Blind spherical deconvolution of multi-shell diffusion MRI to model regional changes in pathology

Motivation: Diffusion-weighted MRI (dMRI) has significantly enhanced our ability to investigate the brain's microstructure, but analysis in pathology remains difficult. Goal(s): This study introduces a voxelwise approach to concurrently estimate the Orientation Distribution Function (ODF) and response function for fiber orientation analysis and tractography. Approach: The proposed blind deconvolution method models the kernel as a sum of axially-symmetric Gaussian functions, defined in spherical harmonics. It is evaluated through simulations and in-vivo experiments in healthy volunteers and glioma patients, demonstrating its efficacy in ODF estimation and data fitting. Results: This novel approach presents better modeling of pathology and offers promising results for white matter analysis. Impact: We introduce a blind deconvolution method for brain microstructure analysis with DWI that concurrently estimates a voxelwise ODF and kernel. This method can aid tractography and provide new image contrasts in the presence of pathology.