Large deformation diffeomorphic registration of diffusion-weighted images with explicit orientation optimization

Area of Science:

• Medical Imaging
• Computational Anatomy
• Neuroimaging

Background:

• Diffusion-weighted imaging (DWI) is crucial for understanding brain microstructure.
• Accurate registration of DWI is challenging due to large deformations and complex diffusion patterns.
• Existing registration methods may limit subsequent diffusion model analysis.

Purpose of the Study:

• To develop a robust diffeomorphic registration framework for large deformation DWI.
• To enable flexible fitting of diffusion models post-registration.
• To improve the accuracy of anatomical and microstructural alignment in DWI.

Main Methods:

• Formulated a large deformation diffeomorphic registration framework from an optimal control perspective.
• Incorporated structural alignment, local fiber reorientation, and deformation regularization.
• Utilized a multi-kernel strategy for multi-scale anatomical structure registration.

Main Results:

• Demonstrated the efficacy of the proposed method using in vivo data.
• Achieved accurate registration under significant anatomical variations.
• Showcased improved alignment compared to existing registration strategies.

Conclusions:

• The developed framework provides accurate and robust registration of diffusion-weighted images.
• This method facilitates advanced, multifaceted analysis of diffusion models.
• The approach offers a significant advancement in neuroimaging registration techniques.