Capturing subject variability in data driven fMRI analysis: A graph theoretical comparison

Recent simulation studies, using functional magnetic resonance imaging (fMRI) like data, have shown that independent vector analysis (IVA) is a superior solution for capturing subject variability when compared to the popular group independent component analysis. This is of fundamental importance for identifying group differences which is a common goal of medical research. Nevertheless, there have not been similar studies on the effectiveness of IVA using real fMRI data. The main difficulties when working with real data are the lack of a ground truth and the high variability among subjects when performing the analysis. In this paper, we present a graph-theoretic approach to effectively compare an algorithm's ability to capture subject variability for real fMRI data and also address the important issue of order selection for capturing subject variability.