Neural Network and Wavelets in Arrhythmia Classification

Abstract

Cardiovascular diseases are a substantial cause of death in the adult population. Changes in the normal rhythmicity of a human heart may result in different cardiac arrhythmias, which may be immediately fatal or cause irreparable damage to the heart, when sustained over long periods of time. In this paper two methods are proposed to efficiently and accurately classify normal sinus rhythm and different arrhythmias through a combination of wavelets and Artificial Neural Networks(ANN). MIT-BIH ECG database has been used for training of ANN. The ability of the wavelet transform to decompose signal at various resolutions allow accurate extraction/detection of features from non-stationary signals like ECG. In the first approach, a set of discrete wavelet transform (DWT) coefficients which contain the maximum information about the arrhythmia is selected from the wavelet decomposition. In the second approach, arrhythmia information is represented in terms of wavelet packet (WP) coefficients. In addition to the information about RR interval, QRS duration, amplitude of R-peak and a set of DWT/WP coefficients are selected from the wavelet decomposition. Multilayer feedforward ANNs employ error backpropagation (EBP) learning algorithm (with hyperbolic tangential activation function), were trained and tested using the extracted parameters. The overall accuracy of classification for 47 patient records in DWT approach (for 13 beats) is 98.02% and in WP approach (for 15 beats) is 99.06%.