Abstract
We present a hybrid approach which combines two classical segmentation algorithms that is implemented in image features instead of the original images and a spatial adaptive technique which is able to delineate the endocardial and epicardial boundaries of the left ventricle in echocardiographic images. Image feature is an approach that acts as a de-noising tool and smoothing operation while the salient boundaries are preserved. Two key techniques are developed: (i) a classical active contour method and a powerful segmentation technique based on level set method embedded in image features are employed for endocardium contour extraction; (ii) a region restricted technique is employed for epicardium contour extraction. Once the endocardium contours are highlighted, the epicardial boundaries are traced by using a contour expanding procedure coupled with a convex hull smoothing operation. In order to validate the effectiveness of the segmentation into image features, the results of both segmentation approaches were compared in term of their correctness. Thus, both the boundary displacement errors and the area error rates were taken into account. A set of 50 echocardiographic images was used. Experimental segmentation results are provided.