Meeting Banner
Abstract #1874

Predicting T1 Information from Diffusion Image Data

Yogesh Rathi1, Oleg Michailovich2, Sylvain Bouix, Martha Shenton, Carl-Fredrik Westin

1Harvard Medical School, Boston, MA, United States; 2ECE, University of Waterloo, Waterloo, Canada

In this work, we propose a novel method for obtaining a T1-weighted MR image from a diffusion MRI scan. Existing measures of diffusion, such as, FA, entropy, etc. do not provide enough contrast between gray matter and CSF regions. The proposed algorithm produces images that can better delineate the different tissue types just like a T1 image. The predicted T1 images can be used in registration, segmentation and visualization of fiber tracts.

Keywords

ability actual allow anisotropy applications around atlas avoid axial background basis becomes better brain bundles certain clear coefficients complex computer constructed contrast cortical creating curvature curve details diffusion directly display distortions eddy either elements energy entropy equation error every fact feature features fiber filtered forest forests forming general generalized good gray hand harmonic improved indeed independent instead intricate invariant issue kurtosis learn learning like machine macroscopic many mapping maps mathematical measures medical methodology modeling motion namely neural next none normalized note notice often orientation orthogonal partial predict predicted prediction processed propose proposed random registration remove requires rigid robust rotations scalar sett since snippets space spatial squared strategy studies subjects superior surfer tensor theory tissue tracts traditionally training trans tree types unseen variance vector ventricles vision visualization visualize volumetric waterloo ween white widely