{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,31]],"date-time":"2026-07-31T11:58:14Z","timestamp":1785499094697,"version":"3.56.0"},"reference-count":61,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,31]],"date-time":"2026-07-31T00:00:00Z","timestamp":1785456000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100004329","name":"The Slovenian Research and Innovation Agency","doi-asserted-by":"publisher","award":["J2-3059"],"award-info":[{"award-number":["J2-3059"]}],"id":[{"id":"10.13039\/501100004329","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100004329","name":"The Slovenian Research and Innovation Agency","doi-asserted-by":"publisher","award":["P2-0232"],"award-info":[{"award-number":["P2-0232"]}],"id":[{"id":"10.13039\/501100004329","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Med."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Inter-scanner variability in magnetic resonance imaging (MRI) adversely affects the diagnostic and prognostic quality of scans and necessitates the development of models that are robust to domain shift arising from the unseen scanner data. A review of recent advances in domain adaptation and domain generalization showed that the efficacy of strategies involving modifications or constraints on the latent space appears to be contingent upon the level and\/or depth of supervision during model training.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We propose an adaptive multi-stage domain unlearning (ADMU) technique to improve robustness to unseen scanner domains. Building on the state-of-the-art segmentation framework nnU-Net, we employ deep supervision at deep encoder stages by applying domain classifier unlearning, sequentially across these stages to reduce domain-discriminative latent features. Following the self-configurable approach of nnU-Net, the auxiliary feedback loop implements an adaptive backpropagation schedule for unlearning. Experiments were conducted on four public datasets (one for training, three for testing) to benchmark white-matter lesion segmentation methods. Five benchmark models and\/or strategies, spanning passive to active domain-robust training strategies, were tested, and five state-of-the-art methods were compared.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>AMDU demonstrated consistent, robust and improved cross-dataset segmentation performance on three test sets versus baseline nn-Unet variants. The advantage of AMDU was in enhanced lesion sensitivity with balanced false detections, resulting in good overall segmentation quality, as measured by segmentation overlap and relative lesion volume error. Compared to continuous domain unlearning, the adaptive scheduling balanced the adverse impact of unlearning onto the main segmentation task. Intensity-based preprocessing was found to be detrimental to segmentation performance.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>\n                      The proposed AMDU strategy was shown to be complementary to data augmentation. Demonstrated for white-matter lesion segmentation it relied only the FLAIR modality, simplifying preprocessing to spatial normalization to brain atlas, with no intensity harmonization, for best cross-dataset segmentation performance. The source code is available at\n                      <jats:ext-link>https:\/\/github.com\/Pubec\/nnunetv2-unlearning.<\/jats:ext-link>\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fmed.2026.1875760","type":"journal-article","created":{"date-parts":[[2026,7,31]],"date-time":"2026-07-31T11:10:27Z","timestamp":1785496227000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Adaptive multi-stage domain unlearning for white-matter lesion segmentation"],"prefix":"10.3389","volume":"13","author":[{"given":"Domen","family":"Prelo\u017enik","sequence":"first","affiliation":[{"name":"Faculty of Electrical Engineering, University of Ljubljana","place":["Ljubljana, Slovenia"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"\u017diga","family":"\u0160piclin","sequence":"additional","affiliation":[{"name":"Faculty of Electrical Engineering, University of Ljubljana","place":["Ljubljana, Slovenia"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,31]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s11831-020-09403-7","article-title":"State-of-the-art segmentation techniques and future directions for multiple sclerosis brain