Atrial Fibrillation Detection Based on Poincaré Plot and P Wave Analysis

Abstract

Atrial fibrillation (AF) is a common supraventricular cardiac arrhythmia characterized by irregular RR intervals and the absence of P waves on the ECG. Since it may significantly increase the risk of stroke, the early diagnosis is crucial which can be supported by today's quickly developing mobile ECG technology. This paper describes an efficient AF detection method which is suitable for the processing of single-channel ECG signals. The algorithm makes the decision about AF based on the Poincaré plot consisting of RR intervals and the presence of the P wave on the average majority cycle. The Poincaré plot analysis was tested earlier on reference ECG signals from the PhysioNet database resulting in the average sensitivity (Se) of 98.69% and average specificity (Sp) of 99.59%. After improving our method by adding P wave analysis, it was built into the WIWE mobile ECG system produced by Sanatmetal Ltd. During clinical validation 29 AF and 315 non-AF WIWE records were processed (mainly to test Sp), 57 of which were non-AF arrhythmia cases. No false decision occurred regarding the clinical measurements in terms of AF.