A new method for automatic segmentation of endocardial boundaries in echocardiographic sequences

This study presents a geometric model for automatic segmentation of left ventricular (LV) boundaries throughout the cardiac cycle in echocardiographic images. The segmentation is obtained by solving a partial differential equation that describes a motion under geodesic curvature. The initial condition for the partial differential equation is computed by applying a maximum likelihood segmentation of the endocardium based on the density probability distribution of the gray levels of the image as well as smoothness constraints. To model gray level behavior of ultrasound images the classic Rayleigh probability distribution is considered. The initial condition is automatically computed for each image throughout the cardiac cycle and the dynamic segmentation of the LV boundary is automatically achieved. Preliminary experiments were conducted with satisfactory segmentation outcomes in comparison with expert manual tracing. The proposed method is fast, no operator-dependent and allows easy noninvasive serial segmentation of endocardial boundaries throughout the cardiac cycle.