lesions","volume":"28","author":"Kaur","year":"2020","journal-title":"Arch Comput Methods Eng"},{"key":"B2","doi-asserted-by":"publisher","first-page":"2078","DOI":"10.1038\/s41467-021-22265-2","article-title":"Identifying multiple sclerosis subtypes using unsupervised machine learning and MRI data","volume":"12","author":"Eshaghi","year":"2021","journal-title":"Nat Commun"},{"key":"B3","doi-asserted-by":"publisher","first-page":"765","DOI":"10.1001\/jama.2020.26858","article-title":"Diagnosis and treatment of multiple sclerosis: a review","volume":"325","author":"McGinley","year":"2021","journal-title":"JAMA"},{"key":"B4","doi-asserted-by":"publisher","first-page":"1082","DOI":"10.1136\/jnnp-2012-304094","article-title":"Brain atrophy and lesion load predict long term disability in multiple sclerosis","volume":"84","author":"Popescu","year":"2013","journal-title":"J Neurol Neurosurg Psychiatry"},{"key":"B5","doi-asserted-by":"publisher","first-page":"104187","DOI":"10.1016\/j.msard.2022.104187","article-title":"Assessment of T2 lesion-based disease activity volume outcomes in predicting disease progression in multiple sclerosis over 10 years","volume":"67","author":"Oship","year":"2022","journal-title":"Mult Scler Relat Disord"},{"key":"B6","doi-asserted-by":"publisher","first-page":"171","DOI":"10.1038\/s41582-020-0314-x","article-title":"MAGNIMS consensus recommendations on the use of brain and spinal cord atrophy measures in clinical practice","volume":"16","author":"Sastre-Garriga","year":"2020","journal-title":"Nat Rev Neurol"},{"key":"B7","doi-asserted-by":"publisher","first-page":"53","DOI":"10.1186\/s40537-021-00444-8","article-title":"Review of deep learning: concepts, CNN architectures, challenges, applications, future directions","volume":"8","author":"Alzubaidi","year":"2021","journal-title":"J Big Data"},{"key":"B8","doi-asserted-by":"publisher","first-page":"1","DOI":"10.5555\/3586589.3586815","article-title":"Underspecification presents challenges for credibility in modern machine learning","volume":"23","author":"D'Amour","year":"2022","journal-title":"J Mach Learn Res"},{"key":"B9","doi-asserted-by":"publisher","first-page":"13650","DOI":"10.1038\/s41598-018-31911-7","article-title":"Objective evaluation of multiple sclerosis lesion segmentation using a data management and processing infrastructure","volume":"8","author":"Commowick","year":"2018","journal-title":"Sci Rep"},{"key":"B10","doi-asserted-by":"publisher","first-page":"9233","DOI":"10.1038\/s41598-026-52150-1","article-title":"Performances of experts and automated methods on new multiple sclerosis lesions detection: insights from the MSSeg2 challenge","volume":"16","author":"Masson","year":"2026","journal-title":"Sci Rep."},{"key":"B11","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":"B12","doi-asserted-by":"publisher","first-page":"1173","DOI":"10.1109\/TBME.2021.3117407","article-title":"Domain adaptation for medical image analysis: a survey","volume":"69","author":"Guan","year":"2022","journal-title":"IEEE Trans Biomed Eng"},{"key":"B13","doi-asserted-by":"publisher","first-page":"2531","DOI":"10.1109\/TMI.2020.2973595","article-title":"Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation","volume":"39","author":"Zhang","year":"2020","journal-title":"IEEE Trans Med Imaging"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fnins.2022.1007453","article-title":"New lesion segmentation for multiple sclerosis brain images with imaging and lesion-aware augmentation","volume":"16","author":"Basaran","year":"2022","journal-title":"Front Neurosci"},{"key":"B15","doi-asserted-by":"crossref","unstructured":"Basaran\n              BD\n            \n            \n              Zhang\n              X\n            \n            \n              Matthews\n              PM\n            \n            \n              Bai\n              W\n            \n          \n          SegHeD: Segmentation of Heterogeneous Data for Multiple Sclerosis Lesions with Anatomical Constraints\n          \n          2024","DOI":"10.1007\/978-3-031-84525-3_5"},{"key":"B16","doi-asserted-by":"publisher","first-page":"102799","DOI":"10.1016\/j.media.2023.102799","article-title":"ImUnity: a generalizable VAE-GAN solution for multicenter MR image harmonization","volume":"88","author":"Cackowski","year":"2023","journal-title":"Med Image Anal"},{"key":"B17","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fnins.2022.964250","article-title":"Triplanar