Paper Title
TWO DEEP LEARNING ARCHITECTURES FOR EPILEPSY SEIZURE PREDICTION

Abstract
Abstract - In this work, we propose and examine two deep learning architectures, for the prediction of epilepsy seizures, using the pieces of information on the electrical activities of the brain, recorded in the electroencephalography (EEG) signals. The architectures proposed are based on deep learning networks, the first one is a recurrent neural network named Long short-term memory (LSTM), and the second, is a non-recurrent network called Deep Feedforward Network (DFN) architecture. To create and perform DFN and LSTM architectures, we are based on 22 features extracted from different EEG segments, to produce a typical dataset consisting of the EEG of 5 patients, which aim to predict the impending epilepsy seizure of patients and predict three states of epilepsy patients. The proposal models give encouraging results in terms of classification rates, for different previous seizure times of 5min to 50min. Keywords - Epilepsy Seizure, EEG, prediction, Deep learning, LSTM.