Effect of Spherical Harmonic Coefficients on Fiber Bundle Segmentation

Nowadays classifying the brain white matter fibers into the distinct object named as bundles inside which the same characteristics on local diffusivity or shape and length of fibers exist, is of a growing interest in neuro-imaging fields. In this paper we present a novel method for segmenting the fiber bundles using Spherical Harmonic Coefficients (SHC) which describe diffusion signal obtained from High Angular Resolution Diffusion Imaging (HARDI) protocols. Using SH coefficients in defining of a similarity measure being used as speed function term in Hamilton-Jacobi equation with Level set framework as an implicit numericalsolution, we have shown that our method has advantages over methods using similarity measures based on DTI field by proper propagating of the front within fiber crossing areas. Without any assumption about diffusion profile or model by dealing with just diffusion signals instead of diffusion probability function most used in other studies, out results on synthetic data as well as real HARD MRI data are surely closer to reality.