A statistical level set framework for segmentation of left ventricle

Abstract

A novel statistical framework for segmentation of the echocardiographic images is presented. The framework begins with presegmentation at a low resolution image and passes the result to the high resolution image for a fast optimal segmentation. We applied Rayleigh distribution to analyze the echocardiographic image, and introduced a posterior probability-based level set model. The model is applied for the pre-segmentation. The pre-segmentation result at the low resolution is used to initialize the front for the high resolution image with a fast scheme. At the high resolution, an efficient statistical active contour model is used to make the curve smoother and drives it closer to the real boundary. Segmentation results show that the statistical framework can extract the boundary accurately and automatically. © Springer-Verlag Berlin Heidelberg 2006.