{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T10:46:24Z","timestamp":1792752384688,"version":"4.5.0"},"reference-count":45,"publisher":"IOP Publishing","issue":"20","license":[{"start":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T00:00:00Z","timestamp":1792713600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T00:00:00Z","timestamp":1792713600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"},{"start":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T00:00:00Z","timestamp":1792713600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"},{"start":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T00:00:00Z","timestamp":1792713600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/publishingsupport.iopscience.iop.org\/questions\/alternative-author-rights-policies\/"}],"funder":[{"name":"National Institute of Biomedical Imaging and Bioengineering","award":["R01EB032680"],"award-info":[{"award-number":["R01EB032680"]}]},{"name":"National Institute of Dental and Craniofacial Research","award":["R01DE033512"],"award-info":[{"award-number":["R01DE033512"]}]},{"name":"National Heart, Lung, and Blood Institute","award":["R01HL163570"],"award-info":[{"award-number":["R01HL163570"]}]},{"name":"National Cancer Institute","award":["R01CA272991"],"award-info":[{"award-number":["R01CA272991"]}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Phys. Med. Biol."],"published-print":{"date-parts":[[2026,10,28]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    <jats:italic>Objective.<\/jats:italic>\n                    This study investigates how domain-aligned self-supervised pretraining influences downstream PET\/CT tumor segmentation, with a focus on disentangling the roles of domain proximity and pretraining dataset scale.\n                    <jats:italic>Approach.<\/jats:italic>\n                    We compared nnU-Net segmentation models initialized from random weights or from DINOv3 encoders pretrained on natural images, x-ray mammography, CT, or PET, using datasets of different sizes. To disentangle the effects of pretraining domain and dataset size, CT-DINOv3 encoders were pretrained using different numbers of 2D CT slices, including subsets matched in size to the pretraining datasets from other domains. Models were evaluated on two PET\/CT benchmarks, autoPET (whole-body, FDG\/prostate-specific membrane antigen (PSMA)) and HECKTOR (head-and-neck primary tumors and metastatic lymph nodes), using five-fold cross-validation with frozen encoders.\n                    <jats:italic>Main results.<\/jats:italic>\n                    CT-pretrained models achieved the highest Dice scores for lymph nodes, FDG lesions, and PSMA lesions. For primary tumors, PET pretraining produced the highest observed Dice score, followed by CT pretraining, suggesting the potential benefit of in-domain representations when PET data are sufficient. Notably, PET pretraining exceeded size-matched CT pretraining for PSMA but remained below models pretrained on larger CT datasets for both PSMA and FDG targets. In contrast, natural-image pretraining did not outperform from-scratch training despite its much larger pretraining scale, indicating that scale alone is insufficient when domain proximity is low. Across comparisons, strong transfer reflected the combination of domain proximity and sufficient pretraining scale, especially for CT-based models with broad data availability.\n                    <jats:italic>Significance.<\/jats:italic>\n                    These results demonstrate that both domain proximity and dataset scale jointly determine the effectiveness of self-supervised pretraining. While higher domain alignment improves performance at comparable scale, sufficient data is required to realize this benefit. CT-based pretraining provides a favorable balance between domain relevance and data availability, leading to robust performance across tasks. This study provides practical guidance for selecting pretraining strategies in PET\/CT segmentation.\n                  <\/jats:p>","DOI":"10.1088\/1361-6560\/aeb29f","type":"journal-article","created":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T22:52:34Z","timestamp":1791586354000},"page":"205022","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Domain-aligned self-supervision: CT-specialized DINOv3 enhances PET\/CT tumor segmentation"],"prefix":"10.1088","volume":"71","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-4938-2205","authenticated-orcid":true,"given":"Yu-Nong","family":"Scarlett Lin","sequence":"first","affiliation":[{"name":"Georgia Institute of Technology"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4208-2035","authenticated-orcid":true,"given":"Shansong","family":"Wang","sequence":"additional","affiliation":[{"name":"The