ABSTRACT
In this chapter, we introduce novel methods for cardiac image segmentation, namely left ventricle in computed tomography and in fetal ultrasound. We explore several segmentation schemes, each adapted to a specific problem, for example, endocardium, epicardium, and image modality. All the segmentation methods presented share two common principles: first, they are all based on deformable models, such as Active Shape Models, Active Appearance Models, or Level Sets, and second, they all have been modified from their original formulation in order to consider a local image descriptor that better guides the shape fitting to the target object. This image descriptor is the Hermite transform, which mimics some of the more important properties of vision perception, such as the Gaussian derivative of early vision and local analysis.