U-Net with lesion-wise voting for the segmentation of new lesions on longitudinal MRI studies","volume":"16","author":"Hitziger","year":"2022","journal-title":"Front Neurosci"},{"key":"B18","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fneur.2026.1704317","article-title":"DenseLes: slice-wise dense network for multiple sclerosis lesion segmentation and classification","volume":"17","author":"Katona","year":"2026","journal-title":"Front Neurol"},{"key":"B19","doi-asserted-by":"publisher","first-page":"103611","DOI":"10.1016\/j.nicl.2024.103611","article-title":"LST-AI: a deep learning ensemble for accurate MS lesion segmentation","volume":"42","author":"Wiltgen","year":"2024","journal-title":"NeuroImage Clin"},{"key":"B20","first-page":"234","article-title":"U-Net: convolutional networks for biomedical image segmentation","volume-title":"Medical Image Computing and Computer-Assisted Intervention \u2013 MICCAI 2015","author":"Ronneberger","year":"2015"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3446618","article-title":"A densely connected network based on U-Net for medical image segmentation","volume":"17","author":"Yang","year":"2021","journal-title":"ACM Trans Multimedia Comput Commun Appl"},{"key":"B22","doi-asserted-by":"publisher","first-page":"102312","DOI":"10.1016\/j.media.2021.102312","article-title":"DeepLesionBrain: towards a broader deep-learning generalization for multiple sclerosis lesion segmentation","volume":"76","author":"Kamraoui","year":"2022","journal-title":"Med Image Anal"},{"key":"B23","doi-asserted-by":"publisher","first-page":"449","DOI":"10.1007\/s10278-017-9983-4","article-title":"Deep learning for brain MRI segmentation: state of the art and future directions","volume":"30","author":"Akkus","year":"2017","journal-title":"J Digit Imaging"},{"key":"B24","unstructured":"Data of theWhiteMatter Hyperintensity (WMH) Segmentation Challenge\n          \n          \n            \n              Kuijf\n              H\n            \n            \n              Biesbroek\n              M\n            \n            \n              De Bresser\n              J\n            \n            \n              Heinen\n              R\n            \n            \n              Chen\n              C\n            \n            \n              Van Der Flier\n              W\n            \n          \n          \n          2022"},{"key":"B25","doi-asserted-by":"publisher","first-page":"118589","DOI":"10.1016\/j.neuroimage.2021.118589","article-title":"Multiple sclerosis lesions segmentation from multiple experts: the MICCAI 2016 challenge dataset","volume":"244","author":"Commowick","year":"2021","journal-title":"Neuroimage"},{"key":"B26","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fneur.2025.1620469","article-title":"Federated learning for lesion segmentation in multiple sclerosis: a real-world multi-center feasibility study","volume":"16","author":"Hindawi","year":"2025","journal-title":"Front Neurol"},{"key":"B27","doi-asserted-by":"publisher","first-page":"198","DOI":"10.1016\/j.neuroimage.2016.02.036","article-title":"Removing inter-subject technical variability in magnetic resonance imaging studies","volume":"132","author":"Fortin","year":"2016","journal-title":"NeuroImage"},{"key":"B28","doi-asserted-by":"publisher","first-page":"149","DOI":"10.1016\/j.neuroimage.2017.08.047","article-title":"Harmonization of multi-site diffusion tensor imaging data","volume":"161","author":"Fortin","year":"2017","journal-title":"Neuroimage"},{"key":"B29","doi-asserted-by":"publisher","first-page":"116450","DOI":"10.1016\/j.neuroimage.2019.116450","article-title":"Harmonization of large MRI datasets for the analysis of brain imaging patterns throughout the lifespan","volume":"208","author":"Pomponio","year":"2020","journal-title":"Neuroimage"},{"key":"B30","doi-asserted-by":"publisher","first-page":"117129","DOI":"10.1016\/j.neuroimage.2020.117129","article-title":"Longitudinal ComBat: a method for harmonizing longitudinal multi-scanner imaging data","volume":"220","author":"Beer","year":"2020","journal-title":"Neuroimage"},{"key":"B31","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s44172-023-00140-w","article-title":"Learning multi-site harmonization of magnetic resonance images without traveling human phantoms","volume":"3","author":"Liu","year":"2024","journal-title":"Commun Eng"},{"key":"B32","doi-asserted-by":"publisher","first-page":"119","DOI":"10.1016\/j.neunet.2023.05.014","article-title":"MSCDA: multi-level semantic-guided contrast improves unsupervised domain adaptation for breast MRI segmentation in small datasets","volume":"165","author":"Kuang","year":"2023","journal-title":"Neural Netw."