Automatic detection of epileptiform activity using wavelet and expert rule base

We present a new method to detect epileptiform activity based on wavelet transform (WT), an artificial neural network and an expert rule system. The method consists of three steps. First, we extract features of spike events on the wavelet subspace. It appears technically feasible to reduce computational complexity. Then, the features are trained and tested to decide epileptic events with three layer feedforward networks employing the backpropagation (BP) learning algorithm. Finally, to confirm and validate epileptiform activity, we apply an expert system based on rule base. The result shows that the wavelet transform reduced data input size and the preprocessed artificial neural network (ANN) are more accurate than those of ANN with the same input size of raw data. In clinical tests, our expert system was capable of rejecting artifacts commonly found in EEG recordings.