{"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/holistic-and-comprehensive-annotation-of","title":"Holistic and Comprehensive Annotation of Clinically Significant Findings on Diverse CT Images: Learning from Radiology Reports and Label Ontology","arxiv_id":"1904.04661","date":"2019-04-09","proceeding":"CVPR 2019 6","authors":["Ke Yan","Yifan Peng","Veit Sandfort","Mohammadhadi Bagheri","Zhiyong Lu","Ronald M. Summers"],"abstract":"In radiologists' routine work, one major task is to read a medical image,\ne.g., a CT scan, find significant lesions, and describe them in the radiology\nreport. In this paper, we study the lesion description or annotation problem.\nGiven a lesion image, our aim is to predict a comprehensive set of relevant\nlabels, such as the lesion's body part, type, and attributes, which may assist\ndownstream fine-grained diagnosis. To address this task, we first design a deep\nlearning module to extract relevant semantic labels from the radiology reports\nassociated with the lesion images. With the images and text-mined labels, we\npropose a lesion annotation network (LesaNet) based on a multilabel\nconvolutional neural network (CNN) to learn all labels holistically.\nHierarchical relations and mutually exclusive relations between the labels are\nleveraged to improve the label prediction accuracy. The relations are utilized\nin a label expansion strategy and a relational hard example mining algorithm.\nWe also attach a simple score propagation layer on LesaNet to enhance recall\nand explore implicit relation between labels. Multilabel metric learning is\ncombined with classification to enable interpretable prediction. We evaluated\nLesaNet on the public DeepLesion dataset, which contains over 32K diverse\nlesion images. Experiments show that LesaNet can precisely annotate the lesions\nusing an ontology of 171 fine-grained labels with an average AUC of 0.9344.","url_abs":"http://arxiv.org/abs/1904.04661v2","url_pdf":"http://arxiv.org/pdf/1904.04661v2.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":"holistic-and-comprehensive-annotation-of","repo_url":"https://github.com/rsummers11/CADLab","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"holistic-and-comprehensive-annotation-of","repo_url":"https://github.com/HowardZhang002/MULAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"holistic-and-comprehensive-annotation-of","repo_url":"https://github.com/leeh43/Singularity_Deeplesion","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"metric-learning","task_name":"Metric Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1904.04661","atlas_url":"https://app.syntology.ai/?focus=1904.04661","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}