{"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/inverse-consistent-deep-networks-for","title":"Inverse-Consistent Deep Networks for Unsupervised Deformable Image Registration","arxiv_id":"1809.03443","date":"2018-09-10","proceeding":null,"authors":["Jun Zhang"],"abstract":"Deformable image registration is a fundamental task in medical image\nanalysis, aiming to establish a dense and non-linear correspondence between a\npair of images. Previous deep-learning studies usually employ supervised neural\nnetworks to directly learn the spatial transformation from one image to\nanother, requiring task-specific ground-truth registration for model training.\nDue to the difficulty in collecting precise ground-truth registration,\nimplementation of these supervised methods is practically challenging. Although\nseveral unsupervised networks have been recently developed, these methods\nusually ignore the inherent inverse-consistent property (essential for\ndiffeomorphic mapping) of transformations between a pair of images. Also,\nexisting approaches usually encourage the to-be-estimated transformation to be\nlocally smooth via a smoothness constraint only, which could not completely\navoid folding in the resulting transformation. To this end, we propose an\nInverse-Consistent deep Network (ICNet) for unsupervised deformable image\nregistration. Specifically, we develop an inverse-consistent constraint to\nencourage that a pair of images are symmetrically deformed toward one another,\nuntil both warped images are matched. Besides using the conventional smoothness\nconstraint, we also propose an anti-folding constraint to further avoid folding\nin the transformation. The proposed method does not require any supervision\ninformation, while encouraging the diffeomoprhic property of the transformation\nvia the proposed inverse-consistent and anti-folding constraints. We evaluate\nour method on T1-weighted brain magnetic resonance imaging (MRI) scans for\ntissue segmentation and anatomical landmark detection, with results\ndemonstrating the superior performance of our ICNet over several\nstate-of-the-art approaches for deformable image registration. Our code will be\nmade publicly available.","url_abs":"http://arxiv.org/abs/1809.03443v1","url_pdf":"http://arxiv.org/pdf/1809.03443v1.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":"inverse-consistent-deep-networks-for","repo_url":"https://github.com/yf817/ICNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"inverse-consistent-deep-networks-for","repo_url":"https://github.com/zhangjun001/ICNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"anatomical-landmark-detection","task_name":"Anatomical Landmark Detection"},{"task_slug":"image-registration","task_name":"Image Registration"},{"task_slug":"medical-image-analysis","task_name":"Medical Image Analysis"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.03443","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}