Non-rigid coregistration of diffusion kurtosis data

Diffusion kurtosis imaging (DKI) is a relatively new model to study the non-Gaussian behavior of water diffusion in the brain white matter which introduces, besides the conventional diffusion tensor, a 4^th order, 3D diffusion kurtosis tensor to describe the diffusion. In this study, a multi-component coregistration algorithm using a viscous fluid model and mutual information is optimized to enable more accurate alignment of the higher order tensor DKI data. The preservation of principle strategy is extended in order to facilitate tensor reorientation of the diffusion and diffusion kurtosis tensors. In addition, experiments demonstrated that involving kurtosis information in the coregistration procedure significantly improves tensor alignment.