Corpus callosum thickness estimation using elastic shape matching

Abstract

We present a shape-based approach for calculating the thickness of the corpus callosum. The corpus callosum is delineated from the MRI midsagittal white matter boundary and represented as a parameterized curve consisting of the top and bottom boundaries by a trained expert. The top and bottom boundaries are first represented in a quotient space of open curves, and then elastically matched under a geometric framework that generates an optimal correspondence between their 'shapes'. This matching is computed using a geodesic between shape representations that are invariant to reparameterizations of the curves. Callosal thickness is given by the distance between matched points on the top and bottom boundaries. Our results within a healthy population of N = 96 subjects show significant differences in callosal thickness computed using elastic matching compared to the direct Euclidean approach.