Automatic Detection of Atrial Fibrillation Using RR Interval from ECG Signals

Abstract

Atrial Fibrillation (AF) is a common arrhythmia, particularly in the elderly and those with heart disease. It can be characterized by an uncoordinated heart electrical activity leading to bad electricity propagation in the upper chambers of the heart. This anomaly can be observed on electrocardiogram (ECG) signals and different statistical methods or timefrequency domain analysis had been explored to distinguish Atrial Fibrillation from other kind of arrhythmias using RR intervals from ECG signals. Examples of such algorithms include the Root Mean Square of the Successive Differences (RMSSD), Sample Entropy and Fast Fourier transform. In this study, the MIT-BIH database such as the Normal Sinus Rhythm (NSR), AF and Arrhythmia ECGs were used. This paper applies a preprocessing algorithm based on Pan- Tompkins method to extract reliable QRS complex from ECG signals and thus accurate RR intervals. In this paper, the proposed automatic detection of AF using RR intervals extracted from ECG is based on a 3 steps approach. The first step uses the computation of the RMSSD on the RR intervals extracted from a 24 hours ECG recording to find whether an arrhythmia had occurred. The second step applies autocorrelation of the squared signal to precisely determine the start time and stop time of the arrhythmia episode within a detected arrhythmia window. The last step consists of computing the Shannon Entropy from the start to stop time extracted on the previous step to discern AF from other type of Arrhythmias. By using the developed algorithm, we were able to accurately detect AF using RR intervals extracted from 24 hours ECGs recording with up to 99.5% accuracy in time resolution.