SMT: Split/merge fiber tractography for MR-DTI

Abstract

Magnetic Resonance Diffusion Tensor Imaging is a recent imaging modalitywhich has shown promise as a non-invasive tool for estimating the orientation andquantity of white matter tracts in vivo. It has been shown that the estimated diffusion tensor̕s principal diffusion direction coincides with the fiber orientations, given that thetensor in question is anisotropic. MR-DTI fiber tractography aims at following theseprincipal diffusion directions to reconstruct fiber paths. The conventional approach isto use integration techniques, i.e. to follow the principal diffusion directions. The goal of this project is to introduce a new technique for estimation and visualizationof fiber tracts. The proposed Split/Merge Tractography (SMT) tries toovercome the disadvantages of existing techniques. SMT̕s approach is to generateshort (thus more reliable) fiber tracts by conventional techniques (Splitting step) andthen group these short tracks according to an estimated distribution (Merging step).SMT allows branching and does not mask the inherent resolution limitation of the data.The aforementioned distribution is estimated via the Metropolis-Hastings Method.