{"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/3d-whole-brain-segmentation-using-spatially","title":"3D Whole Brain Segmentation using Spatially Localized Atlas Network Tiles","arxiv_id":"1903.12152","date":"2019-03-28","proceeding":null,"authors":["Yuankai Huo","Zhoubing Xu","Yunxi Xiong","Katherine Aboud","Prasanna Parvathaneni","Shunxing Bao","Camilo Bermudez","Susan M. Resnick","Laurie E. Cutting","Bennett A. Landman"],"abstract":"Detailed whole brain segmentation is an essential quantitative technique,\nwhich provides a non-invasive way of measuring brain regions from a structural\nmagnetic resonance imaging (MRI). Recently, deep convolution neural network\n(CNN) has been applied to whole brain segmentation. However, restricted by\ncurrent GPU memory, 2D based methods, downsampling based 3D CNN methods, and\npatch-based high-resolution 3D CNN methods have been the de facto standard\nsolutions. 3D patch-based high resolution methods typically yield superior\nperformance among CNN approaches on detailed whole brain segmentation (>100\nlabels), however, whose performance are still commonly inferior compared with\nmulti-atlas segmentation methods (MAS) due to the following challenges: (1) a\nsingle network is typically used to learn both spatial and contextual\ninformation for the patches, (2) limited manually traced whole brain volumes\nare available (typically less than 50) for training a network. In this work, we\npropose the spatially localized atlas network tiles (SLANT) method to\ndistribute multiple independent 3D fully convolutional networks (FCN) for\nhigh-resolution whole brain segmentation. To address the first challenge,\nmultiple spatially distributed networks were used in the SLANT method, in which\neach network learned contextual information for a fixed spatial location. To\naddress the second challenge, auxiliary labels on 5111 initially unlabeled\nscans were created by multi-atlas segmentation for training. Since the method\nintegrated multiple traditional medical image processing methods with deep\nlearning, we developed a containerized pipeline to deploy the end-to-end\nsolution. From the results, the proposed method achieved superior performance\ncompared with multi-atlas segmentation methods, while reducing the\ncomputational time from >30 hours to 15 minutes\n(https://github.com/MASILab/SLANTbrainSeg).","url_abs":"http://arxiv.org/abs/1903.12152v1","url_pdf":"http://arxiv.org/pdf/1903.12152v1.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":"3d-whole-brain-segmentation-using-spatially","repo_url":"https://github.com/MASILab/SLANTbrainSeg","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"caffe2","reach":null},{"paper_slug":"3d-whole-brain-segmentation-using-spatially","repo_url":"https://github.com/MASILab/SLANT_brain_seg","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":null}],"tasks":[{"task_slug":"brain-segmentation","task_name":"Brain Segmentation"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"segmentation","task_name":"Segmentation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1903.12152","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}