{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,27]],"date-time":"2026-10-27T21:40:42Z","timestamp":1793137242945,"version":"4.5.0"},"reference-count":50,"publisher":"Wiley","issue":"6","license":[{"start":{"date-parts":[[2025,3,29]],"date-time":"2025-03-29T00:00:00Z","timestamp":1743206400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["aapm.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2025,6]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Semi\u2010supervised segmentation leverages sparse annotation information to learn rich representations from combined labeled and label\u2010less data for segmentation tasks. The Match\u2010based framework, by using the consistency constraint of segmentation results from different models\/augmented label\u2010less inputs, is found effective in semi\u2010supervised learning. This approach, however, is challenged by the low quality of pseudo\u2010labels generated as intermediate products for training the network, due to the lack of the \u2018\u2018ground\u2010truth\u2019\u2019 reference.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>This study aims to leverage the foundation model, segment anything model (SAM), to assist unsupervised learning of Match\u2010based frameworks. Trained with an extremely large dataset, SAM\u2010based methods generalize better than traditional models to various imaging domains, allow it to serve as an assistant to Match\u2010based frameworks to improve the quality of intermediate pseudo\u2010labels for semi\u2010supervised learning.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We propose SAM\u2010Match, a SAM\u2010guided and Match\u2010based framework for semi\u2010supervised medical image segmentation. Our approach involves two main steps: First, we use pretrained Match\u2010based models to extract high\u2010confidence predictions for prompt generation. Second, these prompts and unlabeled images are input into a fine\u2010tuned SAM\u2010based method to produce high\u2010quality masks as pseudo\u2010labels. And the refined pseudo\u2010labels are further fed back to train the Match\u2010based framework. SAM\u2010Match can be trained in an end\u2010to\u2010end manner, facilitating interactions between the SAM\u2010 and Match\u2010based models.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>SAM\u2010Match demonstrates robust performance across multiple medical imaging datasets, including the ACDC cardiac MRI dataset, the BUSI breast ultrasound dataset, and an in\u2010house liver MRI dataset (MRLiver). We partitioned the datasets into training, validation, and test sets (70%, 10%, and 20% for ACDC; 60%, 9%, and 31% for BUSI; and 62%, 12%, and 25% for MRLiver). On ACDC, with only 3 labeled cases, we achieved a Dice score of 89.36%\u00a0\u00b1\u00a00.06% on 20 test cases. For BUSI, using just 30 labeled samples for training, we attained a Dice score of 59.35%\u00a0\u00b1\u00a00.12% on 170 test samples. On MRLiver, training with only 3 labeled cases resulted in a Dice score of 80.04%\u00a0\u00b1\u00a00.11% on 12 test scans. Wilcoxon signed\u2010rank tests with Bonferroni corrections between the SAM\u2010Match framework and the other comparison methods further demonstrated the statistical significance of SAM\u2010Match's improvement in segmentation accuracy.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>\n                      Our SAM\u2010Match framework shows promising results in semi\u2010supervised semantic segmentation, effectively tackling the challenges of automatic prompt generation for SAM and high\u2010quality pseudo\u2010label generation for Match\u2010based models. It can help accelerate the adoption of semi\u2010supervised learning in segmentation tasks, particularly in data\u2010scarce scenarios. Our data and code will be made available at\n                      <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"https:\/\/github.com\/apple1986\/SAMatch\">https:\/\/github.com\/apple1986\/SAMatch<\/jats:ext-link>\n                      .