Artificial intelligence algorithm for heart disease diagnosis using Phonocardiogram signals

Abstract

An artificial intelligence system has been developed using Artificial Neural Networks (ANN) algorithms to diagnose heart disease from Phonocardiogram (PCG) signals. Four new featured characteristics of the signals, namely activity, complexity, mobility and the spectral peaks from the power spectral density plots are used as input to the neural network. Ninety-four PCG signals for three heart diseases were used in this study to test the accuracy of the neural networks. After the signals are filtered and the feature characteristics are extracted, the features are fed to the neural networks. Classification is carried using the Radial Basis Function (RBF) network and the Back Propagation Network (BPN) techniques. Receiver operating characteristic (ROC) is calculated to measure the accuracy for both structures. The results show that RBF provided 98% accuracy in predicting the disease compared with 90.8% for BPN. The developed artificial intelligence algorithm has been shown to be a powerful technique in automatic diagnosis of heart diseases using PCG signals.