{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/semi-supervised-brain-lesion-segmentation","title":"Semi-Supervised Brain Lesion Segmentation with an Adapted Mean Teacher Model","arxiv_id":"1903.01248","date":"2019-03-04","proceeding":null,"authors":["Wenhui Cui","Yanlin Liu","Yuxing Li","Menghao Guo","Yiming Li","Xiuli Li","Tianle Wang","Xiangzhu Zeng","Chuyang Ye"],"abstract":"Automated brain lesion segmentation provides valuable information for the\nanalysis and intervention of patients. In particular, methods based on\nconvolutional neural networks (CNNs) have achieved state-of-the-art\nsegmentation performance. However, CNNs usually require a decent amount of\nannotated data, which may be costly and time-consuming to obtain. Since\nunannotated data is generally abundant, it is desirable to use unannotated data\nto improve the segmentation performance for CNNs when limited annotated data is\navailable. In this work, we propose a semi-supervised learning (SSL) approach\nto brain lesion segmentation, where unannotated data is incorporated into the\ntraining of CNNs. We adapt the mean teacher model, which is originally\ndeveloped for SSL-based image classification, for brain lesion segmentation.\nAssuming that the network should produce consistent outputs for similar inputs,\na loss of segmentation consistency is designed and integrated into a\nself-ensembling framework. Specifically, we build a student model and a teacher\nmodel, which share the same CNN architecture for segmentation. The student and\nteacher models are updated alternately. At each step, the student model learns\nfrom the teacher model by minimizing the weighted sum of the segmentation loss\ncomputed from annotated data and the segmentation consistency loss between the\nteacher and student models computed from unannotated data. Then, the teacher\nmodel is updated by combining the updated student model with the historical\ninformation of teacher models using an exponential moving average strategy. For\ndemonstration, the proposed approach was evaluated on ischemic stroke lesion\nsegmentation, where it improves stroke lesion segmentation with the\nincorporation of unannotated data.","url_abs":"http://arxiv.org/abs/1903.01248v1","url_pdf":"http://arxiv.org/pdf/1903.01248v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"semi-supervised-brain-lesion-segmentation","repo_url":"https://github.com/wenhui0206/MeanTeacher-DeepMedic","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"ischemic-stroke-lesion-segmentation","task_name":"Ischemic Stroke Lesion Segmentation"},{"task_slug":"lesion-segmentation","task_name":"Lesion Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1903.01248","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}