Screening and prediction of paroxysmal atrial fibrillation by analysis of heart rate variability parameters

This study has been performed within the scope of the CinC-2001 challenge on the detection and prediction of paroxysmal atrial fibrillation (PAF) from 200 paired two-channel ECGs of 30 minutes duration. Different features of heart rate variability (HRV) describing the magnitude as well as the regularity of heart rate fluctuations and the number of supraventricular premature beats (SVPCs) and ventricular premature beats (VPCs) were investigated for their suitability with respect to the classification task using ROC analysis, classification by ranks and linear polynomial classifiers with jackknife validation. Moreover, the time courses of mean parameter values were calculated to identify possible trends. Although promising results of up to more than 80% accuracy in screening and 92% in prediction were achieved on training data, these were not reproducible on an independent test set.