{"id":2377,"date":"2020-01-17T16:07:37","date_gmt":"2020-01-17T21:07:37","guid":{"rendered":"https:\/\/my.vanderbilt.edu\/masi\/?p=2377"},"modified":"2020-01-17T16:07:37","modified_gmt":"2020-01-17T21:07:37","slug":"harmonizing-1-5-t3t-diffusion-weighted-mri-through-development-of-deep-learning-stabilized-microarchitecture-estimators","status":"publish","type":"post","link":"https:\/\/my.vanderbilt.edu\/masi\/2020\/01\/harmonizing-1-5-t3t-diffusion-weighted-mri-through-development-of-deep-learning-stabilized-microarchitecture-estimators\/","title":{"rendered":"Harmonizing 1.5 T\/3T diffusion weighted MRI through development of deep learning stabilized microarchitecture estimators"},"content":{"rendered":"<p><strong>Nath V, Remedios S, Parvathaneni P, Hansen CB, Bayrak RG, Bermudez C, Blaber JA, Schilling KG, Janve VA, Gao Y, Huo Y. Harmonizing 1.5 T\/3T diffusion weighted MRI through development of deep learning stabilized microarchitecture estimators. In Medical Imaging 2019: Image Processing 2019 Mar 15 (Vol. 10949, p. 109490O). International Society for Optics and Photonics.<\/strong><\/p>\n<h2>Abstract<\/h2>\n<p>Diffusion weighted magnetic resonance imaging (DW-MRI) is interpreted as a quantitative method that is sensitive to<br \/>\ntissue microarchitecture at a millimeter scale. However, the sensitization is dependent on acquisition sequences (e.g.,<br \/>\ndiffusion time, gradient strength, etc.) and susceptible to imaging artifacts. Hence, comparison of quantitative DW-MRI<br \/>\nbiomarkers across field strengths (including different scanners, hardware performance, and sequence design<br \/>\nconsiderations) is a challenging area of research. We propose a novel method to estimate microstructure using DW-MRI<br \/>\nthat is robust to scanner difference between 1.5T and 3T imaging. We propose to use a null space deep network (NSDN)<br \/>\narchitecture to model DW-MRI signal as fiber orientation distributions (FOD) to represent tissue microstructure. The<br \/>\nNSDN approach is consistent with histologically observed microstructure (on previously acquired ex vivo squirrel monkey<br \/>\ndataset) and scan-rescan data. The contribution of this work is that we incorporate identical dual networks (IDN) to<br \/>\nminimize the influence of scanner effects via scan-rescan data. Briefly, our estimator is trained on two datasets. First, a<br \/>\nhistology dataset was acquired on three squirrel monkeys with corresponding DW-MRI and confocal histology (512<br \/>\nindependent voxels). Second, 37 control subjects from the Baltimore Longitudinal Study of Aging (67-95 y\/o) were<br \/>\nidentified who had been scanned at 1.5T and 3T scanners (b-value of 700 s\/mm2, voxel resolution at 2.2mm, 30-32 gradient<br \/>\nvolumes) with an average interval of 4 years (standard deviation 1.3 years). After, image registration, we used paired white<br \/>\nmatter (WM) voxels for 17 subjects and 440 histology voxels for training and 20 subjects and 72 histology voxels for<br \/>\ntesting. We compare the proposed estimator with super-resolved constrained spherical deconvolution (CSD) and a<br \/>\npreviously presented regression deep neural network (DNN). NSDN outperformed CSD and DNN in angular correlation<br \/>\ncoefficient (ACC) 0.81 versus 0.28 and 0.46, mean squared error (MSE) 0.001 versus 0.003 and 0.03, and general fractional<br \/>\nanisotropy (GFA) 0.05 versus 0.05 and 0.09. Further validation and evaluation with contemporaneous imaging are<br \/>\nnecessary, but the NSDN is promising avenue for building understanding of microarchitecture in a consistent and device-independent manner.<\/p>\n<figure id=\"attachment_2378\" aria-describedby=\"caption-attachment-2378\" style=\"width: 650px\" class=\"wp-caption alignnone\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/my.vanderbilt.edu\/masi\/wp-content\/uploads\/sites\/2304\n2661\/2020\/01\/Fig_2_Neural_Network-650x299.png\" alt=\"The base regression neural network is depicted in the center which describes the parameters of the fullyconnected dense layers with respective activation functions. The plot at left: Depicts three inputs where the center input comes from corresponding DW-MRI with histology. The other two are pairwise inputs from corresponding voxels of scanner 1.5T and 3T. The plot at right: Depicts the loss function which uses the hypothesis that the outcome\/prediction should be same irrespective of the scanner gradient strength. \" width=\"650\" height=\"299\" class=\"size-large wp-image-2378\" srcset=\"https:\/\/cdn.vanderbilt.edu\/t2-my\/my-prd\/wp-content\/uploads\/sites\/2304\/2020\/01\/Fig_2_Neural_Network-650x299.png 650w, https:\/\/cdn.vanderbilt.edu\/t2-my\/my-prd\/wp-content\/uploads\/sites\/2304\/2020\/01\/Fig_2_Neural_Network-300x138.png 300w, https:\/\/cdn.vanderbilt.edu\/t2-my\/my-prd\/wp-content\/uploads\/sites\/2304\/2020\/01\/Fig_2_Neural_Network-768x354.png 768w, https:\/\/cdn.vanderbilt.edu\/t2-my\/my-prd\/wp-content\/uploads\/sites\/2304\/2020\/01\/Fig_2_Neural_Network.png 1212w\" sizes=\"auto, (max-width: 650px) 100vw, 650px\" \/><figcaption id=\"caption-attachment-2378\" class=\"wp-caption-text\">The base regression neural network is depicted in the center which describes the parameters of the fullyconnected dense layers with respective activation functions. The plot at left: Depicts three inputs where the center input<br \/>comes from corresponding DW-MRI with histology. The other two are pairwise inputs from corresponding voxels of<br \/>scanner 1.5T and 3T. The plot at right: Depicts the loss function which uses the hypothesis that the outcome\/prediction<br \/>should be same irrespective of the scanner gradient strength.<\/figcaption><\/figure>\n","protected":false},"excerpt":{"rendered":"<p>Nath V, Remedios S, Parvathaneni P, Hansen CB, Bayrak RG, Bermudez C, Blaber JA, Schilling KG, Janve VA, Gao Y, Huo Y. Harmonizing 1.5 T\/3T diffusion weighted MRI through development of deep learning stabilized microarchitecture estimators. In Medical Imaging 2019: Image Processing 2019 Mar 15 (Vol. 10949, p. 109490O). International Society for Optics and Photonics&#8230;.<\/p>\n","protected":false},"author":6319,"featured_media":2378,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[130,9,132,144,82,4,65,40],"tags":[],"class_list":["post-2377","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-deep-learning","category-diffusion-weighted-mri","category-harmonization","category-longitudinal","category-magnetic-resonance-imaging","category-neuroimaging","category-registration","category-reproducibility"],"_links":{"self":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2377","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\/6319"}],"replies":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/comments?post=2377"}],"version-history":[{"count":2,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2377\/revisions"}],"predecessor-version":[{"id":2380,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2377\/revisions\/2380"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media\/2378"}],"wp:attachment":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media?parent=2377"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/categories?post=2377"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/tags?post=2377"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}