Multi-fiber reconstruction from diffusion MRI using mixture of wisharts and sparse deconvolution

摘要: In this paper, we present a novel continuous mixture of diffusion tensors model for the diffusion-weighted MR signal attenuation. The relationship between mixing distribution and attenuation is shown to be given by Laplace transform defined on space positive definite tensors. when parameterized Wishart distributions (MOW) possess closed form expression its transform, called Rigaut-type function, which provides an alternative Stejskal-Tanner decay. Our naturally leads deconvolution formulation multi-fiber reconstruction. This requires solution ill-conditioned linear system.We several methods show that nonnegative least squares method outperforms all others in achieving accurate sparse solutions presence noise. performance our reconstruction using MOW demonstrated both synthetic real data along with comparisons state-of-the-art techniques.