3D Segmentation of the Left Venticle in Echocardiographic Images using Deformable Model based on the Geometric Evolution Shapes
Résumé
This work presents a new segmentation method of volumetric echocardiographic images, using a 3D front propagation initialized as a small sphere that grows until arriving to the walls of the left ventricle (LV), adjusting the parameters to the model. The project is developed in three modules denominated: Acquisition, Segmentation and Visualization. 1) The images' acquisition of volumetric sequences of the LV was obtained in an echocardiographic test. 2) The segmentation was divided in three parts: a) A first segmentation method 20->30. It uses a front propagation method initialized with a circle that grows inside the acquired volume. B) A 3D front propagation initialized with a small sphere placed in the interior of the LV, which begins to grow occupying part of the ventricle cavity, c) The surface obtained previously is parameterized. The model is adjusted minimizing the energy of the error between the model and the segmentation carried out with the front propagation. The new idea of this work is the parameterization of the surface using an hierarchy of geometric shapes. With a few values of the model, one can capture the shape, size, position and torsion of the LV