Time-evolution analysis of differential features on 3D surfaces of the heart walls

Differential feature computation such as curvature computation proves itself useful in 3D shape reconstruction, segmentation and is potentially interesting in matching and motion estimation. Several methods have been developed, generally dependent on the surface representation. The authors focus here on discrete voxel-based surface representations and they use the now classical algorithm proposed by Sander and Zucker (1990) for which the authors assess the accuracy. With the aim of analyzing and synthesizing the motion of 3D deformable models, the authors track the evolution of the curvatures through 2 different approaches: one based on global differential features calculation while the other one visualizes the amount of deformation locally. These methods are applied to synthetic data generated from parametric models, and to a sequence of 18 3D X-ray data of the left ventricle of the heart. The simulations try to assess the temporal stability and accuracy of the curvature measures of such deformable surfaces. The experiments on real data of the left ventricular walls exhibit highly deformed regions opposite to quasi-stationary regions that can be used advantageously in 3D motion estimation or in a matching process