Maximum likelihood analysis of late potentials

Maximum likelihood estimation is used for obtaining a function suitable for defining the endpoint of ventricular late potentials (LPs). The estimation procedure is based on a signal model which accounts for basic properties of late potentials in individual electrocardiogram (ECG) leads. The estimator provides a measure of the signal power resulting from eigenvector-based linear filtering of two beat subaverages followed by a nonlinear operation. A simplified model is also described in which only correlation across the ensemble of beats is considered. The performance of the present method was compared to that of conventional time domain analysis. The results obtained from 46 patients showed that large differences in the filtered QRS duration exist between the two methods. The present method was in several cases able to detect the LP endpoint at a lower signal-to-noise ratio. Reproducibility of the present method was studied by computing the endpoint for different beat subaverages.<>