Atlas construction from high angular resolution diffusion imaging data represented by Gaussian Mixture fields

Abstract

Groupwise image registration is an essential part of atlas construction which is a very import and challenging task in medical image analysis. In this paper, we present a novel atlas construction technique using a groupwise registration of high angular resolution diffusion (MR) imaging datasets each of which is represented by a Gaussian Mixture field. To solve the registration problem, an L(2) distance is used to measure the similarity between two Gaussian Mixtures, which leads to an energy function whose gradient can be computed in closed form. A projection method is developed to construct a "sharp" (not blurred) atlas from the result of this groupwise registration. Synthetic and real data experiments are presented to demonstrate the efficacy of the proposed method.