A level set method for building anatomical connectivity paths between brain areas using DTI

In this paper, we present an approach for building paths representing the anatomical connections between some given regions of the brain. The method combines diffusion tensor imaging (DTI) and fast marching techniques, representing an anatomical connection as the shortest path for some global minimization criterion. Paths are generated by growing level sets, looking for saddle points and backpropagating. We show that this method leads to a more robust connection tracking that efficiently deals with fiber crossing. Results on simulations are shown as an illustration.