Diffusion coefficient orientation distribution function for diffusion magnetic resonance imaging.

BACKGROUND: Diffusion magnetic resonance imaging (dMRI) is a popular non-invasive imaging technique applied for the study of nerve fibers in vivo, with diffusion tensor imaging (DTI) and high angular resolution diffusion imaging (HARDI) as the commonly used dMRI methods. However, DTI cannot resolve complex fiber orientations in a local area and HARDI lacks a solid physical basis.

NEW METHOD: We introduce a diffusion coefficient orientation distribution function (DCODF). It has a clear physical meaning to represent the orientation distribution of diffusion coefficients for Gaussian and non-Gaussian diffusion. Based on DCODF, we then propose a new HARDI method, termed as diffusion coefficient orientation distribution transform (DCODT), to estimate the orientation distribution of nerve fibers in voxels.

RESULTS: The method is verified on the simulated data, ISMRM-2015-Tracto-challenge data, and HCP datasets. The results show the superior capability of DCODT in resolving the complex distribution of multiple fiber bundles effectively.

COMPARISON WITH EXISTING METHOD(S): The method is compared to other common model-free HARDI estimators. In the numerical simulations, DCODT achieves a better trade-off between the resolution and accuracy than the counterparts for high b-values. In the comparisons based on the challenge data, the improvement of DCODT is significant in scoring. The results on the HCP datasets show that DCODT provides fewer spurious lobes in the glyphs, resulting in more coherent fiber orientations.

CONCLUSIONS: We conclude that DCODT may be a reliable method to extract accurate information about fiber orientations from dMRI data and promising for the study of neural architecture.