Effect of residual noise level on reproducibility of the signal-averaged ECG

Abstract

Late potentials are detected at various noise levels in clinical studies. The aim of this study was to assess the effect of residual noise level on the reproducibility of the signal-averaged electrocardiogram (ECG). Two consecutive raw 15-minute ECG from each of 188 patients with coronary artery disease were digitized and stored on optical discs. Each raw ECG was analyzed by two signal-averaging procedures to noise level 0.2 μV or 0.4 μV. Standard time-domain parameters were measured: QRS duration (SA-QRS), late potential (LP) duration, and root-mean-square voltage of the terminal 40 ms of the filtered QRS (RMS40). The SA-QRS was prolonged by 12 ± 14 ms by the reduction in noise level from 0.4 μV to 0.2 μV, LP duration was prolonged by 10 ± 10 ms, and RMS40 was reduced by 19 ± 22 μV. The temporal variation of the measured SA-QRS from ECG1 to ECG2 was significantly lower at noise level 0.2 μV (9 ± 13 ms) than at noise level 0.4 μV (13 ± 14 ms) ( P < .001). The LP duration was also more stable at noise level 0.2 μV than at noise level 0.4 μV (0.5 ± 11 ms vs 2 ± 13 ms, P < .05). The presence of any two of three abnormal parameters (SA-QRS > 120 ms, RMS40 < 25 μV, LP duration > 40 ms) was used as the criterion for the presence of LPs. At noise level 0.4 μV, the proportion of patients with diagnosed LPs in ECG1 was 25% and at noise level 0.2 μV it was 62%. At noise level 0.4 μV, 20% were reclassified from LP-negative in ECG1 to LP-positive in ECG2, and 7% were reclassified from LP-positive in ECG1 to LP-negative in ECG2. At noise level 0.2 μV, 20% were reclassified from LP-negative in ECG1 to LP-positive in ECG2, and 9% were reclassified from LP-positive in ECG1 to LP-negative in ECG2. It was concluded that (1) the diagnosis of LPs is significantly dependent on the extent of noise reduction by signal averaging; and (2) the numerical reproducibility of signal-averaged QRS duration and LP duration is lower at noise level 0.4 μV then at noise level 0.2 μV; and the diagnostic reproducibility of LPs is similar at both noise levels.