Probabilistic anatomical connectivity using completion fields

Abstract

Diffusion magnetic resonance imaging has led to active research in the analysis of anatomical connectivity in the brain. Many approaches have been proposed to model the diffusion signal and to obtain estimates of fibre tracts. Despite these advances, the question of defining probabilistic connectivity indices which utilize the relevant information in the diffusion MRI signal to indicate connectivity strength, remains largely open. To address this problem we introduce a novel numerical implementation of a stochastic completion field algorithm, which models the diffusion of water molecules in a medium while incorporating the local diffusion MRI data. We show that the approach yields a valid probabilistic estimate of connectivity strength between two seed regions, with experimental results on the MICCAI 2009 Fibre Cup phantom[1]. © 2010 Springer-Verlag.