Modeling White-Matter Fiber-Orientation Uncertainty for Improved Probabilistic Tractography

Tractography uses fiber-orientation estimates to trace the likely paths of white-matter tracts through the brain, in order to map brain connectivity non-invasively. In this paper, we propose a novel probabilistic framework for modeling fiber-orientation uncertainty and improve probabilistic tractography. The main innovation in the present formulation consists in coupling a particle filtering process with a clustered-mixture model approach to model directional data. Mixtures of von Mises-Fisher (vMF) distributions are used to support the probabilistic estimation of intravoxel fiber directions. The fitted parameters of the clustered vMF mixture at each voxel are then used to estimate white-matter pathways using particle filtering techniques. The technique is validated on simulated as well as on real human brain data experiments.