Segmentation and Analysis of 3D Cardiac Motion from Tagged MRI Images

Abstract

This paper gives an overview of our framework for the automated segmentation and motion analysis of cardiac motion from MRI tagging lines. It consists of a series of novel methods which utilize theory from image processing, deformable models and finite elements. Our framework consists of several steps. In the first step we use Gabor Filter banks and deformable models for the automatic segmentation of tagging lines and cardiac boundaries. The extracted tagging lines and boundaries are then used as input to a volumetric deformable model for the heart's motion estimation analysis. In this step we first extract parameters that can determine the difference between a normal and a pathologic heart motion. Second, using an Expectation-Maximization methodology (EM) we are able to determine a given heart's stress-strain relationship and fiber orientation. Our hypothesis is that the 3D shape and motion analysis of the heart will allow the faster and timely diagnosis of heart disease compared to traditional 2D methods. We present a series of segmentation, shape, motion and tissue property analysis results.