University of Chicago"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuheng","family":"Li","sequence":"additional","affiliation":[{"name":"Georgia Institute of Technology"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3295-328X","authenticated-orcid":true,"given":"Mojtaba","family":"Safari","sequence":"additional","affiliation":[{"name":"The University of Chicago"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0955-3950","authenticated-orcid":true,"given":"Qiang","family":"Li","sequence":"additional","affiliation":[{"name":"Emory University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zach","family":"Eidex","sequence":"additional","affiliation":[{"name":"Emory University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-9337-2564","authenticated-orcid":true,"given":"Yizhou","family":"Wu","sequence":"additional","affiliation":[{"name":"Georgia Institute of Technology"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ryan","family":"Sanford","sequence":"additional","affiliation":[{"name":"Emory University"},{"name":"Georgia Institute of Technology"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8556-5307","authenticated-orcid":true,"given":"Kevin T","family":"Chen","sequence":"additional","affiliation":[{"name":"National Taiwan University"},{"name":"National Taiwan University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9023-5855","authenticated-orcid":true,"given":"Xiaofeng","family":"Yang","sequence":"additional","affiliation":[{"name":"Georgia Institute of Technology"},{"name":"The University of Chicago"},{"name":"Emory University"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"266","published-online":{"date-parts":[[2026,10,23]]},"reference":[{"key":"pmbaeb29fbib1","doi-asserted-by":"publisher","DOI":"10.1109\/nssmicrtsd49126.2023.10338080","type":"conference-proceedings","article-title":"Towards an automated Segment Anything Model (SAM) for lesion segmentation in oncological PET\/CT images","author":"Ahamed","year":"2023"},{"key":"pmbaeb29fbib2","doi-asserted-by":"publisher","DOI":"10.1088\/1361-6560\/ac88b3","type":"journal-article","article-title":"Personalized mid-course FDG-PET based adaptive treatment planning for non-small cell lung cancer using machine learning and optimization","volume":"67","author":"Ajdari","year":"2022","journal-title":"Phys. Med. Biol."},{"key":"pmbaeb29fbib3","doi-asserted-by":"publisher","first-page":"pp 1","DOI":"10.1007\/978-3-031-27420-6_1","type":"conference-proceedings","article-title":"Overview of the HECKTOR challenge at MICCAI 2022: automatic head and neck tumor segmentation and outcome prediction in PET\/CT","author":"Andrearczyk","year":"2023"},{"key":"pmbaeb29fbib4","doi-asserted-by":"crossref","DOI":"10.1109\/ICCV48922.2021.00346","type":"preprint","article-title":"Big self-supervised models advance medical image classification","author":"Azizi","year":"2021"},{"key":"pmbaeb29fbib5","doi-asserted-by":"publisher","first-page":"328","DOI":"10.1007\/s00259-014-2961-x","type":"journal-article","article-title":"FDG PET\/CT: EANM procedure guidelines for tumour imaging: version 2.0","volume":"42","author":"Boellaard","year":"2014","journal-title":"Eur. J. Nucl. Med. Mol. Imaging."},{"key":"pmbaeb29fbib6","type":"preprint","article-title":"On the opportunities and risks of foundation models","volume":"1","author":"Bommasani","year":"2021"},{"key":"pmbaeb29fbib7","first-page":"pp 9650","type":"book","article-title":"Emerging Properties in Self-Supervised Vision Transformers","author":"Caron","year":"2021"},{"key":"pmbaeb29fbib8","article-title":"A simple framework for contrastive learning of visual representations","author":"Chen","year":"2020","type":"preprint"},{"key":"pmbaeb29fbib9","article-title":"DINOv2: learning robust visual features without supervision maxime oquab","author":"Darcet","year":"2024","type":"preprint"},{"key":"pmbaeb29fbib10","doi-asserted-by":"publisher","first-page":"481","DOI":"10.2967\/jnumed.125.270260","type":"journal-article","article-title":"AutoPET challenge on fully automated lesion segmentation in oncologic PET\/CT imaging, part 2: domain generalization","volume":"67","author":"Dexl","year":"2025","journal-title":"J. Nucl. Med."},{"key":"pmbaeb29fbib11","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/s40658-026-00887-z","type":"journal-article","article-title":"Multi-modal segment anything model (mmSAM) for tumor segmentation in multi-tracer oncologic PET\/CT","volume":"13","author":"Fang","year":"2026","journal-title":"EJNMMI Physics"},{"key":"pmbaeb29fbib12","doi-asserted-by":"publisher","first-page":"995","DOI":"10.2967\/jnumed.124.268071","type":"journal-article","article-title":"Treatment response evaluation in prostate cancer using PSMA PET\/CT","volume":"66","author":"Gafita","year":"2025","journal-title":"J. Nucl. Med."