A deep convolutional neural network designed for multi-channel EEG signals achieved 90.5% prediction accuracy for epileptic seizure detection on the SNUH-HYU dataset.
Does a deep convolutional neural network accurately detect epileptic seizures in multi-channel EEG signals?
A deep convolutional neural network designed for multi-channel EEG signals achieves high accuracy in detecting epileptic seizures by considering spatio-temporal correlations.
A new epileptic seizure detection method based on deep convolutional network is proposed. The proposed network is designed for multi-channel EEG signals and considers spatio-temporal correlation, a feature in epileptic seizure detection, using 1D and 2D convolutional layers. 1D convolutional layer considers temporal evolution of the EEG signal of each channel and 2D convolutional layers considers spatial relationships between EEG channels. We make datasets for training and test by extracting the EEG segments from CHB-MIT EEG Scalp database and SNUH-HYU EEG database: the recordings of long-term EEG monitoring at Seoul National University Hospital and Children's Hospital Boston. Our model is trained and tested using the EEG segments with varying durations. We also investigate the effect of artifact elimination on epileptic seizure detection by applying a low-pass filter to the EEG signals. Our model achieves 90.5% prediction accuracy with SNUH-HYU EEG dataset.
Park et al. (Mon,) conducted a other in Epileptic seizure. Deep convolutional neural network was evaluated on Prediction accuracy. A deep convolutional neural network designed for multi-channel EEG signals achieved 90.5% prediction accuracy for epileptic seizure detection on the SNUH-HYU dataset.