Automatic detection of microvolt T-wave alternans in Holter recordings: Effect of baseline wandering

Abstract

ECG deviations from the isoelectric line (baseline wandering) may affect a reliable detection of T-wave alternans (TWA), a phenomenon associated with increased risk of death. The present study was designed to demonstrate, making use of simulated ECG time series, that baseline wandering might cause erroneous TWA detection from TWA-free ECG tracings, or prevent correct TWA detection whenever TWA was present. Our simulated time series were obtained from a real ECG complex (0.75 s recording, sampled at 200 Hz) repeated 170 times. TWA was simulated by varying T-wave amplitude (10, 50, 100 and 500 μV) in a time window of 160 ms centered around the T-wave apex. TWA fundamental frequency was 0.67 Hz. Baseline fluctuations were simulated by sinusoidal waves of 0.1 mV amplitude and frequency of 0.30, 0.67 and 1.50 Hz, respectively. The presence of baseline oscillations at lower (0.30 Hz) and higher (1.50 Hz) frequency than TWA own frequency prevented TWA detection when TWA amplitude was lower or equal to that of baseline oscillations. TWA detection improved after removal of baseline oscillations by application of a third-order spline interpolation, only for frequencies lower than TWA frequency. For baseline oscillations at greater frequency than heart rate, however, spline interpolation became harmful. An improvement over the spline interpolation was obtained by a new heart-rate adapting match filter with a narrow bandwidth around TWA frequency, which allowed detection of TWA almost independently of baseline frequency components when these were different from TWA own frequency.