Towards Visual Mega-Analysis of Voxel-based Measurement in Brain Cohorts

Abstract

We present a visualization prototype for comparative analysis of factional anisotropy (FA) distributions constructed from three-dimensional (3D) brain diffusion tensor imaging (DTI) in brain cohorts. The prototype lets brain scientists examine meta-analysis (the pooled analysis of multiple smaller trials or multi-site studies) results for identifying differences in cohorts. Interactive side-by-side bar charts show multiple statistical results of FA comparisons in regions of interest (ROIs) defined by user-chosen atlas. An occlusion-free two-dimensional (2D) semantic merge tree further displays the global distribution of FA values. Two histograms on each tree arc reveal voxel-based FA distributions represented by that arc branch in cohorts. Interaction techniques support brushing-and-linking of local and global ROIs queries. ROIs can be defined from atlas or select through interaction. We report validation results in a case study and an interview.