Comparison of time-frequency methods using spectral turbulence to identify late potentials in patients with ventricular conduction defects

The link between ventricular late potentials recorded in the signal-averaged electrocardiogram and sustained ventricular tachycardia (VT) has been well established. Patients with ventricular conduction defects are excluded in conventional time-domain analysis. This study aimed to detect the late potentials or abnormal intra-QRS potentials in patients with ventricular conduction defects using spectral turbulence analysis in the time-frequency domain. Patients with either (1) left bundle branch block, (2) right bundle branch block, and (3) intraventricular conduction defects were studied. The time-frequency representation was formed from the signal averaged XYZ leads using the short-time Fourier transform, the wavelet transform, and Choi-Williams distributions. Spectral turbulence analysis of these representations was comprised of four parameters, (1) interslice correlation mean, (2) interslice correlation standard deviation, (3) low-slice correlation ratio, and (4) spectral entropy. Separation of VT from non-VT patients was primarily observed in the X-lead using the short time FFT. Neither the wavelet transform nor the Choi-Williams distribution performed as well as the short time FFT.