{"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/generalised-dice-overlap-as-a-deep-learning","title":"Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentations","arxiv_id":"1707.03237","date":"2017-07-11","proceeding":null,"authors":["Carole H. Sudre","Wenqi Li","Tom Vercauteren","Sébastien Ourselin","M. Jorge Cardoso"],"abstract":"Deep-learning has proved in recent years to be a powerful tool for image\nanalysis and is now widely used to segment both 2D and 3D medical images.\nDeep-learning segmentation frameworks rely not only on the choice of network\narchitecture but also on the choice of loss function. When the segmentation\nprocess targets rare observations, a severe class imbalance is likely to occur\nbetween candidate labels, thus resulting in sub-optimal performance. In order\nto mitigate this issue, strategies such as the weighted cross-entropy function,\nthe sensitivity function or the Dice loss function, have been proposed. In this\nwork, we investigate the behavior of these loss functions and their sensitivity\nto learning rate tuning in the presence of different rates of label imbalance\nacross 2D and 3D segmentation tasks. We also propose to use the class\nre-balancing properties of the Generalized Dice overlap, a known metric for\nsegmentation assessment, as a robust and accurate deep-learning loss function\nfor unbalanced tasks.","url_abs":"http://arxiv.org/abs/1707.03237v3","url_pdf":"http://arxiv.org/pdf/1707.03237v3.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":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/Adityarajora/qwerty","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/IAmSuyogJadhav/Brainy","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/LucasFidon/GeneralizedWassersteinDiceLoss","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/neshitov/UNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/sidify/areal_image_segmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/ubamba98/Brain-Segmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"generalised-dice-overlap-as-a-deep-learning","repo_url":"https://github.com/wolny/pytorch-3dunet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"sensitivity","task_name":"Sensitivity"}],"methods":[{"method_slug":"dice-loss","method_name":"Dice Loss"}],"datasets_introduced":[],"methods_introduced":[{"slug":"dice-loss","name":"Dice Loss","full_name":"Dice Loss"}],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.03237","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}