Wavelet Analysis of Signal‐Averaged Electrocardiograms

Background: Wavelet representation is able to detect low amplitude patterns even if hidden within signals of much higher amplitudes. Method: A software system has been developed that implements wavelet representation of signal‐averaged electrocardiograms (SAECG). In this system, wavelet analysis leads to 4 numerical parameters that characterize the content of low amplitude perturbations found within the high gain QRS complex. In three substudies, these numerical parameters were compared with the standard time‐domain indices of SAECG. Populations: Normal limits were identified from recordings of 104 normal healthy volunteers (54 males, mean age 50 ± 17 years). Short‐term reproducibility of the numerical indices and of abnormal findings was evaluated in a population of 85 subjects (16 healthy volunteers, 22 patients with documented ventricular tachycardia [VT] without structural heart disease, 30 patients with documented sustained postinfarction VT, and 17 survivors of acute myocardial infarction) who were each recorded three times with 5‐minute periods separating individual recordings. The power of wavelet and time‐domain analyses in distinguishing patients with and without sustained VT after myocardial infarction was assessed using recordings of 53 patients with postinfarction VT and of 53 age, sex, and infarct site matched patients without a history of arrhythmic complications after infarction. Results: The studies have shown that (a) the indices of wavelet analysis are more reproducible than the time‐domain indices, (b) the distinction between patients with and without VT after myocardial infarction is similarly powerful by wavelet and time‐domain analyses, and the association of the positive SAECG analysis with postinfarction VT is highly significant with both analyses (P = 3.94 × 10–14 for wavelet analysis and 2.55 × 10 −9 for time‐domain analysis), the indices of wavelet analysis differ significantly between normals and patients with an uncomplicated history of myocardial infarction (P = 0.02–0.005), while time‐domain indices do not (all parameters NS), (d) in contrast to the time‐domain analysis, wavelet analysis was similarly powerful in identifying VT patients with anterior and inferior infarction (P = 1.4 × 10 −9 , n = 30, and P = 2.0 × 10 −15 , n = 23, respectively). Conclusion: Wavelet analysis is a highly reproducible method for SAECG processing which (a) is as powerful as the time‐domain analysis for the identification of ischemic VT patients, (b) compared to the time‐domain analysis, is not dependent on infarct site, and is able to distinguish postmyocardial infarction patients without VT from healthy subjects. A.N.E. 2000,5(1):4–19