An automated three-dimensional plus time registration framework for dynamic MR renography

Dynamic contrast-enhanced 3D images of the kidneys, or 3D MR renography, has the potential for broad clinical applications, but suffers from respiratory motion that limits analysis and interpretation. Manual registration is prohibitively labor-intensive. In this paper, a fully automated technique, Wavelet Representation and the Fourier Transform (WRFT) method, that corrects for translation and rotation motion in 3D MR renography is presented. The method was composed by anisotropic denoising, wavelet-based feature extraction, and Fourier-based registration. This was first evaluated on a set of simulated MR renography images with defined degrees of kidney motion. The method was then tested on 24 clinical patient MR renography data sets. Results of clinical testing were compared with the results obtained using a mutual information registration method. Based on intrarenal time-intensity curves, our method showed robust and consistent agreement with the results of manually coregistered data sets.