A machine learning-based method for estimating the number and orientations of major fascicles in diffusion-weighted magnetic resonance imaging

Davood Karimi, Lana Vasung, Camilo Jaimes, Fedel Machado-Rivas, Shadab Khan, Simon K. Warfield, Ali Gholipour

Abstract: Highlights•We propose a data-driven method for determination of orientations of brain fascicles.•We compute feature vectors that represent signal attenuation with orientation.•An MLP uses the feature vector to predict the angle to the closest fascicle.•Our method accurately estimates the number and orientations of fascicles in a voxel.•Our method also leads to superior tractography, assessed by experts.