Abstract (EN)

We present a new approach to surface tracking applied to 3D medical data with a deformable model. It is based on a parametric model composed of a superquadric fit followed by a Free-Form Deformation (FFD), that gives a compact representation of a set of points in a 3D image. We present three different approaches to track surfaces in a sequence of 3D cardiac images. From the tracking, we infer quantitative parameters which are useful for the physician, like the ejection fraction, the variation of the heart wall thickness and of the volume during a cardiac cycle or the torsion component in the deformation of the ventricle. Experimental results are shown for automatic shape tracking and motion analysis of a time sequence of Nuclear Medicine images.