Deformation based representation of groupwise average and variability

Abstract

This paper presents a novel method for creating an unbiased and geometrically centered average from a group of images. The morphological variability of the group is modeled as a set of deformation fields which encode differences between the group average and individual members. We demonstrate the algorithm on a group of 27 MR images of mouse brains. The average image is highly resolved as a result of excellent groupwise registration. Local and global groupwise variability estimates are discussed. © Springer-Verlag Berlin Heidelberg 2004.