A large deformation diffeomorphic metric mapping solution for diffusion spectrum imaging datasets.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Spatial transformation is crucial for group analysis of diffusion spectrum imaging (DSI) data.
• Existing methods face challenges with conventional reorientation problems in diffusion-weighted datasets.

Purpose of the Study:

• To develop a novel spatial transformation method for DSI datasets.
• To generalize large deformation diffeomorphic metric mapping (LDDMM) to the 6D space of DSI data, termed LDDMM-DSI.
• To overcome conventional reorientation issues by allowing free deformation in q-space.

Main Methods:

• Developed LDDMM-DSI, a 6D extension of the LDDMM framework tailored for DSI datasets.
• Treated data-matching as a variational problem to find optimal velocity fields.
• Ensured generated transformations are diffeomorphic and geodesic.

Main Results:

• LDDMM-DSI successfully registered real brain DSI datasets.
• The method achieved accurate alignment of both global structural shapes and local diffusion profiles.
• Demonstrated the capability to handle the 6D nature of DSI data effectively.

Conclusions:

• LDDMM-DSI facilitates robust group analyses of DSI datasets.
• The method aids in the generation of standardized DSI templates.
• This approach enhances the analysis and comparability of diffusion spectrum imaging data.