Fusion of a priori model for the 3D segmentation of the left ventricule in echocardiographics images

This paper proposes a new segmentation method for the echocardiogaphic images based on data fusion theory. This proposal consists on a data fusion system with two inputs: The "A" input corresponds to the gradient image after having made the pre-processing links to the acquisition, filtering and interpolation. The "B" input corresponds a parametric equation of a priori model whose equation can be an ellipsoid or a superquadrics. These two inputs are fusion together by using the Bayesian methodology of probabilities with the purpose to obtain a new function "F". That will be used to stop the Level Set Method to segment the inside of the internal wall of the left ventricle (LV). Finally the 3D visualization of the LV volume is obtained using the method of marching cubes, and the 3D global model is adjusted to the previous reconstruction. The analysis of the mean square error (MSE) between the calculated model and the real LV volume is smaller than 5%, which confirms the precision of computing of the global medical parameters. In conclusion, the segmentation method based on data fusion allows to obtain simultaneously the global model and the local model of deformation.