PDE and tensor based approach for parietal dynamic tracking in ultrasound sequences

A method for pericardial and endocardial edge extraction of the left ventricle (LV) in 2-D short axis echocardiographic sequence is presented. The proposed algorithm consists in two steps: nonlinear anisotropic diffusion filtering and directional geodesic active contours. These step are based on both PDE and tensors. Tensor formalism gives access to structure orientation of the acoustic density field. This provides a more robust algorithm by using the whole image information. Sample experimental results are shown for myocardium edge detection in 2-D short-axis ultrasound sequences.