\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.1002\/mp.17785","type":"journal-article","created":{"date-parts":[[2025,4,15]],"date-time":"2025-04-15T04:55:56Z","timestamp":1744692956000},"page":"4513-4527","update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":16,"title":["A segment anything model\u2010guided and match\u2010based semi\u2010supervised segmentation framework for medical imaging"],"prefix":"10.1002","volume":"52","author":[{"given":"Guoping","family":"Xu","sequence":"first","affiliation":[{"name":"The Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology University of Texas Southwestern Medical Center  Dallas Texas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaoxue","family":"Qian","sequence":"additional","affiliation":[{"name":"The Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology University of Texas Southwestern Medical Center  Dallas Texas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hua\u2010Chieh","family":"Shao","sequence":"additional","affiliation":[{"name":"The Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology University of Texas Southwestern Medical Center  Dallas Texas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jax","family":"Luo","sequence":"additional","affiliation":[{"name":"Massachusetts General Hospital Harvard Medical School  Boston Massachusetts USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Weiguo","family":"Lu","sequence":"additional","affiliation":[{"name":"The Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology University of Texas Southwestern Medical Center  Dallas Texas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"You","family":"Zhang","sequence":"additional","affiliation":[{"name":"The Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology University of Texas Southwestern Medical Center  Dallas Texas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2025,3,29]]},"reference":[{"key":"e_1_2_9_2_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.compmedimag.2020.101851"},{"key":"e_1_2_9_3_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.14240"},{"key":"e_1_2_9_4_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298965"},{"key":"e_1_2_9_5_1","first-page":"234","volume-title":"U\u2010Net: Convolutional Networks for Biomedical Image Segmentation","author":"Ronneberger O","year":"2015"},{"key":"e_1_2_9_6_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2017.2699184"},{"key":"e_1_2_9_7_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2023.107840"},{"key":"e_1_2_9_8_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2024.108238"},{"key":"e_1_2_9_9_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.eswa.2023.123052"},{"key":"e_1_2_9_10_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00699"},{"key":"e_1_2_9_11_1","first-page":"31","article-title":"Realistic evaluation of deep semi\u2010supervised learning algorithms","author":"Oliver A","year":"2018","journal-title":"Adv Neural Inf Process Syst"},{"key":"e_1_2_9_12_1","first-page":"596","article-title":"Fixmatch: simplifying semi\u2010supervised learning with consistency and confidence","volume":"33","author":"Sohn K","year":"2020","journal-title":"Adv Neural Inf Process Syst"},{"key":"e_1_2_9_13_1","first-page":"1","volume-title":"2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI)","author":"Liu M","year":"2022"},{"key":"e_1_2_9_14_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01108"},{"key":"e_1_2_9_15_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-32226-7_90"},{"key":"e_1_2_9_16_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.17399"},{"key":"e_1_2_9_17_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00371"},{"key":"e_1_2_9_18_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2023.102918"},{"key":"e_1_2_9_19_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2023.103061"},{"key":"e_1_2_9_20_1","doi-asserted-by":"publisher","DOI":"10.1109\/JBHI.2024.3406871"},{"key":"e_1_2_9_21_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-44824-z"},{"key":"e_1_2_9_22_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2025.103547"},{"key":"e_1_2_9_23_1","first-page":"552","volume-title":"SAM\u2010Med3D\u2010MOE: Towards a Non\u2010Forgetting Segment Anything Model Via Mixture of Experts for 3D Medical Image Segmentation","author":"Wang G","year":"2024"},{"key":"e_1_2_9_24_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW60793.2023.00273"},{"key":"e_1_2_9_25_1","unstructured":"ShaharabanyT DahanA GiryesR WolfL.AutoSAM: adapting SAM to medical images by overloading the prompt encoder.ArXiv Preprint arXiv:2306.06370.2023."