Late potential recognition using Wigner-Ville distribution and neural network

In this paper, we present a method for the recognition of simulated late potentials in ECG by using a Wigner-Ville Transform (WVT) associated with a Neural Network (NN). We extract 64 ms of the ECG signal after the J point. This part of the signal is then processed by the WVT. The NN is trained with the energy value of the time-frequency transform obtained from 75 simulated ECG with late potentials and 25 simulated ECG without late potentials. The pattern recognition of the NN is then tested with 75 ECG, and we achieve in 94% of sensitivity and 88% of specificity.