Abstract #4474

Mapping white matter specificity captured by diffusion tractography through deep learning on structural MRI

Qi Yang¹, Colin Hansen¹, Francois Rheault², Bramsh Qamar³, Owen Williams⁴, Susan Resnick⁴, Eleftherios Garyfallidis³, Adam W Anderson^5,6, Maxime Descoteaux², Bennett A Landman^5,6,7,8, and Kurt G Schilling⁵

¹Computer Science, Vanderbilt University, Nashville, TN, United States, ²Sherbrooke Connectivity Imaging Laboratory (SCIL), Universite de Sherbrooke, Sherbrooke, QC, Canada, ³Intelligent Systems Engineering, Indiana University Bloomington, Bloomington, IN, United States, ⁴Laboratory of Behavioral Neuroscience, National Institute on Aging, Baltimore, MD, United States, ⁵Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, ⁶Biomedical Engineering, Vanderbilt University, Nashville, TN, United States, ⁷Radiology and Radiological Sciences, Vanderbilt University Medical Center, Nashville, TN, United States, ⁸Electrical Engineering, Vanderbilt University, Nashville, TN, United States

We investigate the value added by diffusion fiber tractography dissection of white matter pathways over standard T1w images in which WM is largely homogenous. We analyze structural differences in pathway segmentations between a deep learning network trained to label WM directly on T1w images and the same pathways dissected using tractography, and find that although the core of the stem of many fiber bundles is accurately delineated, the value of tractography is in delineating terminal connections and determining fine-scale stem geometrical properties. Thus, while localizing regions-of-interest is possible in structural images, tractography is needed for increased pathway and connection information.

This abstract and the presentation materials are available to members only; a login is required.

Join Here