},{"key":"e_1_2_9_26_1","doi-asserted-by":"crossref","unstructured":"LinX XiangY ZhangL YangX YanZ YuL.Beyond adapting SAM: Towards end\u2010to\u2010end ultrasound image segmentation via auto prompting.International Conference on Medical Image Computing and Computer\u2010Assisted Intervention (MICCAI).2024:24\u201034.","DOI":"10.1007\/978-3-031-72111-3_3"},{"key":"e_1_2_9_27_1","doi-asserted-by":"crossref","unstructured":"ZhangY ChengY QiY.SemiSAM: exploring SAM for enhancing semi\u2010supervised medical image segmentation with extremely limited annotations.2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM).IEEE;2024:3982\u20103986.","DOI":"10.1109\/BIBM62325.2024.10821951"},{"key":"e_1_2_9_28_1","doi-asserted-by":"crossref","unstructured":"ZhangY ZhouT WangS WuY GuP ChenDZ.SamDSK: combining segment anything model with domain\u2010specific knowledge for semi\u2010supervised learning in medical image segmentation.Pattern Recognition and Computer Vision (PRCV).2024:343\u2010357.","DOI":"10.1007\/978-981-97-8496-7_24"},{"key":"e_1_2_9_29_1","doi-asserted-by":"crossref","unstructured":"LiX DengR TangY BaoS YangH HuoY.Leverage weekly annotation to pixel\u2010wise annotation via zero\u2010shot segment anything model for molecular\u2010empowered learning. In:Proceedings Medical Imaging 2024: Digital and Computational Pathology.SPIE;2024;12933:133\u2010139.","DOI":"10.1117\/12.3006577"},{"key":"e_1_2_9_30_1","unstructured":"LaineS AilaT.Temporal ensembling for semi\u2010supervised learning.ArXiv Preprint arXiv:1610.02242.2016."},{"key":"e_1_2_9_31_1","first-page":"30","article-title":"Mean teachers are better role models: weight\u2010averaged consistency targets improve semi\u2010supervised deep learning results","author":"Tarvainen A","year":"2017","journal-title":"Adv Neural Inf Process Syst"},{"key":"e_1_2_9_32_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-32245-8_67"},{"key":"e_1_2_9_33_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2021.10.008"},{"key":"e_1_2_9_34_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2022.102517"},{"key":"e_1_2_9_35_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00421"},{"key":"e_1_2_9_36_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00683"},{"key":"e_1_2_9_37_1","first-page":"18408","article-title":"Flexmatch: boosting semi\u2010supervised learning with curriculum pseudo labeling","volume":"34","author":"Zhang B","year":"2021","journal-title":"Adv Neural Inf Process Syst"},{"key":"e_1_2_9_38_1","unstructured":"WangY ChenH HengQ et\u00a0al.Freematch: self\u2010adaptive thresholding for semi\u2010supervised learning.arXiv preprint arXiv:2205.07246.2022."},{"key":"e_1_2_9_39_1","unstructured":"DeVriesT.Improved regularization of convolutional neural networks with cutout.arXiv preprint arXiv:1708.04552.2017."},{"key":"e_1_2_9_40_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00612"},{"key":"e_1_2_9_41_1","first-page":"22106","article-title":"Semi\u2010supervised semantic segmentation via adaptive equalization learning","volume":"34","author":"Hu H","year":"2021","journal-title":"Adv Neural Inf Process Syst"},{"key":"e_1_2_9_42_1","first-page":"42","volume-title":"LeViT\u2010UNet: Make Faster Encoders With Transformer for Medical Image Segmentation","author":"Xu G","year":"2023"},{"key":"e_1_2_9_43_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.dib.2019.104863"},{"key":"e_1_2_9_44_1","first-page":"3071","volume-title":"2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)","author":"Liao W","year":"2020"},{"key":"e_1_2_9_45_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-66179-7_47"},{"key":"e_1_2_9_46_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00262"},{"key":"e_1_2_9_47_1","unstructured":"LiH KadavA DurdanovicI SametH GrafHP.Pruning filters for efficient convnets.arXiv preprint arXiv:1608.08710.2016."},{"key":"e_1_2_9_48_1","doi-asserted-by":"publisher","DOI":"10.1145\/3486618"},{"key":"e_1_2_9_49_1","unstructured":"RaviN GabeurV HuY\u2010T et\u00a0al.SAM 2: segment anything in images and videos.arXiv preprint arXiv:2408.00714.2024."},{"key":"e_1_2_9_50_1","unstructured":"ZhuJ QiY WuJ.Medical SAM 2: segment medical images as video via segment anything model 2.arXiv preprint arXiv:2408.00874.2024."},{"key":"e_1_2_9_51_1","first-page":"542","volume-title":"FastSAM3D: An Efficient Segment Anything Model for 3D Volumetric Medical Images","author":"Shen Y","year":"2024"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.17785","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,12]],"date-time":"2025-11-12T08:03:27Z","timestamp":1762934607000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.17785"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,3,29]]},"references-count":50,"journal-issue":{"issue":"6","published-print":{"date-parts":[[2025,6]]}},"alternative-id":["10.1002\/mp.17785"],"URL":"https:\/\/doi.org\/10.1002\/mp.17785","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,3,29]]},"assertion":[{"value":"2024-11-25","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-03-11","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-03-29","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}