Surface modeling in heart motion analysis

This paper presents a new approach for left ventricle surface modeling in the analysis of dynamic behavior of the heart from biplane cineangiograms. We decompose the motion and deformation analysis of the left ventricle into two stages by coarse-to-fine modeling the moving surface of the left ventricle. Such a two-step surface modeling enables us to formulate the complex motion analysis as a series of well-defined parameter estimation algorithms. We model the globally deformable surface by a parametrized family of surfaces known as superquadrics that are able to model expansion/contraction and twisting deformations. The deformation parameters are obtained by fitting the given 3-D data to the superquadric modeling primitives. The residues of such a fitting are the measure of local deformations that global modeling primitives are unable to abstract. These residues are then interpolated by spherical harmonic surface modeling primitives to form a residue surface. Local deformation tensor analysis of the left ventricle is therefore based on the overall surfaces constructed by superposition of local spherical harmonic residue surfaces on top of the global superquadric surfaces. Animations of the estimated dynamic surface are generated using scientific visualization techniques in order to vividly examine the complex spatial and temporal nature of the left ventricle. These animations are consistent with a priori knowledge of the left ventricle and hence show the success of the surface modeling algorithm.