Automatic Seizure Detection based on a Convolutional Neural Network-Recurrent Neural Network Model
Epilepsy is one of the most common diseases that impact 1-2% of the world's population. Detecting seizures through electroencephalogram (EEG) data is a common way for epilepsy diagnoses. Nowadays, there are some automatic systems for detecting seizures through EEG signals. However, such systems have their shortcomings. Some of the automatic systems use manual feature extraction, which is not suitable for generalization in the future. Furthermore, the automatic systems without manual feature extraction do not have high performances. In this paper, a convolutional neural network combined with a recurrent neural network, known as the CNN-RNN model, is developed to detect seizures under the EEG raw signal dataset “CHB-MIT”. The proposed model is designed to achieve a high performance without manual feature extraction. The promising results with a high accuracy of 99.24%, specificity of 99.29%, and recall of 99.16% show good validity and effectiveness of the proposed model.
Meeting number: 2621 884 9664 Password: FEutbk5NJ33