Detection of Ventricular Late Potentials by Adaptive Filtering and Cross-Spectral Analysis

This paper presents a novel technique to detect ventricular late potentials in frequency domain. The approach is based on the assumption that ventricular late potentials are associated with the occurrence of signal components at higher frequencies than those present in the high-resolution ECG of healthy subjects. This justifies the adoption of an adaptive high pass filter whose cut-off frequencies in set according to the characteristics of the averaged high-resolution ECG. Subsequent segments selected on the filter output are cross-spectral analyzed to outline the correlation between ventricular late potentials and the abrupt modification of the spectro-temporal representation of the terminal portion of the QRS complex and the early ST segment. An index is defined to detect anomalous frequency components. This is shown effective by processing real recording as well as signals in which simulated waveforms were superimposed on the ECG. The index is applied to the analysis of the high-resolution ECG of patients with Arrhythmogenic Right Ventricular Disease. The results suggest that the proposed technique may be utilized in the stratification of the seriousness of Arrhythmogenic Right Ventricular Disease