Neural network detection of ventricular late potentials in ECG signals using wavelet transform extracted parameters

After recovery from acute myocardial infarction (MI), a significant number of patients remain at risk of sudden death, which is attributed to ventricular tachycardia (VT). Ventricular Late Potentials (VLPs) are associated with VT. VLPs are low amplitude high frequency signals that appear at the end of the QRS complex of an ECG recording. In this work, discrete Wavelet Transform (DWT) and Artificial Neural Networks (ANN) are applied in the analysis of ECG signals in order to identify VLPs. Results of this analysis are used to classify patients with and without VLPs in their ECGs. DWT were computed for a total of (38) different ECG records that included control signals and signals for patients with VT. A set of parameters were extracted from WT and used as inputs to neural networks for the classification. Multilayer feedforward ANNs employing the back-propagation (BP) learning algorithm were trained and tested using the WT extracted parameters.