhanced contrastive learning for semi-supervised endoscopic image segmentation. https:\/\/arxiv.org\/abs\/2412.02314"},{"key":"12324_CR21","unstructured":"Xiang J, Li Z, Wang W, Xia Q, Zhang S (2021) Self-ensembling contrastive learning for semi-supervised medical image segmentation. https:\/\/arxiv.org\/abs\/2105.12924"},{"key":"12324_CR22","doi-asserted-by":"crossref","unstructured":"Wang K, Zhan B, Zu C et al (2022) Semi-supervised medical image segmentation via a tripled-uncertainty guided mean teacher model with contrastive learning. Med Image Anal 79:102447","DOI":"10.1016\/j.media.2022.102447"},{"key":"12324_CR23","doi-asserted-by":"crossref","unstructured":"Cui W, Liu Y, Li Y, et al (2019) Semi-supervised brain lesion segmentation with an adapted mean teacher model. In: Lecture Notes in Computer Science. Springer International Publishing, Cham, pp 554\u2013565","DOI":"10.1007\/978-3-030-20351-1_43"},{"key":"12324_CR24","doi-asserted-by":"publisher","first-page":"e0300039","DOI":"10.1371\/journal.pone.0300039","volume":"19","author":"B Li","year":"2024","unstructured":"Li B, Xu Y, Wang Y, Li L, Zhang B (2024) The student teacher framework guided by self-training and consistency regularization for semi-supervised medical image segmentation. PLoS One 19:e0300039. https:\/\/doi.org\/10.1371\/journal.pone.0300039","journal-title":"PLoS One"},{"key":"12324_CR25","unstructured":"Lindeijer TN, Ytredal TM, Eftest\u00f8l T et al (2023) Leveraging multi-view data without annotations for prostate MRI segmentation: a contrastive approach. https:\/\/arxiv.org\/abs\/2308.06477"},{"key":"12324_CR26","doi-asserted-by":"publisher","first-page":"104622","DOI":"10.1016\/j.bspc.2023.104622","volume":"83","author":"Z Li","year":"2023","unstructured":"Li Z, Fang J, Qiu R et al (2023) Cda-net: a contrastive deep adversarial model for prostate cancer segmentation in MRI images. Biomed Signal Process Control 83:104622","journal-title":"Biomed Signal Process Control"},{"key":"12324_CR27","unstructured":"Jeong J, Lee S, Kim J, Kwak N (2019) Consistency-based Semi-supervised Learning for Object detection. In: Advances in Neural Information Processing Systems. Curran Associates, Inc. 32"},{"key":"12324_CR28","doi-asserted-by":"publisher","unstructured":"Singh P, Chukkapalli R, Chaudhari S et al (20241) Shifting to machine supervision: annotation-efficient semi and self-supervised502learning for automatic medical image segmentation and classification. Sci Rep 14:10820. https:\/\/doi.org\/10.1038\/s41598-024-61822-9","DOI":"10.1038\/s41598-024-61822-9"},{"key":"12324_CR29","doi-asserted-by":"crossref","unstructured":"Hering A, de Boer S, Saha A et al (2024) Deformable MRI sequence registration for AI-based prostate cancer diagnosis. In: International Workshop on Biomedical Image Registration. Lecture Notes in Computer Science. Springer, pp. 148\u2013162","DOI":"10.1007\/978-3-031-73480-9_12"},{"key":"12324_CR30","doi-asserted-by":"crossref","unstructured":"Chen A, Lay N, Harmon S et al (2024) Location-based radiology report guided semi-supervised learning for prostate cancer detection. In: 2024 IEEE International Symposium on Biomedical Imaging (ISBI), IEEE. pp. 1\u20135","DOI":"10.1109\/ISBI56570.2024.10635907"}],"container-title":["European Radiology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s00330-026-12324-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s00330-026-12324-x","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s00330-026-12324-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,27]],"date-time":"2026-05-27T05:51:12Z","timestamp":1779861072000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s00330-026-12324-x"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,28]]},"references-count":30,"journal-issue":{"issue":"6","published-online":{"date-parts":[[2026,6]]}},"alternative-id":["12324"],"URL":"https:\/\/doi.org\/10.1007\/s00330-026-12324-x","relation":{},"ISSN":["1432-1084"],"issn-type":[{"value":"1432-1084","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,28]]},"assertion":[{"value":"18 August 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"31 December 2025","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"28 January 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Compliance with ethical standards"}},{"value":"The scientific guarantor of this publication is Ivo Schoots.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Guarantor"}},{"value":"R.G.H.B.T. is a member of the scientific editorial board of European Radiology (section: Oncology) and, as such, did not participate in the selection or review processes for this article. The remaining authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Conflict of interest"}},{"value":"Not applicable.","order":4,"name":"Ethics","group":{"name":"EthicsHeading","label":"Statistics and biometry"}},{"value":"Written informed consent was not required for this study because we used publicly available datasets.","order":5,"name":"Ethics","group":{"name":"EthicsHeading","label":"Informed consent"}},{"value":"Institutional Review Board approval was not required because we use publicly available datasets.","order":6,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethical approval"}},{"value":"This study utilizes publicly available datasets (PI-CAI, PROMIS, Prostate158) and newly generated annotations on subsets of these data. To the best of our knowledge, there is no subject overlap with prior publications or work currently under review from our group using these specific patient cohorts for different analyses. The novelty of this work lies in the systematic comparison of supervised, semi-supervised (including our proposed mtU-Net), and fully supervised DL strategies for aggressive prostate cancer segmentation using these datasets, particularly focusing on performance generalizability via external validation.","order":7,"name":"Ethics","group":{"name":"EthicsHeading","label":"Study subjects or cohorts overlap"}},{"order":8,"name":"Ethics","group":{"name":"EthicsHeading","label":"Methodology"}}]}}