},{"key":"B33","doi-asserted-by":"publisher","first-page":"102789","DOI":"10.1016\/j.media.2023.102789","article-title":"SynthSeg: segmentation of brain MRI scans of any contrast and resolution without retraining","volume":"86","author":"Billot","year":"2023","journal-title":"Med Image Anal"},{"key":"B34","doi-asserted-by":"publisher","first-page":"117471","DOI":"10.1016\/j.neuroimage.2020.117471","article-title":"A contrast-adaptive method for simultaneous whole-brain and lesion segmentation in multiple sclerosis","volume":"225","author":"Cerri","year":"2021","journal-title":"Neuroimage"},{"key":"B35","doi-asserted-by":"publisher","first-page":"115","DOI":"10.1016\/j.patrec.2025.12.008","article-title":"Towards robust and reliable multi-modal 3D segmentation of multiple sclerosis lesions","volume":"200","author":"Coppola","year":"2026","journal-title":"Pattern Recognit Lett"},{"key":"B36","doi-asserted-by":"publisher","first-page":"117689","DOI":"10.1016\/j.neuroimage.2020.117689","article-title":"Deep learning-based unlearning of dataset bias for MRI harmonisation and confound removal","volume":"228","author":"Dinsdale","year":"2021","journal-title":"Neuroimage"},{"key":"B37","doi-asserted-by":"publisher","first-page":"102076","DOI":"10.1016\/j.media.2021.102076","article-title":"Multi-site MRI harmonization via attention-guided deep domain adaptation for brain disorder identification","volume":"71","author":"Guan","year":"2021","journal-title":"Med Image Anal"},{"key":"B38","first-page":"725","article-title":"Learn to ignore: domain adaptation for multi-site MRI analysis","volume-title":"Medical Image Computing and Computer Assisted Intervention-MICCAI 2022","author":"Wolleb","year":"2022"},{"key":"B39","article-title":"Medical Image Computing and Computer Assisted Intervention - MICCAI 2023. MICCAI 2023","volume-title":"Lecture Notes in Computer Science","author":"Bi","year":"2023"},{"key":"B40","first-page":"416","article-title":"SwinUNETR-V2: stronger swin transformers with stagewise convolutions for 3D medical image segmentation","volume-title":"Medical Image Computing and Computer Assisted Intervention-MICCAI 2023","author":"He","year":"2023"},{"key":"B41","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fncom.2025.1713766","article-title":"MS-DASPNet: multiple sclerosis lesion segmentation from brain MRI using dual attention and spatial pyramid pooling with transfer learning","volume":"19","author":"Jain","year":"2026","journal-title":"Front Comput Neurosci"},{"key":"B42","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3699513","article-title":"Ischemic stroke segmentation by transformer and convolutional neural network using few-shot learning","volume":"20","author":"Alshehri","year":"2024","journal-title":"ACM Trans Multimedia Comput Commun Appl"},{"key":"B43","first-page":"488","article-title":"nnU-Net revisited: a call for rigorous validation in 3D medical image segmentation","volume-title":"Medical Image Computing and Computer Assisted Intervention-MICCAI 2024","author":"Isensee","year":"2024"},{"key":"B44","first-page":"124","article-title":"Confidence intervals uncovered: are we ready for real-world medical imaging AI?","volume-title":"Medical Image Computing and Computer Assisted Intervention-MICCAI 2024","author":"Christodoulou","year":"2024"},{"key":"B45","doi-asserted-by":"publisher","first-page":"565","DOI":"10.1109\/3DV.2016.79","article-title":"V-Net: fully convolutional neural networks for volumetricmedical image segmentation","author":"Milletari","year":"2016"},{"key":"B46","doi-asserted-by":"publisher","first-page":"310","DOI":"10.1007\/s00521-005-0467-y","article-title":"Revisiting squared-error and cross-entropy functions for training neural network classifiers","volume":"14","author":"Kline","year":"2005","journal-title":"Neural Comput Appl"},{"key":"B47","doi-asserted-by":"publisher","first-page":"2096","DOI":"10.5555\/2946645.2946704","article-title":"Domain-adversarial training of neural networks","volume":"17","author":"Ganin","year":"2016","journal-title":"J Mach Learn Res"},{"key":"B48","article-title":"Deep domain