Tract-oriented statistical group comparison of diffusion in sheet-like white matter

Abstract

Identifying specific structures of the brain where pathology differs between groups of subjects may aid to develop imaging-based markers for disease diagnosis. We propose a new technique for doing multivariate statistical analysis on white matter tracts with sheet like shapes. Previous works assume tube-like shapes, not always suitable for modelling the white matter tracts of the brain. The tract-oriented technique aimed at group studies, integrates the usage of multivariate features and outputs a single value of significance indicating tract-specific differences. This is in contrast to voxel based analysis techniques which outputs a significance per voxel basis, and requires multiple comparison correction. We demonstrate our technique by comparing a group of controls with a group of Multiple Sclerosis subjects obtaining significant differences on 11 different fascicle structures.