},{"key":"pmbaeb29fbib13","doi-asserted-by":"publisher","first-page":"10","DOI":"10.1038\/s41597-022-01718-3","type":"journal-article","article-title":"A whole-body FDG-PET\/CT dataset with manually annotated tumor lesions","volume":"9","author":"Gatidis","year":"2022","journal-title":"Sci. Data"},{"key":"pmbaeb29fbib14","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR.2016.265","type":"other","article-title":"Image style transfer using convolutional neural networks","author":"Gatys","year":"2016"},{"key":"pmbaeb29fbib15","article-title":"Bootstrap your own latent: a new approach to self-supervised learning","author":"Grill","year":"2020","type":"preprint"},{"key":"pmbaeb29fbib16","doi-asserted-by":"publisher","first-page":"759","DOI":"10.1016\/j.ijrobp.2008.04.059","type":"journal-article","article-title":"Combined 18F-FDG-PET\/CT imaging in radiotherapy target delineation for head-and-neck cancer","volume":"73","author":"Guido","year":"2009","journal-title":"Int. J. Radiat. Oncol. Biol. Phys."},{"key":"pmbaeb29fbib17","article-title":"Masked autoencoders are scalable vision learners","author":"He","year":"2021","type":"preprint"},{"key":"pmbaeb29fbib18","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1038\/s41592-020-01008-z","type":"journal-article","article-title":"nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation","volume":"18","author":"Isensee","year":"2020","journal-title":"Nat. Methods"},{"key":"pmbaeb29fbib19","doi-asserted-by":"publisher","first-page":"01","DOI":"10.1148\/ryai.220047","type":"journal-article","article-title":"The EMory BrEast imaging Dataset (EMBED): a racially diverse, granular dataset of 3.4 million screening and diagnostic mammographic images","volume":"5","author":"Jeong","year":"2023","journal-title":"Radiol. Artif. Intell."},{"key":"pmbaeb29fbib20","doi-asserted-by":"publisher","first-page":"pp 47","DOI":"10.1007\/978-3-031-27420-6_4","type":"conference-proceedings","article-title":"A General web-based platform for automatic delineation of head and neck gross tumor volumes in PET\/CT images","author":"Jiang","year":"2023"},{"key":"pmbaeb29fbib21","doi-asserted-by":"publisher","DOI":"10.1016\/j.artmed.2021.102078","type":"journal-article","article-title":"Transfer learning in medical image segmentation: new insights from analysis of the dynamics of model parameters and learned representations","volume":"116","author":"Karimi","year":"2021","journal-title":"Artif. Intell. Med."},{"key":"pmbaeb29fbib22","doi-asserted-by":"crossref","DOI":"10.1109\/ICCV51070.2023.00371","type":"preprint","article-title":"Segment anything","author":"Kirillov","year":"2023"},{"key":"pmbaeb29fbib23","doi-asserted-by":"publisher","first-page":"2500","DOI":"10.1056\/NEJMoa022136","type":"journal-article","article-title":"Staging of non\u2013small-cell lung cancer with integrated positron-emission tomography and computed tomography","volume":"348","author":"Lardinois","year":"2003","journal-title":"New Engl. J. Med."},{"key":"pmbaeb29fbib24","article-title":"MedDINOv3: how to adapt vision foundation models for medical image segmentation?","author":"Li","year":"2025","type":"preprint"},{"key":"pmbaeb29fbib25","article-title":"Does DINOv3 set a new medical vision standard?","author":"Liu","year":"2025","type":"preprint"},{"key":"pmbaeb29fbib26","doi-asserted-by":"publisher","first-page":"01","DOI":"10.1038\/s41467-024-44824-z","type":"journal-article","article-title":"Segment anything in medical images","volume":"15","author":"Ma","year":"2024","journal-title":"Nat. Commun."},{"key":"pmbaeb29fbib27","doi-asserted-by":"publisher","first-page":"321","DOI":"10.1002\/mp.17470","type":"journal-article","article-title":"Necessity and impact of specialization of large foundation model for medical segmentation tasks","volume":"52","author":"Nguyen","year":"2024","journal-title":"Med. Phys."},{"key":"pmbaeb29fbib28","doi-asserted-by":"publisher","first-page":"5825","DOI":"10.21037\/qims-2024-2826","type":"journal-article","article-title":"A narrative review of foundation models for medical image segmentation: zero-shot performance evaluation on diverse modalities","volume":"15","author":"Noh","year":"2025","journal-title":"Quant. Imaging Med. Surg."},{"key":"pmbaeb29fbib29","article-title":"Developing a PET\/CT foundation model for cross-modal anatomical and functional imaging","author":"Oh","year":"2025","type":"preprint"},{"key":"pmbaeb29fbib30","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2021.102336","type":"journal-article","article-title":"Head and neck tumor segmentation in PET\/CT: the HECKTOR challenge","volume":"77","author":"Oreiller","year":"2022","journal-title":"Med. Image Anal."