Deep Learning Models for Seizure Prediction using the Wavelet Transform

Epilepsy is a disease that affects 1% of the world’s population, and a third of those afflicted suffer from seizures which cannot be controlled by medical intervention. Epilepsy causes seizures, temporary confusion, and psychological symptoms. A seizure is an unanticipated, synchronous and recurrent brain cell discharge. Seizures can be detected using electroencephalogram (EEG) signals. In this paper, three ML models are proposed to tackle the pertinent issue of seizure prediction, with training done on both a patient-specific and general basis. The models are a Convolutional Neural Network (CNN), a Long Short-Term Memory Convolutional Neural Network (CNNLSTM) and an Artificial Neural Network (ANN). The best model, the CNN-LSTM, produced an AUC of 0.769 and an accuracy of 0.906 on one of the patient’s data.