Frequency analysis of the signal-averaged ECG of postinfarction patients for prediction of cardiac death

Frequency domain analyses based on the Fast Fourier Transform (FFT), Wavelet decomposition (WD) and Power law scaling (PLS) were performed on the signal-averaged ECG (SAECG) of 2461 postinfarction patients for the purpose of predicting cardiac death (CD) and compared to time domain (TD) analysis. During follow-up (2 years), 158 patients had CD. FFT was used to estimate the power area ratio from (20 to 50 Hz)/(0 to 20 Hz) and from (40 to 100 Hz)/(0 to 40 Hz). The relative energy (RE) of the SAECG was calculated in seven frequency bands ranging from 0 to 250 Hz using WD. PLS slope was estimated from the spectral log-log plot. Multivariate analysis (Cox) showed QRS duration as the best time domain parameter (risk=3.2, p<0.0005) with RE below 3.9 Hz (risk=2.2, p<0.0006) and RE from 7.8 to 15.6 Hz (risk=2.3, p<0.0006) as independent and significant predictors of CD. FFT and PLS parameters were excluded from the model. WD parameters of the SAECG used for predicting of CD after MI allowed a significant discrimination between CD patients and survivors. The statistical results obtained with the WD parameters were comparable with those obtained by TD parameters