Left ventricular border recognition in echocardiographic images using modular neural networks and Sugeno integral measures

Abstract

The Echocardiography and the 2D ultrasound images are widely used to assess patients with heart diseases. The Observer (cardiologist) qualitatively deduces the heart morphology and left and right ventricular functions. In this paper we use the modular neural networks and Sugeno Measures to find patterns in echocardiogram images to recognize left ventricular borders of the heart and derive quantitative parameters. We studied 39 echocardiographic images that are used as an input to modular neural networks to find patterns and recognize the left ventricular border and also to a monolithic neural network to compare the results. We used the percentage of error recognition to evaluate the two neural networks, where modular neural networks offered better results with a 98% of recognition versus 80% recognition of monolithic Neural Network. Modular neural networks proved that they are an effective technique to recognize the left ventricular border of the heart.