A new similarity measure for non-rigid breathing motion compensation of myocardial perfusion MRI

TLDR: This work proposes a measure based on Normalized Gradient Fields (NGF) and uses the B-spline based approach to describe the transformation of the image space to ensure smoothness and shows that by using NGF good registration results can be obtained for image exhibiting a high intensity contrast.

Abstract: Breathing movements during the image acquisition of first-pass gadolinium enhanced, myocardial perfusion Magnetic Resonance Imaging (MRI) hinder a direct automatic analysis of the blood flow of the myocardium. In addition, a qualitative readout by visual tracking is more difficult as well. Non-rigid registration can be used to compensate for these movements in the image series. Because of the local contrast and intensity change over time, the registration criterion needs to be chosen carefully. We propose a measure based on Normalized Gradient Fields (NGF) in order to obtain registration. Since this measure requires strong gradients in the images, we also test combining the measure with the Sum of Squared Differences (SSD) to maintain registration forces over the whole image area. To ensure smoothness, we employ a Laplacian regularizer and use the B-spline based approach to describe the transformation of the image space. Our experiments show that by using NGF good registration results can be obtained for image exhibiting a high intensity contrast. For images with a low intensity contrast, combining NGF and SSD improves the registration results significantly over using NGF only. Both measures are differentiable making possible the application of fast, gradient based optimizers.