{"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/joint-weakly-and-semi-supervised-deep","title":"Joint Weakly and Semi-Supervised Deep Learning for Localization and Classification of Masses in Breast Ultrasound Images","arxiv_id":"1710.03778","date":"2017-10-10","proceeding":null,"authors":["Seung Yeon Shin","Soochahn Lee","Il Dong Yun","Sun Mi Kim","Kyoung Mu Lee"],"abstract":"We propose a framework for localization and classification of masses in\nbreast ultrasound (BUS) images. We have experimentally found that training\nconvolutional neural network based mass detectors with large, weakly annotated\ndatasets presents a non-trivial problem, while overfitting may occur with those\ntrained with small, strongly annotated datasets. To overcome these problems, we\nuse a weakly annotated dataset together with a smaller strongly annotated\ndataset in a hybrid manner. We propose a systematic weakly and semi-supervised\ntraining scenario with appropriate training loss selection. Experimental\nresults show that the proposed method can successfully localize and classify\nmasses with less annotation effort. The results trained with only 10 strongly\nannotated images along with weakly annotated images were comparable to results\ntrained from 800 strongly annotated images, with the 95% confidence interval of\ndifference -3.00%--5.00%, in terms of the correct localization (CorLoc)\nmeasure, which is the ratio of images with intersection over union with ground\ntruth higher than 0.5. With the same number of strongly annotated images,\nadditional weakly annotated images can be incorporated to give a 4.5% point\nincrease in CorLoc, from 80.00% to 84.50% (with 95% confidence intervals\n76.00%--83.75% and 81.00%--88.00%). The effects of different algorithmic\ndetails and varied amount of data are presented through ablative analysis.","url_abs":"http://arxiv.org/abs/1710.03778v2","url_pdf":"http://arxiv.org/pdf/1710.03778v2.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":"joint-weakly-and-semi-supervised-deep","repo_url":"https://github.com/syshin1014/wssdl_bus","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}