Atrial fibrillation detection algorithms for very long term ECG monitoring

Abstract

In this paper, we describe two algorithms suitable for the detection of Atrial Fibrillation episodes in very long terms (weeks) ECG monitoring, were the need of on-board implementation requires the development of reliable but simple and easy-to-implement methods. The proposed algorithms are based on the extraction of simple geometric features from the histogram of RR prematurity and delta RR. On the MIT Atrial fibrillation database, the RR prematurity algorithm provides the following performances: episodes sensitivity (S) 91%, episode positive Predictivity (P+) 92%, duration S 93%, duration P+ 97%. For the delta-RR algorithm the results were: episodes S 92%, episode P+ 78%, duration S 89%, duration P+ 90%. © 2005 IEEE.