Detection of Cardiac Arrhythmias using Electrocardiogram Signals

Abstract

ECG is an important non-invasive clinical tool for the diagnosis of heart diseases.The detection of cardiac arrhythmias is a challenging task since the small variations in ECG signals cannot be distinguished precisely by human eye. The objective of this work is to detect cardiac arrhythmias with highest detection accuracy. Cardiac arrhythmias are classified using discrete wavelet transform (DWT) and dual tree complex wavelet transform (DTCWT) technique. The DWT feature set comprises of statistical features extracted from the sub bands obtained after decomposition of QRS complex signals up to 5 scales whereas the DTCWT feature set comprises of wavelet coefficients extracted from the 4th and 5th scale decomposition of QRS complex signals. The two sets of features are appended individually by four other features (AC power, kurtosis, skewness and timing information) extracted from the QRS complex signal of each cardiac cycle. These feature sets are independently classified using a multi layered perceptron (MLP) neural network based on back propagation algorithm. In this work, five types of ECG beat (Normal (N), Paced (P), Right Bundle Branch Block (R), Left Bundle Branch Block (L) and Premature Ventricular Contraction (V)) are classified from the 48 files of MIT-BIH arrhythmia database. The experimental results indicate that the DTCWT technique classifies ECG beats with an overall sensitivity of 94.64% while DWT technique classifies with an overall sensitivity of 91.23%.