{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,7]],"date-time":"2026-08-07T15:16:32Z","timestamp":1786115792904,"version":"build-2736575974"},"reference-count":31,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,6]],"date-time":"2026-08-06T00:00:00Z","timestamp":1785974400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["P01 CA261669"],"award-info":[{"award-number":["P01 CA261669"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>Radiation necrosis (RN) is an important complication of radiation therapy (RT) and is challenging to assess radiographically because lesions are prone to inconsistent delineation, often necessitating intracranial surgery to obtain a diagnosis. This study evaluated how established deep learning segmentation models can be retrained on an institutional RN cohort, assessed external generalization, and integrated the best components into RN-D3, an end-to-end pipeline for RN detection, differentiation, and delineation.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods and materials<\/jats:title>\n                    <jats:p>\n                      We trained models on a cohort of 52 patients with RN after proton beam RT treated at Massachusetts General Hospital (2004\u20132016) and tested these models on an external cohort of 29 patients with 39 RN lesions from the MOLAB brain metastasis dataset (photon-based RT, five Spanish institutions, 2005\u20132021). Six architectures were evaluated: four convolutional neural networks (nnU-Net, STU-Net, TRIAD PlainConv, and Spark3D) and two transformers (SwinUNETR and TRIAD SwinB). Training strategies included training from scratch, supervised pretraining, self-supervised pretraining, and foundation model initialization. Matched-encoder pairs enabled direct comparison of pretraining effects. Performance was evaluated by Dice similarity coefficient (DSC), surface Dice at 2\u202fmm, and lesion detection rate. A post-processing pipeline combining encoder features and shape radiomics was assessed for reducing false positives. The best segmentation model and this classifier were combined into RN-D\n                      <jats:sup>3<\/jats:sup>\n                      , an end-to-end pipeline evaluated on 115 brain metastasis (BM) lesions from MOLAB for RN-versus-non-RN differentiation.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      Spark3D and STU-Net achieved the highest external DSC (median 0.69 [IQR 0.50\u20130.84] and 0.65 [0.29\u20130.88]), followed by baseline nnU-Net (0.61 [0.39\u20130.83]). For Spark3D and STU-Net, large lesions (\u22651\u202fmL,\n                      <jats:italic>n<\/jats:italic>\n                      \u202f=\u202f21) were detected at 100% (median DSC 0.77 and 0.73) and small lesions (&amp;lt;1\u202fmL,\n                      <jats:italic>n<\/jats:italic>\n                      \u202f=\u202f18) at 78 and 67% (DSC 0.60 and 0.58). The remaining models detected 17\u201322% of small lesions with a median DSC near zero. False-positive filtering removed 2\/13 and 7\/14 false positives for Spark3D and STU-Net, retaining all true detections for Spark3D and 85% for STU-Net. RN-D\n                      <jats:sup>3<\/jats:sup>\n                      achieved an end-to-end sensitivity of 0.90 and a specificity of 0.88 on 115 BM lesions for RN-versus-non-RN differentiation. The two transformer-based architectures evaluated showed larger drops from internal to external cohorts (\u0394DSC \u22120.36 to \u22120.42 vs.\u202f\u2212\u202f0.12 to \u22120.20 for CNNs), though all architectures degraded externally.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>\n                      RN-D\n                      <jats:sup>3<\/jats:sup>\n                      integrates a CNN segmentation backbone with an RN-versus-non-RN classifier into an end-to-end pipeline addressing detection, differentiation, and delineation of RN. On external data, RN-D\n                      <jats:sup>3<\/jats:sup>\n                      achieved high end-to-end sensitivity and high specificity in patients with brain metastases, with detection of small lesions the main determinant of missed cases. Small-lesion detection remains the primary translational bottleneck.\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2026.1891883","type":"journal-article","created":{"date-parts":[[2026,8,6]],"date-time":"2026-08-06T05:31:51Z","timestamp":1785994311000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["RN-D3: detection, differentiation, and delineation of radiation necrosis on post-contrast MRI"],"prefix":"10.3389","volume":"9","author":[{"given":"Patrick","family":"Salome","sequence":"first","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]},{"name":"Clinical Cooperation Unit Translational Radiation Oncology, German Cancer Research Center (DKFZ)","place":["Heidelberg, Germany"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicol\u00f2","family":"Cogno","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hoyeon","family":"Lee","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]},{"name":"Department of Diagnostic Radiology and Centre of Cancer Medicine, LKS Faculty of Medicine, University of Hong Kong","place":["Hong Kong, Hong Kong