Research On Epilepsy Detection Method Based On EEG Signal

Epilepsy is a chronic neurological disease with recurrent and indefinite seizures.About 70 million patients of different ages worldwide are deeply affected by it.Unpredictable damage may occur,which may result in the death of the patient in severe cases.Doctors need to make a diagnosis based on 24 hours or more of EEG recordings obtained from the patient’s brain,but this heavy workload often leads to misjudgment by doctors.Therefore,an automated epilepsy detection system is urgently needed to assist neurologists in making a fast and accurate diagnosis.In this paper,the algorithm research of epilepsy detection and prediction based on neural network is proposed.The innovation of this paper is that it proposes an epilepsy detection method based on temporal feature gating network(TFGN),and proposes an unsupervised learning deep convolutional generative adversarial network(DCGAN).The accuracy of epilepsy detection and prediction is improved,and prediction,and the model is comprehensive and reliable.The specific work of this paper is as follows:(1)The traditional epilepsy detection model inputs the extracted features into the classifier to achieve epilepsy detection.If the feature vector is to be unified with the input format of the classifier,it is necessary to reduce the dimension of the feature vector.Epilepsy detection method based on Temporal Feature Gating Network(TFGN),TFGN model combines extracted features and feature dimensionality reduction in one detection model,which saves the calculation amount of the model and improves the efficiency of the epilepsy detection model.Two kinds of Bonn datasets and CHB-MIT datasets with different sampling frequencies,different ages,and different channels are used to verify the comprehensiveness and reliability of the TFGN detection model.(2)The traditional epilepsy prediction model requires label data corresponding to the dataset,and the unsupervised learning deep convolutional generative adversarial network(DCGAN)prediction model saves the tedious work of labeling a large number of signals,while obtaining more accurate features to improve model prediction ability.Using an unsupervised deep convolutional generative adversarial network prediction model to obtain 100% recall rate provides a new idea for future epileptic seizure prediction.