A DTI-Derived Measure of Cortico-Cortical Connectivity

Abstract

We arm researchers with a simple method to chart a macroscopic cortico-cortical connectivity network in living human subjects. The researcher provides a diffusion-magnetic resonance imaging (MRI) data set and N cortical regions of interest. In return, we provide an N xN structural adjacency matrix (SAM) quantifying the relative connectivity between all cortical region pairs. We also return a connectivity map for each pair to enable visualization of interconnecting fiber bundles. The measure of connectivity we devise is: 1) free of length bias, 2) proportional to fiber bundle cross-sectional area, and 3) invariant to an exchange of seed and target. We construct a 3-D lattice scaffolding (graph) for white-matter by drawing a link between each pair of voxels in a 26-voxel neighborhood for which their two respective principal eigenvectors form a sufficiently small angle. The connectivity between a cortical region pair is then measured as the maximum number of link-disjoint paths that can be established between them in the white-matter graph. We devise an efficient Edmonds-Karp-like algorithm to compute a conservative bound on the maximum number of link-disjoint paths. Using both simulated and authentic diffusion-tensor imaging data, we demonstrate that the number of link-disjoint paths as a measure of connectivity satisfies properties 1)-3), unlike the fraction of intersecting streamlines-the measure intrinsic to most existing probabilistic tracking algorithms. Finally, we present connectivity maps of some notoriously difficult to track longitudinal and contralateral fasciculi.