Epilepsy-Net: attention-based 1D-inception network model for epilepsy detection using one-channel and multi-channel EEG signals

In this paper, we propose and evaluate Epilepsy-Net, a collection of deep learning EEG signal processing tools to detect epileptic seizures against non-epileptic seizures without any handcrafted features extraction. In the Epilepsy-Net model, the 1D-convolutional neural networks (CNN), the recurrent neural network (RNN) and the attention mechanism are combined, where each algorithm is represented by the ResNet and Inception, the gated recurrent unit and the convolutional block attention module respectively; without any handcrafted features. To the best of our knowledge, Epilepsy-Net is the first EEG signal processing work to detect epileptic seizures by combining the attention mechanism with the Inception deep network algorithm.We validate our Epilepsy-Net through several large public EEG signal datasets. The results of our experiments show that the proposed attention deep learning approach is an effective tool for epilepsy detection using EEG signals with high accuracy of 100%, 99.05% and 98.22% for the Bonn EEG dataset, variant of the Bonn EEG dataset, and CHB-MIT dataset, respectively.