Functional and anatomical medical image matching

Abstract

Robust solutions for correspondence matching of two data sets are prerequisite for many applications, particularly in the medical domain. In this paper, we examine the task of matching images from two Magnetic resonance imaging (MRI) acquisitions; cine dynamic sequence that allows studying the contractile function and Delayed Enhancement (DE) acquisition that allows myocardial viability assessment. Detecting the endocardial contour directly on DE images involves a major difficulty owing to the fact that blood and infarcted zone have almost the same grey level intensity.The purpose of this study was to present software to segment automatically DE images, using Cine sequence because of their high contrast between cavity and myocardium. Given the DE image as a source and the cine sequence as a target, three steps are performed in order to automatically bring the source image into register with the target image. A temporal registration followed by a spatial registration allow matching the two data sets taking into account all acquisition differences concerning time, position, orientation and resolution. An automated segmentation method is then applied to cine images in order to get an accurate endocardial border including papillary muscles into the left ventricle cavity. Resulting border is finally registered and superimposed on DE images.Very encouraging results were provided by the proposed method. However, the evaluation must be extended to a larger data base, to compare automatic segmentation on DE images against manual tracing by experts.