Application of Neural Networks in Diagnosis of Valve Physiological Heart Disease from Heart Sounds

Abstract

Classification of the sound heart into different valve-physiological heart disease categories is a complex pattern recognition task. In this paper application of various types of neural networks are introduced for diagnosing heart disease). At first a method is described for extracting useful features from the sound hearts and then a simple algorithm is introduced for heart sounds recognition. In fact, feature vectors are formed based on a wavelet decomposition of the sounds. The heart sound diseases are classified into normal heart sound and the other six valve physiological heart categories. Different types of artificial neural networks (ANNs) are used for this purpose. Those are Multilayer perceptron (MLP) with back propagation training algorithm, Elman Neural Network (ENN) and Radial Basis Function (RBF) Network. Expensive experimental results show an average recognition score of 81.25% to 96.42%.