Registration of dynamic renal MR images using neurobiological model of saliency

Abstract

In this paper we propose the use of a neurobiology-based saliency measure to improve the performance of a quantitative- qualitative measure of mutual information for rigid registration of 4D renal perfusion MR images. Our registration method assigns greater importance to more salient voxels by applying a soft thresholding function to normalized saliency values. The resulting saliency map is a better representation of what is truly visually salient than an entropy-based saliency map. Our tests on real patient datasets show that incorporating this saliency measure produces better registration results than traditional entropy-based approaches.