Screening patients with paroxysmal atrial fibrillation (PAF) from non-PAF heart rhythm using HRV data analysis

The idea of this research is to determine can we tell from the HRV data without paroxysmal atrial fibrillation present at the recording if the patient suffers from this arrhythmia. The benefit is we can provide time and cost effective preliminary screening procedure during short time visit to the clinic. To achieve this goal we used Fourier analysis of the 30 minute HRV segment duration. We found statistically significant increase in the energy above 0.1 Hz for the patients with documented PAF history. This suggests that people with this arrhythmia has increased parasympathetic activity. For automatic classification of the patient we trained artificial neural networks on the HRV Fourier spectrum of AFPDB database. Testing on the AFDB (66.5 hours of HRV data from PAF patients) and NSRDB (352 hours of HRV data from healthy ones) databases produced Se 94.5% and Sp 96.5%.