Population Shape Collapse in Large Deformation Registration of MR Brain Images

This paper examines the shape collapse problem that occurs when registering a pair of images or a population of images of the brain to a reference (target) image coordinate system using diffeomorphic image registration. Shape collapse occurs when a foreground or background structure in an image with non-zero volume is transformed into a set of zero or near zero volume as measured on a discrete voxel lattice in the target image coordinate system. Shape collapse may occur during image registration when the moving image has a structure that is either missing or does not sufficiently overlap the corresponding structure in the target image[4]. Such a problem is common in image registration algorithms with large degrees of freedom such as many diffeomorphic image registration algorithms. Shape collapse is a concern when mapping functional data. For example, loss of signal may occur when mapping functional data such as fMRI, PET, SPECT using a transformation with a shape collapse if the functional signal occurs at the collapse region. This paper proposes an novel shape collapse measurement algorithm to detect the regions of shape collapse after image registration in pairwise registration. We further compute the shape collapse for a population of pairwise transformations such as occurs when registering many images to a common atlas coordinate system. Experiments are presented using the SyN diffeomorphic image registration algorithm. We demonstrate how changing the input parameters to the SyN registration algorithm can mitigate some of the collapse image registration artifacts.