Aligning 2-D Cardiac MR Images in 3-D Space using Normalized Mutual Information

Cardiac Magnetic Resonance images offer the opportunity to study the heart in detail. One of the main issues in its modelling is to create an accurate 3-D reconstruction of the left ventri- cle from 2-D views. A first step to achieve this goal is the correct registration among the different image planes due to patient movements. In this article, we present an accurate method to correct displacement artifacts using the Normalized Mu- tual Information. Here, the image views are treated as planes in order to diminish the approximation error caused by the association of a certain thick- ness, and moved simultaneously to avoid any kind of bias in the alignment process. This method has been validated using real and syntectic plane dis- placements, yielding promising results.