Improved event interval reconstruction in synthetic electrocardiograms.

Abstract

Mathematical models for synthesizing ECGs are important tools in providing reproducible standard signals for biomedical signal processing research and technology. The ECG synthesis model ECGsyn and its extensions are the state of the art and realistically capture: 1) the ECG morphology; 2) the spectrum of heart rate variability; and 3) QT-interval adaptation to heart rate. This paper demonstrates ECGsyn's time-domain limitations at reconstructing an ECG time series from ECG wave annotations. An alternative algorithm, ECGfm, is therefore presented that recreates the event intervals with greater fidelity by applying the wave annotations as a phase constraint. Results on data drawn from the Physionet Physobank QT database demonstrate the improved performance of the ECGfm algorithm.