A Framework for Linear and Non-Linear Registration of Diffusion-Weighted MRIs Using Angular Interpolation.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Diffusion-weighted magnetic resonance images (DW-MRIs) registration is crucial for population studies and brain atlas construction.
• Current methods often rely on scalar images (FA, b0) or complex models (DTI, ODF), potentially losing directional information or facing computational challenges.

Purpose of the Study:

• To compare registration algorithms based on angular interpolation (AI), b0 images, and diffusion tensor imaging (DTI).
• To generalize AI for non-linear registration and evaluate its performance against existing state-of-the-art methods.

Main Methods:

• Developed and implemented a generalized angular interpolation (AI) algorithm for non-linear DW-MRI registration within FSL.
• Compared AI registration with b0 and DTI-based registration methods.

Main Results:

• Angular interpolation (AI) registration demonstrates superior accuracy compared to b0 and DTI-based methods in many cases.
• The generalized AI method effectively registers raw DW-MRI data, preserving directional information.

Conclusions:

• AI-based registration is a powerful and accurate alternative for DW-MRIs.
• This approach enhances registration accuracy and provides valuable raw data for subsequent neuroimaging analyses.