SAR, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gregory","family":"Buti","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Keyur D.","family":"Shah","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]},{"name":"Department of Radiation Oncology and Winship Cancer Institute, Emory University","place":["Atlanta, Georgia"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Felix","family":"Ehret","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]},{"name":"Department of Radiation Oncology, Charit\u00e9\u2014Universit\u00e4tsmedizin Berlin, Freie Universit\u00e4t Berlin and Humboldt-Universit\u00e4t zu Berlin","place":["Berlin, Germany"]},{"name":"German Cancer Consortium (DKTK), Partner Site Berlin, a Partnership Between DKFZ and Charit\u00e9 \u2013 Universit\u00e4tsmedizin Berlin","place":["Berlin, Germany"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Amy Jordan","family":"Wisdom","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Helen A.","family":"Shih","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Harald","family":"Paganetti","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ibrahim","family":"Chamseddine","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology, Mass General Brigham Cancer Institute and Harvard Medical School","place":["Boston, MA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,6]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"66","DOI":"10.1007\/s11912-019-0818-y","article-title":"Cerebral radiation necrosis: incidence, pathogenesis, diagnostic challenges, and future opportunities","volume":"21","author":"Ali","year":"2019","journal-title":"Curr. Oncol. Rep."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"4113","DOI":"10.3390\/cancers15164113","article-title":"A multidimensional Connectomics- and Radiomics-based advanced machine-learning framework to distinguish radiation necrosis from true progression in brain metastases","volume":"15","author":"Cao","year":"2023","journal-title":"Cancers (Basel)"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"4424","DOI":"10.1158\/1078-0432.CCR-24-1215","article-title":"Decoding patient heterogeneity influencing radiation-induced brain necrosis","volume":"30","author":"Chamseddine","year":"2024","journal-title":"Clin. Cancer Res."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"842","DOI":"10.1016\/j.ijrobp.2025.09.059","article-title":"A mechanistic model of brain necrosis progression based on vascular heterogeneity","volume":"124","author":"Cogno","year":"2026","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"ref5","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1016\/j.radonc.2021.12.035","article-title":"Radiation induced contrast enhancement after proton beam therapy in patients with low grade glioma\u2014how safe are protons?","volume":"167","author":"Harrabi","year":"2022","journal-title":"Radiother. Oncol."},{"key":"ref6","doi-asserted-by":"publisher","first-page":"1152","DOI":"10.1093\/neuonc\/noad261","article-title":"Automated segmentation of ablated lesions using deep convolutional neural networks: a basis for response assessment following laser interstitial thermal therapy","volume":"26","author":"Haskell-Mendoza","year":"2024","journal-title":"Neuro-Oncology"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2201.01266","article-title":"Swin UNETR: swin transformers for semantic segmentation of brain tumors in MRI images","author":"Hatamizadeh","year":"2022","journal-title":"arXiv"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"218","DOI":"10.1186\/s12880-025-01796-w","article-title":"Deep learning for automated segmentation of radiation-induced changes in cerebral arteriovenous malformations following radiosurgery","volume":"25","author":"Ho","year":"2025","journal-title":"BMC Med. Imaging"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2304.06716","article-title":"STU-net: scalable and transferable medical image segmentation models empowered by large-scale supervised pre-training","author":"Huang","year":"2023","journal-title":"arXiv"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1038\/s41592-020-01008-z","article-title":"nnU-net: a self-configuring method for deep learning-based biomedical image segmentation","volume":"18","author":"Isensee","year":"2021","journal-title":"Nat. Methods"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"4952","DOI":"10.1002\/hbm.24750","article-title":"Automated brain extraction of multisequence MRI using artificial neural networks","volume":"40","author":"Isensee","year":"2019","journal-title":"Hum. Brain Mapp."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"S20","DOI":"10.1016\/j.ijrobp.2009.02.091","article-title":"Radiation dose-volume effects in the brain","volume":"76","author":"Lawrence","year":"2010","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"ref13","doi-asserted-by":"publisher","first-page":"654","DOI":"10.1038\/s41467-024-44824-z","article-title":"Segment anything in medical images","volume":"15","author":"Ma","year":"","journal-title":"Nat. Commun."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"654","DOI":"10.1038\/s41467-024-44824-z","article-title":"Segment anything in medical images","volume":"15","author":"Ma","year":"","journal-title":"Nat. Commun."