Segmentacion del Ventriculo izquierdo en imagenes Ecocardiograficas utizando Snakes, modelos Deformables y Frentes de propagacion
Résumé
This work presents some segmentation methods based on dynamical representation of contours applied to echocardiographic images of the left ventricle. The project was developed in for stages named: acquisition, segmentation, 3D reconstruction and VRML visualization. 1) Images coming from volume sequences of the left ventricle were acquired during an echocardiographic exam. 2) Three segmentation methods were developed on the basis of features of the image: a) Contour active models "Snakes", b) Models using a propagation front based on Sethian ideas, c) a third model that uses 3D geometry for the acquisition of a volumetric sequence. This global is capable of segmenting and capturing the global parameters of size, shape and orientation required by the medical practitioner. 3) The 3D reconstruction and visualization is performed taking advantage of the cylindrical geometry. In conclusion, the results obtained are consistent with the apparent movement observed in the left ventricle proving in this way that the superquadratic can be applied for segmenting and modeling the left ventricle at the same time. Medical practitioners will be able of seeing and detecting a pathology even better: for example, an infarct or ischemic episode based on the dynamical of the model developed.