Bispectral analysis for the detection of ventricular late potentials

The surface detection of delayed fragmented waveforms-ventricular late potentials-continuous with the end of the QRS complex during sinus rhythm in post-myocardial infarction subjects has been considered as a noninvasive risk indicator for the possible development of sustained ventricular tachycardia. In this paper, bispectral analysis (using nonparametric techniques) of synthesized ECG data containing late potentials characterised by Simson's (1981) method are used to identify late signal components terminal in the QRS. The bispectrum estimation technique is shown to offer an advantage over the conventional power spectra in that it represents the true phase characteristics produced by the nonlinear electrical excitation of the heart. Additionally, forming the bispectrum of the data set automatically suppressed any Gaussian signal components (the original measurement noise was modelled as being Gaussian in nature), and consequently resulted in automatic noise reduction.<>