Spherical Deconvolution Algorithm For High Angular Resolution Diffusion Imaging

Using diffusion weighted magnetic resonance imaging method to reconstruct the white matter nerve fiber is an important way to display the distribution of the white matter nerve fiber between functional regions of the brain in vivo. It has vital significance to the brain function cognitive, brain surgery navigation and mental illness research. Diffusion tensor imaging is the most commonly used fiber reconstruction method in the recent, but the inherent limitation of it is that it can’t resolve multiple fibers within one voxel. So many kinds of multi-fiber reconstruction methods are proposed to solve the reconstruction problem of crossing fibers. Among them, the spherical deconvolution method is an effective method. Improving the angular resolution of the spherical deconvolution model is the research hotspot in recent years.In order to improve the angular resolution, accuracy, robustness, and efficiency of spherical deconvolution methods, including the discrete fiber orientation distribution function based methods and continuous fiber orientation distribution function based methods, this paper mainly works as follows:Firstly, a deconvoluton kernel of signal response function is proposed in a kind of simple forms and estimation algorithm. Based on the principle of the deconvolution algorithm, the of the effect to the fiber orientation distribution function(FOD) get from the deviated response function is obtain through experiments, and further the effect of the DWI data acquisition parameter to the FOD is obtained. The experimental results show the validation of the proposed simple form response function.Secondly, regard to discrete fiber orientation distribution function based spherical deconvolution algorithm with better angular resolution but bad discrete deviation error drawback, a mining fitting procession method is proposed. Firstly, using the discrete fiber orientation distribution function to establish the spherical deconvolution model, and then using the spherical harmonic functions to fitting the discrete fiber orientation distribution function to compensate the discrete error. The experimental results show that, this algorithm has a higher angular resolution than the continuous fiber orientation distribution function based algorithm, and can improve the efficiency more than13times.Finally, in view of the continuous fiber orientation distribution function based spherical deconvolution algorithm, which the angular resolution is rely on the frequency of the spherical harmonic components, as a result affecting the reconstruction ability to small angular crossing fibers, a heuristic iterative algorithm and fiber orientation distribution information heuristic information extract strategy to improve the stability of the super-resolution solution. The simulated dataset and real clinical dataset demonstrated the proposed algorithm can improve the angular resolution under the premise of keep the robustness and accuracy.