Importance of the endpoint of noise reduction in analysis of the signal-averaged electrocardiogram.

Reduction of random noise by signal averaging is required to uncover ventricular late potentials (LPs). Noise reduction is dependent upon the ambient noise before the study and the number of signal-averaged QRS complexes. Prior studies have used a fixed number of QRS complexes (e.g., 150 to 200) for performing signal-averaged electrocardiograms (SAECGs). Because of variable background noise levels, it was hypothesized that variable noise levels after processing could interfere with detection of LP. Accordingly, SAECGs were performed for each patient to 2 prespecified noise endpoints (expressed as the standard deviation/square root of number of beats): 1.0 microV, which has been used previously as a minimal residual noise level, and 0.3 microV, a low level that generally can be attained in less than 450 beats. Root mean square-voltage noises in the 40-Hz high pass filtered vector magnitude for these studies were 1.36 +/- 0.57 and 0.58 +/- 0.28 microV, respectively. The relative prevalence of LP was evaluated in 3 groups. Group I was comprised of 26 patients with sustained ventricular tachyarrhythmias, group II included 59 patients after myocardial infarction and group III had 14 normal volunteers. The prevalence of LP was greater in group I (69 vs 46%, p less than 0.001) and group II (34 vs 24%, p less than 0.01) with the 0.3-microV studies. In group III, the prevalence did not change (7 vs 7%, difference not significant). The greater detection of LP was due to improved resolution of the terminal low-amplitude QRS segment. Therefore, using the 0.3 microV level instead of 1.0 microV increased sensitivity of LP without loss of specificity.(ABSTRACT TRUNCATED AT 250 WORDS)