Frequency versus time domain analysis of signal-averaged electrocardiograms. I. Reproducibility of the results

Abstract

Three repeated signal-averaged electrocardiographic (ECG) recordings were made in 40 subjects (15 healthy volunteers. 10 patients with ventricular tachycardia without apparent heart disease and 15 patients with ventricular tachycardia after myocardial infarction). In each subject, the three recordings were made within 25 min. The recordings were subsequently analyzed by 1) the conventional time domain method of signal-averaged ECG analysis with use of filter settings of 25 to 250 Hz and 40 to 250 Hz, 2) a spectral analysis method computing the energy area within the spectral boundaries of 40 to 140 Hz and the ratio between the energies of areas within the boundaries 40 to 140 Hz and 0 to 40 Hz, and 3) spectral temporal mapping computing the “normality factor.”The study compared the reproducibility of these three approaches to the analysis of signal-averaged ECGs. First, the reproducibility of the diagnostic conclusions (that is, of the diagnosis of late potentials) was compared for the time domain method and for spectral temporal mapping. Second, the reproducibility of the numeric values of individual indexes provided by different methods was compared for all methods by computing the ratios between standard deviations of measurements in individual patients and standard deviation of all measurements.The reproducibility of diagnostic conclusions was significantly higher for the time domain method than for spectral temporal mapping (p < 0.05, sign test). The numeric reproducibility of the normality factors produced by spectral temporal mapping was significantly lower than the numeric reproducibility of the values of all indexes provided by time domain, spectral area and spectral area ratio methods (p < 0.05 to 0.00005, Wilcoxon tests). Spectral temporal mapping was the least reproducible method for the analysis of signal-averaged ECGs.