},{"key":"pmbaeb29fbib31","doi-asserted-by":"publisher","first-page":"p 121","DOI":"10.1007\/978-3-031-27420-6_13","type":"conference-proceedings","article-title":"Simplicity is all you need: out-of-the-Box nnUNet followed by binary-weighted radiomic model for segmentation and outcome prediction in head and neck PET\/CT","author":"Rebaud","year":"2023"},{"key":"pmbaeb29fbib32","article-title":"Segment anything model for head and neck tumor segmentation with CT, PET and MRI multi-modality images","author":"Ren","year":"2024","type":"preprint"},{"key":"pmbaeb29fbib33","article-title":"Spreading vectors for similarity search","author":"Sablayrolles","year":"2018","type":"preprint"},{"key":"pmbaeb29fbib34","doi-asserted-by":"publisher","first-page":"pp 230","DOI":"10.1007\/978-3-031-27420-6_23","type":"conference-proceedings","article-title":"Deep learning and machine learning techniques for automated PET\/CT segmentation and survival prediction in head and neck cancer","author":"Salmanpour","year":"2023"},{"key":"pmbaeb29fbib35","article-title":"DINOv3","author":"Sim\u00e9oni","year":"2025","type":"preprint"},{"key":"pmbaeb29fbib36","doi-asserted-by":"publisher","first-page":"p 107","DOI":"10.1007\/978-3-031-27420-6_11","type":"conference-proceedings","article-title":"Multi-scale fusion methodologies for head and neck tumor segmentation","author":"Srivastava","year":"2023"},{"key":"pmbaeb29fbib37","article-title":"Measuring robustness to natural distribution shifts in image classification","author":"Taori","year":"2020","type":"preprint"},{"key":"pmbaeb29fbib38","doi-asserted-by":"publisher","DOI":"10.1016\/j.adro.2023.101212","type":"journal-article","article-title":"Positron emission tomography (PET)\/computed tomography (CT) imaging in radiation therapy treatment planning: a review of PET imaging tracers and methods to incorporate PET\/CT","volume":"8","author":"Trotter","year":"2023","journal-title":"Adv. Radiat. Oncol."},{"key":"pmbaeb29fbib39","doi-asserted-by":"publisher","first-page":"107","DOI":"10.1016\/j.radonc.2019.07.025","type":"journal-article","article-title":"18F-FDG-PET\/CT-based treatment planning for definitive (chemo)radiotherapy in patients with head and neck squamous cell carcinoma improves regional control and survival","volume":"142","author":"van den Bosch","year":"2020","journal-title":"Radiother. Oncol."},{"key":"pmbaeb29fbib40","article-title":"Primus: enforcing attention usage for 3D medical image segmentation","author":"Wald","year":"2025","type":"preprint"},{"key":"pmbaeb29fbib41","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR52688.2022.00943","type":"preprint","article-title":"SimMIM: a simple framework for masked image modeling","author":"Xie","year":"2022"},{"key":"pmbaeb29fbib42","article-title":"Benchmarking DINOv3 for multi-task stroke analysis on non-contrast CT","author":"Zhang","year":"2025a","type":"preprint"},{"key":"pmbaeb29fbib43","doi-asserted-by":"crossref","DOI":"10.1109\/ICCV51701.2025.01961","type":"preprint","article-title":"SegAnyPET: universal promptable segmentation from positron emission tomography images","author":"Zhang","year":"2025b"},{"key":"pmbaeb29fbib44","article-title":"A comprehensive survey on pretrained foundation models: a history from BERT to ChatGPT","author":"Zhou","year":"2023","type":"preprint"},{"key":"pmbaeb29fbib45","article-title":"iBOT: image BERT pre-training with online tokenizer","author":"Zhou","year":"2022","type":"preprint"}],"container-title":["Physics in Medicine &amp; Biology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T10:35:37Z","timestamp":1792751737000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/1361-6560\/aeb29f"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,23]]},"references-count":45,"journal-issue":{"issue":"20","published-online":{"date-parts":[[2026,10,23]]},"published-print":{"date-parts":[[2026,10,28]]}},"URL":"https:\/\/doi.org\/10.1088\/1361-6560\/aeb29f","relation":{},"ISSN":["0031-9155","1361-6560"],"issn-type":[{"value":"0031-9155","type":"print"},{"value":"1361-6560","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,23]]},"assertion":[{"value":"Domain-aligned self-supervision: CT-specialized DINOv3 enhances PET\/CT tumor segmentation","name":"article_title","label":"Article Title"},{"value":"Physics in Medicine & Biology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2026 The Author(s). Published on behalf of Institute of Physics and Engineering in Medicine by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2026-01-04","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-10-09","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-10-23","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}