Normalized gradient fields and mutual information for motion correction of DCE-MRI images

Abstract

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) of the kidney typically displays spatial motion and undesired artefacts due to unavoidable patient movement and physiological pulsations, with the effect of corrupting voxel-wise signal intensity changes arising from contrast agent wash-in and wash-out. Image registration is a necessary tool to counteract such motion artefacts and to estimate physiological parameters reliably. In this work, we present a fluid-registration-based method for deformable multimodal image registration based on normalized gradients, particularly well suited to handle the motion challenges in DCE-MRI time series. We evaluate and confirm that both normalized gradients and mutual information are high-performing cost functionals for co-registration of DCE-MRI time series. Further, there are indications that normalized gradients have better performance than mutual information on this kind of images. These results promote normalized gradients as a promising tool for proper motion correction of DCE-MRI images applied in the clinic or in biomedical research.