Bayesian inference of intravoxel structure in diffusion MRI

While diffusion tensor imaging (DTI) provides a powerful tool to reconstruct neural pathways in vivo, the standard diffusion tensor model is limited to resolve a single fiber direction within each voxel. To overcome this difficulty, high angular resolution diffusion imaging (HARDI) has recently been proposed to investigate intravoxel fiber heterogeneity. In this paper we propose a novel method for mixture model decomposition of the HARDI signal based on Bayesian inference and trans-dimensional Markov Chain simulation. The method is applied to both synthetic and real data.