{"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/learning-interpretable-anatomical-features","title":"Learning Interpretable Anatomical Features Through Deep Generative Models: Application to Cardiac Remodeling","arxiv_id":"1807.06843","date":"2018-07-18","proceeding":null,"authors":["Carlo Biffi","Ozan Oktay","Giacomo Tarroni","Wenjia Bai","Antonio de Marvao","Georgia Doumou","Martin Rajchl","Reem Bedair","Sanjay Prasad","Stuart Cook","Declan O'Regan","Daniel Rueckert"],"abstract":"Alterations in the geometry and function of the heart define well-established\ncauses of cardiovascular disease. However, current approaches to the diagnosis\nof cardiovascular diseases often rely on subjective human assessment as well as\nmanual analysis of medical images. Both factors limit the sensitivity in\nquantifying complex structural and functional phenotypes. Deep learning\napproaches have recently achieved success for tasks such as classification or\nsegmentation of medical images, but lack interpretability in the feature\nextraction and decision processes, limiting their value in clinical diagnosis.\nIn this work, we propose a 3D convolutional generative model for automatic\nclassification of images from patients with cardiac diseases associated with\nstructural remodeling. The model leverages interpretable task-specific anatomic\npatterns learned from 3D segmentations. It further allows to visualise and\nquantify the learned pathology-specific remodeling patterns in the original\ninput space of the images. This approach yields high accuracy in the\ncategorization of healthy and hypertrophic cardiomyopathy subjects when tested\non unseen MR images from our own multi-centre dataset (100%) as well on the\nACDC MICCAI 2017 dataset (90%). We believe that the proposed deep learning\napproach is a promising step towards the development of interpretable\nclassifiers for the medical imaging domain, which may help clinicians to\nimprove diagnostic accuracy and enhance patient risk-stratification.","url_abs":"http://arxiv.org/abs/1807.06843v1","url_pdf":"http://arxiv.org/pdf/1807.06843v1.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":"learning-interpretable-anatomical-features","repo_url":"https://github.com/UK-Digital-Heart-Project/lvae_mlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"diagnostic","task_name":"Diagnostic"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[{"method_slug":"interpretability","method_name":"Interpretability"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}