Effect of noise on T-wave alternans measurement in ambulatory ECGs using modified moving average versus spectral method

Abstract

Background: The modified moving average (MMA) and spectral method (SM) are commonly used to measure T-wave alternans (TWA), but their accuracy has not been compared in ambulatory electrocardiograms (ECGs) where TWA signal-to-noise ratio is low. Our objective was to compare the effect of noise and signal nonstationarity on the accuracy of TWA measurement using MMA versus SM when applied to synthetic and ambulatory ECGs.

Methods: Periodic and nonperiodic noise were added to noiseless synthetic ECGs. Simulated TWA (0-20 microV) was added to synthetic ECGs and ambulatory ECG recordings. TWA was measured using SM and MMA, and the measurement error relative to added TWA was compared. An MMA ratio was used to discriminate TWA signal from noise. Signal nonstationarity was simulated by changing heart rate, TWA magnitude, and TWA phase.

Results: With no added TWA, MMA falsely measured TWA in synthetic and ambulatory ECGs, while false measurement was not seen with SM. An MMA ratio > 1.2 eliminated false TWA detection. In the presence of low TWA magnitude (<10 microV), TWA was overestimated by MMA and underestimated by SM in proportion to the noise level. In synthetic ECGs with periodic noise and 10-microV added TWA, MMA was less accurate than SM. The effects of simulated signal nonstationarity on the TWA magnitude measured with MMA versus SM were similar using a 64-beat analysis window.

Conclusions: In the presence of noise, MMA falsely detects or overestimates simulated TWA in ambulatory ECG recordings. In this setting, the proposed MMA ratio improves the specificity of MMA.