Signal processing options for detecting conduction abnormalities in ischemic ventricles.

Abstract

Sources of error in averaging for late ventricular potentials include variations in the magnitude and timing of late potentials as well as variability in the temporal alignment of successive cardiac cycles during averaging. Since these errors can both attenuate high frequency signals and add artifactual components to the ECG, we developed methods to estimate and minimize them. Studies on patients indicated that small misalignments can obliterate high frequency components of the QRS, while broadening the low amplitude tail of the QRS, possibly leading to erroneous interpretations. The digital cross-correlation method of alignment was evaluated using simulated signals, and was shown to be relatively insensitive to gross variations in waveshape, and appeared to be less accurate than real-time pattern recognition schemes. A real-time alignment method based on a mathematical model was developed that could measure its accuracy, and was implemented in hardware. Our method improves alignment and average fidelity. Direct recordings from human and canine ventricles revealed that beat-to-beat variability in activation patterns of ischemic regions often followed a regular pattern such as 2:1 block. Spectral analysis of ECG recordings showed that a sub-harmonic, indicative of a 2:1 pattern, could be detected in the ECG during periods when ischemic regions of the ventricle were experiencing 2:1 block. These results suggest the possible utility of sub-harmonic analysis as a tool for detection of abnormal electrophysiological conduction.