Meeting Banner
Abstract #2899

Identification of Mild Traumatic Brain Injuries by Comparison of Free-Water Corrected Z-Distributions

Ofer Pasternak1, Sylvain Bouix1, Yogesh Rathi1, Craig A. Branch2, Carl-Fredrik Westin1, Martha E. Shenton1, 3, Michael L. Lipton4

1Harvard Medical School, Boston, MA, United States; 2The Gruss Magnetic Resonance Research Center, Albert Einstein College of Medicine, Bronx, NY, United States; 3VA Boston Healthcare System, Brockton, MA, United States; 4 The Gruss Magnetic Resonance Research Center, Albert Einstein College of Medicine, Bronx, NY, United States

Detection of brain abnormalities following mild traumatic brain injury (mTBI) is hard since tissue changes are subtle and heterogeneously located, thus not well suited to be detected using group analyses. We study a diagnosis approach in which free-water corrected diffusion MRIs of acute mTBI subjects are individually compared to a normative atlas. By analyzing the distribution of z-scores we are able to demonstrate significant group differences, which were not detected using conventional group analyses. We locate ranges in the distribution that are most specific to the injury, and these infer the types of abnormalities that can be expected in mTBI,

Keywords

able abnormal abnormalities abnormality accurate addition affected albeit allow analyses appear appears application applied applying around award basis behavior blue brain branch characterize circumvent clinical collected college common comparing construct control controls corpus corrected created days decisions decreased defense degeneration detect detection deviation diagnose diagnostic differentiation diffusion discrepancy disorder distribution distributions domains drive establish evaluating expected extreme false feature features findings form free future gradient grant green healthy heterogeneity heterogeneous identification identified identifies included indicate indicates individual injuries injury insight learn learning location many maps measure measures medicine mild moreover negative noted occur often onto optimized pathologies patient patients patterns plot plots population positive positives post predicts problem projected protocol registered regressed requires resolution robust scanned school score scores screening seem sensitive sensitivity significantly since skeleton software spatial specificity stars statistically subject subjects suggest suggesting suited summing sums support system template thresholded tomography tool traumatic twenty types utilized water white yield