Linear prediction-based reconstruction of electrocardiogram with premature ventricular contraction for heart rate variability analysis

Abstract

A method to detect the locations of premature ventricular contraction (PVC) and reconstruct electrocardiogram (ECG) is proposed in this work. The ECG reconstruction is based on the linear prediction method. For the sake of known ground truths, a set of PVC beats was synthetically inserted to ECG from the MIT-BIH normal sinus rhythm database. The experimental results show that the heart rate variability (HRV) measures derived from the reconstructed ECG's are very close to the ground truths in both time domain and frequency domain. The comparison to the methods commonly used by experts, i.e., removing the abnormal RR intervals or averaging RR intervals before calculating the HRV measures, also shows the better performance by the proposed method. © 2013 Springer Science+Business Media Dordrecht.