Research on Algorithm for Detecting Atrial Fibrillation Using RR Intervals

Abstract

Atrial fibrillation is a kind of arrhythmias that is common but fatal. It causes many kinds of mortality such as stroke, transient ischemia attack and heart failure. In addition, as the age increases, the incidence rate of AF increases substantially. So it is very important to detect AF as early as possible. In this paper, an algorithm for detecting AF is proposed. The algorithm uses the feature of RR intervals and mainly includes two processes. Firstly, we pre-process the raw RR intervals and implement symbolic dynamic transformation and multiscale Shannon entropy to measure the disorder of the RR intervals to get a preliminary detection result. Then the delta RR interval distribution difference curve is used to modify the boundary between AF and normal beats to get a more stable result. Experiments show that in similar algorithms, our method achieves the best performance (sensitivity of 98.81% and specificity of 96.53% respectively). This method is suitable for ECG comprehensive analysis (detecting AF after collecting long ECG signals) in clinical use, which can help improve doctors' work efficiency. At the end of the article, we also discussed where the algorithm can be improved.