Extracting left ventricular contour by MVN_CNN, UBN_CNN and region based level set method

Abstract

In this paper, the detection of the boundaries and extraction of the area of left ventricular are proposed. The echocardiographic image is preprocessed to enhance the contrast and smoothness by utilizing Multi Valued Neural Cellular Neural Networks (MVN_CNN) non linear filter. UBN_CNN is applied to the smoothed image to detect the heart boundaries. A non threshold Boolean function with nine variables is utilized to detect the edges corresponding to the upward and downward brightness overleaps. A region based level set method is applied to extract the area of left ventricular. Some experimental results are given for different echocardiographic images. The detection and extraction combination of UBN_CNN and region based level set approach showed better results for extracting the LV boundaries.