by Image Inversion

This paper presents left ventricle (LV) endocardial segmentation from contrast 3D echocardiography (C3DE) images. The C3DE image segmentation is a very challenging problem. Though the image quality is perceived to be improved for visual analysis, the image quality actually deteriorates for the purpose of automatic/semi-a utomatic analysis due to high speckle noise. To overcome the speckle noise and low contrast between myocardium and LV cavity, we employ a simple inverse image construction method to ease the segmentation challenge. This reduces the failure rate of segmentation which was otherwise very high on C3DE. The image inversion method inverts the C3DE image appearance, thus having high contrast and low noise appearance while the endocardial delineation becomes clearer. This inverted image information is utilized in a semi-automatic image-driven LV segmentation method. The segmentation results of LV boundary delineation are compared to a manual LV boundary delineation and a good match is observed.