A fast and robust time-series based decision rule for identification of atrial fibrillation arrhythmic patterns in the ECG

Atrial fibrillation (AF) is an arrhythmic behaviour of the heart, which occurs when the myocardium of the atrial chambers enter into a sustained chaotic and fractionated muscular contraction dynamic. Reliable detection of AF episodes in ECG monitoring devices, is important for early treatment and health risks reduction. A decision rule for identifying AF arrhythmic patterns was derived from RR-intervals analysis of time-series generated from ECG recordings before, during and after AF episodes. Time-series elements were obtained by consecutive RR intervals time differences (zIRR). In the proposed decision rule, two arguments must be satisfied for identifying an AF pattern within a window of 35 beats: (1) the number of zIRR elements above 50 ms absolute value, is >;10, and (2) there is a uniform dispersion of all the corresponding RR-interval elements within the same 35 beat window. Detection of AF using the proposed decision rule scheme was achieved with 96% exactitude, 93% sensitivity and 97% specificity. The longest case of processing time per ECG beat was of 129ms. This computing time requirement can enable real-time ECG processing algorithms for AF identification.