Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) of the kidney provides important information for the diagnosis of renal dysfunction. To this end, a time series of image volumes is acquired after injection of a contrast agent. The interpretation and pharmacokinetic analysis of the time series data is highly sensitive to motion artifacts. Registration of these data is a challenging task as contrast uptake adds new image features and gives rise to intensity changes over time within the kidneys.
This paper presents a new registration pipeline for a time series of 3D DCE-MRI. The pipeline combines state-of-art modules such as a weighted and robust least squares type distance measure, a regularization that is based on hyperelasticity and thus ensures diffeomorphic transformations and enables the incorporation of local rigidity constraints on the kidneys. We provide results that indicate the necessity of these constraints and illustrate the superiority of the proposed pipeline as compared to other approaches.