Brain image registration based on cortical contour mapping

Abstract

The authors address the problem of brain image registration, and they present a new, nonlinear registration technique. In the first step of the authors' technique they obtain a point-to-point mapping between the outer cortical contours of the brain images using an elastic string algorithm. In the second step the authors register the two images based on the point-to-point correspondence established in the first step. They propose a new, nonlinear registration transformation, which is based on a spatially variable scaling and relation that can describe highly nonlinear deformations. Finally, the authors test their algorithm on two different registration problems: they first consider the registration of a postmortem photograph of a baboon brain cross-section and then an MR image of approximately the same cross-section.<>