{"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/high-resolution-breast-cancer-screening-with","title":"High-Resolution Breast Cancer Screening with Multi-View Deep Convolutional Neural Networks","arxiv_id":"1703.07047","date":"2017-03-21","proceeding":null,"authors":["Krzysztof J. Geras","Stacey Wolfson","Yiqiu Shen","Nan Wu","S. Gene Kim","Eric Kim","Laura Heacock","Ujas Parikh","Linda Moy","Kyunghyun Cho"],"abstract":"Advances in deep learning for natural images have prompted a surge of\ninterest in applying similar techniques to medical images. The majority of the\ninitial attempts focused on replacing the input of a deep convolutional neural\nnetwork with a medical image, which does not take into consideration the\nfundamental differences between these two types of images. Specifically, fine\ndetails are necessary for detection in medical images, unlike in natural images\nwhere coarse structures matter most. This difference makes it inadequate to use\nthe existing network architectures developed for natural images, because they\nwork on heavily downscaled images to reduce the memory requirements. This hides\ndetails necessary to make accurate predictions. Additionally, a single exam in\nmedical imaging often comes with a set of views which must be fused in order to\nreach a correct conclusion. In our work, we propose to use a multi-view deep\nconvolutional neural network that handles a set of high-resolution medical\nimages. We evaluate it on large-scale mammography-based breast cancer screening\n(BI-RADS prediction) using 886,000 images. We focus on investigating the impact\nof the training set size and image size on the prediction accuracy. Our results\nhighlight that performance increases with the size of training set, and that\nthe best performance can only be achieved using the original resolution. In the\nreader study, performed on a random subset of the test set, we confirmed the\nefficacy of our model, which achieved performance comparable to a committee of\nradiologists when presented with the same data.","url_abs":"http://arxiv.org/abs/1703.07047v3","url_pdf":"http://arxiv.org/pdf/1703.07047v3.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":"high-resolution-breast-cancer-screening-with","repo_url":"https://github.com/nyukat/BIRADS_classifier","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"high-resolution-breast-cancer-screening-with","repo_url":"https://github.com/saidbm24/CNN-for-BIRADS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"breast-cancer-detection","task_name":"Breast Cancer Detection"},{"task_slug":"medical-diagnosis","task_name":"Medical Diagnosis"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}