Spherical deconvolution high order tensor multi-fiber reconstruction method

We proposed a spherical deconvolution algorithm to recover the multi-fiber structure. We introduced the high order tensor as the fiber orientation distribution function model which is able to represent complex functions on the sphere and has advantage of simple polynomial form. The spherical doconvolution method regard the diffusion weighted signal as the convolution of the response function with a fiber orientation distribution function on the sphere, then use deconvolution algorithm to recover the fiber orientation distribution function from the diffusion weighted signal. As the common problem of ill condition in the spherical deconvolution, we propose a heuristic algorithm to solve it. We studied the algorithm using simulated diffusion weighted signals of two fibers crossing at different angles and different noise levels. We compared our algorithm to the CSD algorithm and the experimental results demonstrate that our algorithm has higher angular resolution.