Ranking fMRI time courses by minimum spanning trees: assessing coactivation in fMRI.

In fMRI, time courses with similar temporal "activation" patterns may belong to different brain regions (i.e., these regions are functionally connected, coactivated). A group of time courses (TCs) corresponding to a particular type of temporal activation pattern should be maximally self-consistent (homogeneous). We demonstrate that ordering a group of multidimensional fMRI time courses by a minimum spanning tree (MST) may be used to investigate the temporal homogeneity of a group of TCs. We show the utility of MST ranking for data-driven analysis methods in investigating coactivation in fMRI. MST ranking is equally useful for hypothesis-led methods. Furthermore, MST ranking enables pairwise comparisons of groups/clusters (i.e., any collection of TCs, no matter how derived) of fMRI time courses.