{"id":3183,"date":"2022-12-04T23:26:16","date_gmt":"2022-12-05T04:26:16","guid":{"rendered":"https:\/\/my.vanderbilt.edu\/masi\/?p=3183"},"modified":"2022-12-04T23:35:20","modified_gmt":"2022-12-05T04:35:20","slug":"deep-whole-brain-segmentation-of-7t-structural-mri","status":"publish","type":"post","link":"https:\/\/my.vanderbilt.edu\/masi\/2022\/12\/deep-whole-brain-segmentation-of-7t-structural-mri\/","title":{"rendered":"Deep whole brain segmentation of 7T structural MRI"},"content":{"rendered":"

Karthik Ramadass, Xin Yu, Leon Y. Cai, Yucheng Tang, Shunxing Bao, Cailey Kerley, Micah D\u2019Archangel, Laura A Barquero, Allen T. Newton, Isabel Gauthier, Rankin Williams McGugin, Benoit M.\u00a0 Dawant, Laurie E. Cutting, Yuankai Huo, Bennett A. Landman SPIE Medical Imaging 2023<\/p>\n

Abstract<\/h2>\n

7T magnetic resonance imaging<\/span> (MRI) has<\/span> the potential to drive<\/span> our understanding of human brain function<\/span> through new<\/span> contrast and enhanced resolution<\/span>. Whole<\/span> brain segmentation is a key neuroimaging technique that allows for region<\/span>–<\/span>by<\/span>–<\/span>region analysis of the brain.<\/span> Segmentation<\/span> is also an important preliminary step that provides spatial and volumetric <\/span>information for<\/span> running other neuroimaging pipeline<\/span>s<\/span>. Spatial<\/span>ly localized atlas network tiles<\/span> (SLANT) is a popular 3D<\/span> convolut<\/span>ional<\/span> neural network <\/span>(CNN) tool that breaks the whole brain segmentation task into localized sub<\/span>–<\/span>tasks. Each<\/span> sub<\/span>–<\/span>task involves a specific spatial location handled by an independent 3D convolu<\/span>t<\/span>ional<\/span> network to provide high<\/span> resolution whole brain segmentation results. SLANT has been widely used to generate whole brain segmentations from <\/span>structural scans acquired on 3T MRI. However, the use of SLANT for whole brain segmentation from structural 7T<\/span> MRI<\/span> scans has not been successful<\/span> due to the inhomogeneous image contrast usually seen across the brain in 7T MRI<\/span>.<\/span> For<\/span> instance, w<\/span>e<\/span> demonstrate<\/span> the<\/span> mean<\/span> percent<\/span> difference<\/span> of SLANT<\/span> label volumes<\/span> between a<\/span> 3T<\/span> scan<\/span>–<\/span>rescan<\/span> is<\/span> approximately 1.73%, whereas<\/span> its 3T<\/span>–<\/span>7T<\/span> scan<\/span>–<\/span>rescan<\/span> counterpart<\/span> has higher differences around 15.13%.<\/span> Our approach<\/span> to address this problem is to<\/span> register<\/span> the whole brain segmentation<\/span> performed on<\/span> 3T MRI<\/span> to<\/span> 7T MRI and use this<\/span> information to finetune SLAN<\/span>T for structural 7T MRI. With the finetuned SLANT pipeline, we observe<\/span> a<\/span> lower mean<\/span> relative difference<\/span> in the label volumes<\/span> of ~8.43%<\/span> acquired from structural 7T MRI data.<\/span> Dice similarity coefficient<\/span> between SLANT segmentation on the 3T MRI scan and the a<\/span>fter finetuning SLANT segmentation on the 7T MRI<\/span> increased from 0.79 to 0.83 with p<0.01. These results suggest finetuning of SLANT is a viable solution for improving <\/span>whole brain segmentation on high resolution 7T structural imaging.<\/span><\/span><\/p>\n

\"The
The sagittal and coronal slices with overlays of segmentations from SLANT on 3T MRI, SLANT on 7T MRI and finetuned SLANT on 7T MRI are shown. Some of the labels such as the left cerebellum white matter show clear improvement in the segmentation boundaries in the sagittal slices.<\/figcaption><\/figure>\n","protected":false},"excerpt":{"rendered":"

Karthik Ramadass, Xin Yu, Leon Y. Cai, Yucheng Tang, Shunxing Bao, Cailey Kerley, Micah D\u2019Archangel, Laura A Barquero, Allen T. Newton, Isabel Gauthier, Rankin Williams McGugin, Benoit M.\u00a0 Dawant, Laurie E. Cutting, Yuankai Huo, Bennett A. Landman SPIE Medical Imaging 2023 Abstract 7T magnetic resonance imaging (MRI) has the potential to drive our understanding of…<\/p>\n","protected":false},"author":7298,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-3183","post","type-post","status-publish","format-standard","hentry","category-news"],"_links":{"self":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/3183","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/users\/7298"}],"replies":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/comments?post=3183"}],"version-history":[{"count":2,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/3183\/revisions"}],"predecessor-version":[{"id":3186,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/3183\/revisions\/3186"}],"wp:attachment":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media?parent=3183"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/categories?post=3183"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/tags?post=3183"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}