FOD-based registration for susceptibility distortion correction in brainstem connectome imaging

Area of Science:

• Neuroimaging
• Diffusion MRI
• Human Connectome Project

Background:

• High-resolution diffusion MRI (dMRI) from the Human Connectome Project (HCP) enables detailed mapping of brainstem fiber pathways.
• Susceptibility-induced distortions in dMRI data, particularly in the brainstem, present a major challenge for accurate tractography.
• Current HCP Pipeline tools, like topup, offer partial correction but leave significant residual distortions due to reliance on B0 images with limited white matter contrast.

Purpose of the Study:

• To develop and validate a novel method for estimating and correcting residual distortions in HCP dMRI data.
• To leverage the full information from multi-shell dMRI with opposite phase encodings (PEs) for improved distortion correction.
• To enhance the accuracy of brainstem connectome mapping by addressing persistent distortion artifacts.

Main Methods:

• Computed fiber orientation distributions (FODs) from dMRI data acquired with opposite PEs.
• Developed a novel FOD-based registration method to estimate and correct residual distortions.
• Validated the method using dMRI data from 94 Human Connectome Project subjects.

Main Results:

• The proposed FOD-based registration method significantly reduces misalignment of primary fiber directions in the brainstem.
• Quantitative analysis showed a 21% reduction in main fiber direction misalignment compared to the HCP Pipeline's topup tool.
• The method effectively utilizes rich dMRI information from opposite PEs for enhanced distortion correction.

Conclusions:

• FOD-based registration offers a superior approach to correcting residual diffusion MRI distortions in the human brainstem.
• This novel method is compatible with the existing HCP Pipeline, facilitating its integration into connectome imaging research.
• The improved distortion correction enhances the reliability and precision of mapping fine-grained neural pathways.