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"109","DOI":"10.1016\/j.ijrobp.2020.08.058","article-title":"Brain necrosis in adult patients after proton therapy: is there evidence for dependency on linear energy transfer?","volume":"109","author":"Niemierko","year":"2021","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"ref16","doi-asserted-by":"publisher","first-page":"154","DOI":"10.1016\/j.radonc.2019.09.008","article-title":"Volumetric and actuarial analysis of brain necrosis in proton therapy using a novel mixture cure model","volume":"142","author":"Niyazi","year":"2020","journal-title":"Radiother. Oncol."},{"key":"ref17","doi-asserted-by":"publisher","first-page":"208","DOI":"10.1038\/s41597-023-02123-0","article-title":"A comprehensive dataset of annotated brain metastasis MR images with clinical and radiomic data","volume":"10","author":"Oca\u00f1a-Tienda","year":"2023","journal-title":"Sci. Data"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"101618","DOI":"10.1016\/j.eclinm.2022.101618","article-title":"Mortality of early treatment for radiation-induced brain necrosis in head and neck cancer survivors: a multicentre, retrospective, registry-based cohort study","volume":"52","author":"Pan","year":"2022","journal-title":"EClinicalMedicine"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"3029","DOI":"10.1093\/neuonc\/noaf090","article-title":"AI differentiates radionecrosis from true progression in brain metastasis upon stereotactic radiosurgery: analysis of 124 histologically assessed lesions","volume":"27","author":"Ressa","year":"2025","journal-title":"Neuro-Oncology"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1505.04597","article-title":"U-net: convolutional networks for biomedical image segmentation","author":"Ronneberger","year":"2015","journal-title":"arXiv"},{"key":"ref21","doi-asserted-by":"publisher","first-page":"5685","DOI":"10.1109\/JBHI.2024.3406256","article-title":"Brain tumor recurrence vs. radiation necrosis classification and patient survivability prediction","volume":"28","author":"Sadique","year":"2024","journal-title":"IEEE J. Biomed. Health Inform."},{"key":"ref22","doi-asserted-by":"publisher","first-page":"486","DOI":"10.1097\/COC.0000000000001036","article-title":"Differentiating radiation necrosis and metastatic progression in brain tumors using Radiomics and machine learning","volume":"46","author":"Salari","year":"2023","journal-title":"Am. J. Clin. Oncol."},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.3390\/cancers15061820","article-title":"MR-class: a Python tool for brain MR image classification utilizing one-vs-all DCNNs to Deal with the open-set recognition problem","volume":"15","author":"Salome","year":"","journal-title":"Cancer"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"965","DOI":"10.3390\/cancers15030965","article-title":"MR intensity normalization methods impact sequence specific Radiomics prognostic model performance in primary and recurrent high-grade glioma","volume":"15","author":"Salome","year":"","journal-title":"Cancer"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"1343","DOI":"10.1148\/rg.325125002","article-title":"Radiation necrosis in the brain: imaging features and differentiation from tumor recurrence","volume":"32","author":"Shah","year":"2012","journal-title":"Radiographics"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"e104","DOI":"10.1158\/0008-5472.CAN-17-0339","article-title":"Computational Radiomics system to decode the radiographic phenotype","volume":"77","author":"van Griethuysen","year":"2017","journal-title":"Cancer Res."},{"key":"ref27","doi-asserted-by":"publisher","first-page":"395","DOI":"10.3389\/fonc.2018.00395","article-title":"Diagnosis and Management of Radiation Necrosis in patients with brain metastases","volume":"8","author":"Vellayappan","year":"2018","journal-title":"Front. Oncol."},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2410.23132","article-title":"Revisiting MAE pre-training for 3D medical image segmentation","author":"Wald","year":"2025","journal-title":"arXiv"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"e230024","DOI":"10.1148\/ryai.230024","article-title":"TotalSegmentator: robust segmentation of 104 anatomic structures in CT images","volume":"5","author":"Wasserthal","year":"2023","journal-title":"Radiology"},{"key":"ref30","volume-title":"Triad: Vision Foundation Model for 3D Magnetic Resonance Imaging","author":"Yang","year":"2025"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"101840","DOI":"10.1016\/j.media.2020.101840","article-title":"Models Genesis","volume":"67","author":"Zhou","year":"2021","journal-title":"Med. Image Anal."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1891883\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,6]],"date-time":"2026-08-06T05:31:53Z","timestamp":1785994313000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1891883\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,6]]},"references-count":31,"alternative-id":["10.3389\/frai.2026.1891883"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1891883","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,6]]},"article-number":"1891883"}}