Demons registration of high angular resolution diffusion images

In this work we present a method for non-rigid registration of high angular resolution diffusion imaging (HARDI) datasets that are modeled by a field of antipodally symmetric spherical functions, represented by their expansion in the real spherical harmonic (RSH) basis. We use a multichannel demons algorithm which utilizes a computationally simple, rotationally invariant similarity function defined in the RSH space. Additionally, we describe a finite strain based algorithm for reorientation of the HARDI data model. We validate our framework on simulated fiber orientation distribution datasets and on human in-vivo data.