Atrial Fibrillation Detection Based on Poincaré plot of RR Intervals

Atrial fibrillation (AF) is one of the most common types of arrhythmia which significantly increases the risk factor of stroke – especially in elderly population. In this paper an algorithm is presented which is suitable for the effective detection of AF, by using merely heart rate data as input. The method is based on the processing of Poincaré plots constructed from the set of 30 RR intervals. During the analysis of Poincaré plots the dispersion of points around the diagonal line is calculated and the number of clusters is determined by a self-developed cluster analyzer. The decision criterion of AF relies on these two parameters. On the one hand, the algorithm was tested on 10 AF and 10 normal rhythm ECG signals of the PhysioNet Database, achieving the average sensitivity (Se) of 98.69% and the average specificity (Sp) of 99.59%. On the other hand, 10 AF and 10 normal clinically confirmed records of a heart rate meter were also processed, resulting the average Se and Sp of 96.89% and 99.00%, respectively.