Analysis of education effects on cognitive reserve based on network flow

Abstract

In this thesis, we investigate the effects of education on cognitive reserve using WM brain networks in order to support hypothesized neural mechanisms.We first model a brain system as a WM brain network of which the node and edge sets represent the brain regions and the fiber tract bundles in WM connecting them, respectively.To trace the boundary of each region, we develop a novel triangulation-invariant method for computing the anisotropic geodesic (AG) map under the metric of curvature-minimizing speed function on a mesh of the cerebral surface.To make the fiber tracts in each edge have geometric homogeneity, we develop an example-based method for automatically classifying the fiber tracts into anatomic bundles according to their shapes and trajectories.We then identify two sub-networks of WM brain networks of which the connectivities are correlated with education levels, based on a graph-theoretical notion of maximum flow.We believe that these sub-networks be regarded as an important evidence that supports the education effects on cognitive reserve from a neural mechanism point of view.The contributions of this thesis are three-fold:First, we present a triangulation-invariant method for efficiently computing the AG map on a cerebral surface mesh to post process the boundary of a brain region.Next, we propose an example-based method for automatic classification of WM fiber tracts to construct the edges connecting the nodes.Finally, we identify two sub-networks to support hypothesized neural mechanisms, by analyzing the WM brain network based on maximum flow.