An EM Strategy for Ventricular Wall Motion Analysis

Quantitative assessment of the motion and deformation variability of the heart has important implications for the diagnosis and treatment of cardiac diseases. In this paper, we propose a novel multiframe framework for left ventricular deformation analysis based upon expectation maximization (EM) principles. Myocardial boundaries are extracted using modified levelset strategy, and sparse corresponding boundary points are derived based on matching geometrical similarities. The left ventricle is modeled as an isotropic linear elastic material, and EM algorithm is employed to track tissue deformation over the entire cardiac cycle. Initial experiments have been performed with 2D cardiac MR image sequences.