Research On Eeg Key Sequence Extraction And Epilepsy Detection Based On Attention Mechanism

Epilepsy is caused by abnormal discharge of brain neurons.During the seizure attack,the patient is easy to be injured as the patient’s body is out of control.This kind of repeated torture seriously affects the life of the patient and his family.The purpose of epileptic seizure prediction research is to give early warning before seizure,in order to minimize the harm,which has very important application value.At present,epilepsy prediction research has made great progress,but it is far from meeting the needs of practical clinical application.It means a lot to study the information in the process of brain transition from normal to seizure from multiple feature dimensions and perspectives.On the one hand,this paper constructs a network model from the perspective of attention mechanism,which has achieved good results in epilepsy prediction;on the other hand,from the perspective of brain network evolution pattern analysis,this paper studies the EEG（Electroencephalogram）sequences before seizures,and detects the key sequences that can distinguish seizures.Firstly,the research history and current situation of epileptic seizure prediction and epileptic EEG key time series are systematically summarized.Through the in-depth analysis of the research ideas and results of the existing researchers,the advantages and disadvantages of different algorithms are shown.In the end of this section the common methods and attention mechanism of constructing epileptic brain connection are introduced.Secondly,in this section,we propose a method of epileptic EEG prediction based on the combination of attention mechanism and brain network characteristics.Besides a new mixed band attention artificial neural network model is proposed,which can be used to identify the change rule in the process of brain network connection.The proposed method is then realize the task of epileptic seizure prediction.The experiment shows that the method has achieved good results.Finally,this paper proposes a key time series detection method of epileptic EEG based on the combination of brain network evolution model and attention mechanism.The proposed method is used to construct brain network evolution model.In this section,the brain network connection is used to reconstruct the evolution pattern of brain network related to each electrode.Combined with the attention mechanism,the key time series with discrimination information causing the change of attention weight are found for the classification task of pre seizure and inter seizure by artificial neural network.A series of experiments show that this paper makes a practical and meaningful exploration on the evolution pattern of brain network.