confusion: maximizing for domain invariance","author":"Tzeng","year":"2014","journal-title":"arXiv"},{"key":"B49","doi-asserted-by":"crossref","unstructured":"Deep learning and the information bottleneck principle\n          \n          1\n          5\n          \n            \n              Tishby\n              N\n            \n            \n              Zaslavsky\n              N\n            \n          \n          2015 IEEE information theory workshop (ITW)\n          \n          2015","DOI":"10.1109\/ITW.2015.7133169"},{"key":"B50","doi-asserted-by":"publisher","first-page":"1526","DOI":"10.1007\/s11263-022-01611-x","article-title":"Curriculum learning: a survey","volume":"130","author":"Soviany","year":"2022","journal-title":"Int J Comput Vis"},{"key":"B51","doi-asserted-by":"publisher","first-page":"51","DOI":"10.1007\/s12021-017-9348-7","article-title":"A novel public MR image dataset of multiple sclerosis patients with lesion segmentations based on multi-rater consensus","volume":"16","author":"Lesjak","year":"2018","journal-title":"Neuroinformatics"},{"key":"B52","doi-asserted-by":"publisher","first-page":"77","DOI":"10.1016\/j.neuroimage.2016.12.064","article-title":"Longitudinal multiple sclerosis lesion segmentation: resource and challenge","volume":"148","author":"Carass","year":"2017","journal-title":"Neuroimage"},{"key":"B53","doi-asserted-by":"publisher","first-page":"S102","DOI":"10.1016\/S1053-8119(09)70884-5","article-title":"Unbiased nonlinear average age-appropriate brain templates from birth to adulthood","volume":"47","author":"Fonov","year":"2009","journal-title":"Neuroimage"},{"key":"B54","doi-asserted-by":"publisher","first-page":"1310","DOI":"10.1109\/TMI.2010.2046908","article-title":"N4ITK: improved N3 bias correction","volume":"29","author":"Tustison","year":"2010","journal-title":"IEEE Trans Med Imaging"},{"key":"B55","doi-asserted-by":"publisher","first-page":"192","DOI":"10.1002\/jmri.22003","article-title":"Adaptive non-local means denoising of MR images with spatially varying noise levels","volume":"31","author":"Manj\u00f3n","year":"2010","journal-title":"J Magn Reson Imaging"},{"key":"B56","doi-asserted-by":"publisher","first-page":"9068","DOI":"10.1038\/s41598-021-87564-6","article-title":"The ANTsX ecosystem for quantitative biological and medical imaging","volume":"11","author":"Tustison","year":"2021","journal-title":"Sci Rep"},{"key":"B57","doi-asserted-by":"publisher","first-page":"101638","DOI":"10.1016\/j.nicl.2018.101638","article-title":"One-shot domain adaptation in multiple sclerosis lesion segmentation using convolutional neural networks","volume":"21","author":"Valverde","year":"2019","journal-title":"NeuroImage Clin"},{"key":"B58","doi-asserted-by":"publisher","first-page":"80","DOI":"10.2307\/3001968","article-title":"Individual comparisons by ranking methods","volume":"1","author":"Wilcoxon","year":"1945","journal-title":"Biometrics Bulletin"},{"key":"B59","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1007\/978-3-031-80139-6_1","article-title":"ICPR 2024 competition on multiple sclerosis lesion segmentation\u2013methods and results","volume-title":"Pattern Recognition. Competitions","author":"Rondinella","year":"2025"},{"key":"B60","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fnagi.2023.1303036","article-title":"A deep learning model for brain age prediction using minimally preprocessed T1w images as input","volume":"15","author":"Dartora","year":"2024","journal-title":"Front Aging Neurosci"},{"key":"B61","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fneur.2025.1640514","article-title":"Unsupervised semi-automated MRI segmentation detects cortical lesion expansion in chronic traumatic brain injury","volume":"16","author":"Freeman","year":"2025","journal-title":"Front Neurol"}],"container-title":["Frontiers in Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1875760\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,31]],"date-time":"2026-07-31T11:10:30Z","timestamp":1785496230000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1875760\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,31]]},"references-count":61,"alternative-id":["10.3389\/fmed.2026.1875760"],"URL":"https:\/\/doi.org\/10.3389\/fmed.2026.1875760","relation":{},"ISSN":["2296-858X"],"issn-type":[{"value":"2296-858X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,31]]},"article-number":"1875760"}}