Rigid registration of Delayed-Enhancement and Cine Cardiac MR images using 3D Normalized Mutual Information

For the viability assessment in patients with infarcts, myocardial hyperenhancement on Delayed-Enhancement Cardiac MRI (DE) can be anatomically localized using Cine Cardiac MRI (Cine). An automatic rigid registration method, using the Normalized Mutual Information (NMI) maximization was proposed to refine the registration of the functional (DE) and anatomical (Cine) images. The full process including: 1) a coarse DE/Cine spatiotemporal alignment, 2) a refined DE/Cine registration, 3) a GVF-Snakes segmentation of myocardial contours on Cine images and 4) a myocardial infarction extent (MIE) quantification using a fuzzy c-means algorithm, was applied to ten patients with myocardial infarction. The qualitative and quantitative evaluation of the registration method showed high-quality alignment. The comparison between visual and automatic MIE quantification agreement (±1 grade) showed an improvement from 86% before to 91% after the refined registration step.