Surface-based modeling of white matter fasciculi with orientation encoding

Abstract

In this paper, we describe a novel technique for modeling sheet-like white matter (WM) fasciculi using continuous medial representation (cm-rep). In the cm-rep framework, the skeleton of a fasciculus is described by a parametric surface patch. This modeling scheme is particularly appropriate for sheet-like structures, because the shapes of such objects can be effectively captured by their skeletons. We show that dimensionality reduction can be achieved without much loss of spatial specificity by projecting data along the "less interesting" thickness direction onto the skeletons. We demonstrate that local fiber orientation of the modeled fasciculi can be encoded in our framework and show how this information can be leveraged for deriving and analyzing brain connectivity patterns on the skeleton themselves.