High-resolution ECG analysis: a fuzzy approach to detect ventricular late potentials using a wavelet-based vector magnitude waveform

The objective of this paper is to investigate the performance of a fuzzy classifier, designed using nearest neighborhood clustering, in detection of ventricular late potentials (VLPs) when it uses the feature vectors extracted from a vector magnitude (VM) waveform based on the discrete wavelet transform (DWT). VLPs are low-amplitude, high-frequency signals which appear at the terminal part of the QRS complex in the high-resolution ECG (HRECG) signal and may be used as a non-invasive marker for patients prone to ventricular tachycardia (VT). In this research, the fuzzy classifier performance was investigated with two types of the time-domain feature vectors were extracted from the end part of the QRS complex in the wavelet-based VM waveform. These feature vectors were fed to the fuzzy classifier and a multilayer perceptron (MLP) simultaneously. The results show that the fuzzy classifier can detect VLPs better than the MLP neural network with less computational complexity.