An optimization approach to recognition of epileptogenic data using neural networks with simplified input layers

This study introduces a simplified approach for the implementation of artificial neural networks (ANN) for the recognition of epileptic data in electroencephalograph (EEG) recordings. The training set construction is based on a trend-adaptive polygon which simplifies the search process as it reduces the size of the training set. This data reduction, at a sampling rate of 200 Hz, yielded a reduction ratio of 34% as a minimum to an 81% in the best case scenario. With a higher sampling rate of 500 Hz, a reduction ratio of 73% as a minimum to an impressive 92% in the best case scenario was achieved. The outcome is thus a computationally attractive classifier with a simpler design implementation and with higher prospects for accurate diagnosis. The algorithm was trained and tested with EEG data from four epileptic patients using the k-fold cross-validation technique.