Diffusion MRI registration using orientation distribution functions.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Diffusion-weighted MRI (DW-MRI) is crucial for mapping white matter architecture.
• Accurate registration of DW-MRI datasets is challenging due to complex fiber orientations and intravoxel crossing.
• Existing registration methods may not fully preserve the integrity of diffusion orientation distribution functions (ODFs).

Purpose of the Study:

• To develop and validate a linear-elastic registration method for DW-MRI data based on mapping ODFs.
• To improve the accuracy of DW-MRI registration by incorporating ODF reorientation and elastic constraints.
• To enhance the consistency of principal diffusion directions in registered ODF maps.

Main Methods:

• Reconstruction of ODFs using q-ball imaging (QBI) to handle intravoxel fiber crossings.
• Registration based on mapping ODF maps represented by spherical harmonics for computational efficiency.
• Application of ODF reorientation matrices derived from the local Jacobian to spherical harmonic coefficients.
• Definition of registration similarity cost using ODF shape distance.
• Regularization of transformation fields using linear elastic constraints.

Main Results:

• The proposed linear-elastic registration method effectively improved upon affine alignment.
• The method significantly reduced ODF shape differences between registered datasets.
• Incorporating reorientation during registration led to more consistent principal diffusion directions compared to methods without reorientation.
• Validation using both synthetic and real DW-MRI data confirmed the method's efficacy.

Conclusions:

• Linear-elastic registration of ODF maps offers a robust approach for DW-MRI data.
• ODF reorientation is essential for maintaining directional consistency during registration.
• The method provides improved accuracy and preservation of white matter architecture information in registered DW-MRI datasets.