Multi-structure whole brain registration and population average.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate registration of brain images is crucial for understanding brain structure and function.
• Existing magnetic resonance (MR) image registration methods may not optimally align critical subcortical and cortical regions.
• Incorporating anatomical information can potentially enhance registration accuracy.

Purpose of the Study:

• To develop and validate a novel whole brain MR image registration method.
• To explicitly improve the alignment of cortical and subcortical brain structures.
• To generate a high-quality population average brain image and cortical surface.

Main Methods:

• A novel whole brain MR image registration technique was developed.
• The method utilizes anatomic segmentation of multiple cortical and subcortical structures.
• Registration is performed using variational optimization within the large deformation diffeomorphic metric matching (LDDMM) framework.

Main Results:

• Quantitative evaluation on 10 manually labeled brains demonstrated improved subcortical and cortical alignment compared to traditional single-channel MR image registration.
• The new method successfully generated a crisp volumetric and cortical surface-based population average brain.
• The resulting average image facilitated reconstruction and labeling of the cortical surface.

Conclusions:

• The proposed MR image registration method enhances the alignment of key brain regions.
• This technique yields superior population average brain images, aiding in neuroanatomical studies.
• The method offers a valuable advancement for neuroimaging analysis and brain mapping.