Epileptic Seizure Onset Detection Algorithm Using Dynamic Cascade Feed-Forward Neural Networks

Effective feature extraction and accurate classification of EEG signals have important role in successful of epileptic seizure onset detection algorithms. In this paper, a seizure onset detection algorithm based on dynamic cascade feed-forward neural networks (DCFNN) is proposed. In this algorithm, spectral and spatial features are extracted from the L-second seizure and non-seizure EEG signals. Then a DCFNN is used to determine an optimal nonlinear decision boundary. This algorithm has two advantages: (1) the extracted features can create maximum distinction between two classes. (2) the used DCFNN classifier have an inherently parallel structure and guaranteed to converge to a optimal classifier as the size of the representative training set increases. The performance of algorithm is evaluated based on three measures, sensitivity, specificity and latency. The results indicate that our algorithm obtains a higher sensitivity and smaller latency in relation to other algorithms.