A Fiber Tracking Algorithm Based on Non-Local Constrained Spherical Deconvolution

Abstract

Fiber tracking with diffusion magnetic resonance imaging provides a powerful tool for non-invasive observation of white matter in the brain. Constrained spherical deconvolution (CSD) is a multi-fiber tracking model, which can model the orientation of fibers in the voxel and achieve brain fiber reconstruction. This paper proposes a deterministic fiber tracking algorithm based on a non-local CSD model that combines neighborhood information and fractional regularization. The algorithm aimed to solve the ill-posed problem and loss detailed information in the conventional CSD model. The nonlocality of fractional order reduced the errors of fiber orientation distribution estimation, and the neighborhood information was used to ensure spatial consistency, reducing the effects of random noise. Simulation data and experimental human brain data were used to compare the performance of the proposed algorithm and the conventional CSD deterministic tracking algorithm. The results demonstrated that the proposed algorithm produced not only better overall visual effect, but also more complete and accurate reconstruction of the crossing fibers.

Keywords

• diffusion magnetic resonance imaging
• fiber tracking
• constrained spherical deconvolution
• neighborhood information
• fractional regularization