A Noise Resistive Transtexture Based Delineation of Left Ventricle in Echocardiographic Images

Echocardiography is the predominant technique for quantitative assessment of cardiac function. This paper presents a new noise resistive method for segmenting echocardiographic images using fuzzy logic and texture properties. In echocardiographic images, the transition between blood pool and myocardium is significant, which reflects as variations in texture properties. Our edge detection method incorporates the transition characteristics and texture properties of echo images (transtexture) for boundary tracking. The texture properties such as energy and homogeneity are given as input for the fuzzy system, and output is the segmented image. The proposed algorithm is tested for about 200 images, with added speckle noise and compared with other edge detection operators. The results show that, integration of texture properties with fuzzy logic makes the delineation robust and reliable up to 20% of speckle noise.