Development and Validation of a Multimodal 3D Brain Image Registration Algorithm for Alzheimer’s disease and Related Dementias

Motivation: Multimodal registration of brain images enables comprehensive integration of data between imaging modalities, enhancing our understanding of brain structure, function(s), and disease coping mechanisms. Goal(s): This study develops and validates a pipeline for multimodal image registration of MRI and PET data from ADRD subjects to a common atlas space. Approach: This framework utilizes a multi-stage approach, combining rigid and deformable registration, to align input images to the MNI152+ atlas for comparisons of brain anatomy and function. Results: This study generated a 3D registration framework for multimodal images, achieving a Structural Similarity Index Measure (SSIM) score of 0.9143±0.0288(n=337) across subjects with different ADRD stages. Impact: This study presents a 3D multimodal registration framework invariant to image contrast, spatial resolutions, and anatomical and functional variations; thus, enabling direct comparisons of brain anatomy and function(s) across subjects, while minimizing inter-subject variability, permitting quantitative multimodal analysis.