{"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/diagnose-like-a-radiologist-attention-guided","title":"Diagnose like a Radiologist: Attention Guided Convolutional Neural Network for Thorax Disease Classification","arxiv_id":"1801.09927","date":"2018-01-30","proceeding":null,"authors":["Qingji Guan","Yaping Huang","Zhun Zhong","Zhedong Zheng","Liang Zheng","Yi Yang"],"abstract":"This paper considers the task of thorax disease classification on chest X-ray\nimages. Existing methods generally use the global image as input for network\nlearning. Such a strategy is limited in two aspects. 1) A thorax disease\nusually happens in (small) localized areas which are disease specific. Training\nCNNs using global image may be affected by the (excessive) irrelevant noisy\nareas. 2) Due to the poor alignment of some CXR images, the existence of\nirregular borders hinders the network performance. In this paper, we address\nthe above problems by proposing a three-branch attention guided convolution\nneural network (AG-CNN). AG-CNN 1) learns from disease-specific regions to\navoid noise and improve alignment, 2) also integrates a global branch to\ncompensate the lost discriminative cues by local branch. Specifically, we first\nlearn a global CNN branch using global images. Then, guided by the attention\nheat map generated from the global branch, we inference a mask to crop a\ndiscriminative region from the global image. The local region is used for\ntraining a local CNN branch. Lastly, we concatenate the last pooling layers of\nboth the global and local branches for fine-tuning the fusion branch. The\nComprehensive experiment is conducted on the ChestX-ray14 dataset. We first\nreport a strong global baseline producing an average AUC of 0.841 with\nResNet-50 as backbone. After combining the local cues with the global\ninformation, AG-CNN improves the average AUC to 0.868. While DenseNet-121 is\nused, the average AUC achieves 0.871, which is a new state of the art in the\ncommunity.","url_abs":"http://arxiv.org/abs/1801.09927v1","url_pdf":"http://arxiv.org/pdf/1801.09927v1.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":"diagnose-like-a-radiologist-attention-guided","repo_url":"https://github.com/guide2157/ChulaXrayClassifier","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1801.09927","atlas_url":"https://app.syntology.ai/?focus=1801.09927","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}