Predicting Epileptic Seizures from Intracranial EEG Using LSTM-Based Multi-task Learning

摘要: Epilepsy afflicts nearly 1% of the world’s population, and is characterized by occurrence spontaneous seizures. It’s important to make prediction before seizures, so that epileptic can prevent seizures taking place on some specific occasions avoid suffering from great damage. The previous work in seizure paid less attention time-series information their performances may also restricted small training data. In this study, we proposed a Long Short-Term Memory (LSTM)-based multi-task learning (MTL) framework for prediction. LSTM unit was used process sequential data MTL applied perform latency regression simultaneously. We evaluated method American Society Seizure Prediction Challenge dataset obtained an average accuracy 89.36%, which 3.41% higher than reported state-of-the-art. addition, input output middle layers were visualized. visual experiment results demonstrated superior performance our LSTM-MTL