Introduces the first framework-agnostic method to leverage a learned quality prediction model to guide semi-supervised learning (SSL). Two complementary mechanisms for integrating as drop-in enhancements to existing SSL frameworks: differentiable quality regularization and pseudo-label re-weighting. Extensive evaluation across multiple medical imaging datasets, modalities, SSL paradigms, and model architectures.</p>\n","updatedAt":"2026-06-05T04:09:42.649Z","author":{"_id":"63855ef0ac61472e5e96d77e","avatarUrl":"/avatars/283c639b1f50760fb9d73c96c48d896e.svg","fullname":"Kumar Abhishek","name":"kabhishe","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8735871315002441},"editors":["kabhishe"],"editorAvatarUrls":["/avatars/283c639b1f50760fb9d73c96c48d896e.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.01753","authors":[{"_id":"6a224ad73490a593e87b157d","name":"Kumar Abhishek","hidden":false},{"_id":"6a224ad73490a593e87b157e","name":"Ghassan Hamarneh","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/63855ef0ac61472e5e96d77e/RXrCb4mV8PzlVXPG8-hvM.png"],"publishedAt":"2026-06-01T00:00:00.000Z","submittedOnDailyAt":"2026-06-05T00:00:00.000Z","title":"Quality-Guided Semi-Supervised Learning for Medical Image Segmentation","submittedOnDailyBy":{"_id":"63855ef0ac61472e5e96d77e","avatarUrl":"/avatars/283c639b1f50760fb9d73c96c48d896e.svg","isPro":false,"fullname":"Kumar Abhishek","user":"kabhishe","type":"user","name":"kabhishe"},"summary":"Training accurate medical image segmentation models requires large amounts of densely annotated data, which is costly and time-consuming to obtain. Semi-supervised learning (SSL) alleviates this by learning from both abundant unlabeled data and limited labeled data. However, most modern SSL methods rely on pseudolabels for unlabeled data, and typically assess their reliability through model confidence or uncertainty, measures that are self-referential and lack explicit grounding in segmentation quality. Instead, we propose a quality-guided SSL framework that trains a dedicated network to estimate segmentation quality from image-mask pairs. The predictor is trained on variable-quality masks generated through synthetic corruptions augmented with imperfect outputs from partially trained segmentation models, capturing realistic error patterns encountered during training. We integrate the quality predictor into SSL through two complementary mechanisms: a quality-aware regularization loss and a quality-based pseudolabel sample reweighting scheme. We show that our method serves as a drop-in enhancement to existing SSL frameworks. Extensive experiments across five datasets and multiple architectures demonstrate consistent improvements over competing SSL methods, advancing the state-of-the-art in semi-supervised medical image segmentation.","upvotes":0,"discussionId":"6a224ad73490a593e87b157f","githubRepo":"https://github.com/sfu-mial/QG-SSL","githubRepoAddedBy":"user","ai_summary":"A quality-guided semi-supervised learning framework for medical image segmentation that uses a dedicated quality predictor to improve pseudolabel reliability and segmentation performance.","ai_keywords":["semi-supervised learning","pseudolabels","segmentation quality","quality predictor","regularization loss","sample reweighting","synthetic corruptions","partially trained models"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":0,"organization":{"_id":"6385602cb2906edaf839745b","name":"sfu-mial","fullname":"Medical Image Analysis Lab, SFU","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1669685287218-63855ef0ac61472e5e96d77e.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[],"acceptLanguages":["en"],"organization":{"_id":"6385602cb2906edaf839745b","name":"sfu-mial","fullname":"Medical Image Analysis Lab, SFU","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1669685287218-63855ef0ac61472e5e96d77e.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.01753.md"}">
Quality-Guided Semi-Supervised Learning for Medical Image Segmentation
Abstract
A quality-guided semi-supervised learning framework for medical image segmentation that uses a dedicated quality predictor to improve pseudolabel reliability and segmentation performance.
Training accurate medical image segmentation models requires large amounts of densely annotated data, which is costly and time-consuming to obtain. Semi-supervised learning (SSL) alleviates this by learning from both abundant unlabeled data and limited labeled data. However, most modern SSL methods rely on pseudolabels for unlabeled data, and typically assess their reliability through model confidence or uncertainty, measures that are self-referential and lack explicit grounding in segmentation quality. Instead, we propose a quality-guided SSL framework that trains a dedicated network to estimate segmentation quality from image-mask pairs. The predictor is trained on variable-quality masks generated through synthetic corruptions augmented with imperfect outputs from partially trained segmentation models, capturing realistic error patterns encountered during training. We integrate the quality predictor into SSL through two complementary mechanisms: a quality-aware regularization loss and a quality-based pseudolabel sample reweighting scheme. We show that our method serves as a drop-in enhancement to existing SSL frameworks. Extensive experiments across five datasets and multiple architectures demonstrate consistent improvements over competing SSL methods, advancing the state-of-the-art in semi-supervised medical image segmentation.
Community
Introduces the first framework-agnostic method to leverage a learned quality prediction model to guide semi-supervised learning (SSL). Two complementary mechanisms for integrating as drop-in enhancements to existing SSL frameworks: differentiable quality regularization and pseudo-label re-weighting. Extensive evaluation across multiple medical imaging datasets, modalities, SSL paradigms, and model architectures.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.01753 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.01753 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.01753 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.