Multistructure Diffeomorphic Brain Registration

Abstract

A new method to constrain brain MRI registration, and perform experiments evaluating the alignment of manually traced structures, reduction of intersubject variance, and morphmetric results using group wise registration is compared against traditional MRI. Structure-specific information is be used to initialize and constrain registration to improve accuracy and robustness. By using image registration, it restrict the scope of structures to those that can be represented volumetrically as images, such as MRI volumes or anatomical segmentations. This method uses Mean Shift algorithm to efficiently segment the gray and white matter and thus the average gray matter was taken to find the dice overlap. It uses concurrent subcortical and cortical shape matching to guide the overall registration