Epileptic Seizure Prediction Using Convolutional Autoencoder Based Deep Learning

Epilepsy is one of the most common neurological diseases in the world. Prediction of epileptic seizures are of great importance for both health care providers and epilepsy patients. In this study, seizure prediction is performed using Electroencephalographic (EEG) data. Unlike the traditional feature extraction methods, an autoencoder architecture which has been trained without supervision, is employed. The feature extraction system developed with this architecture is combined with a deep learning based classifier. By way of deriving an automatic feature extractor and classifier, an end-to-end seizure prediction system has been developed. The highest accuracy rate achiveved in the study is 92.2%.