| Epilepsy is a chronic neurological disease affecting about 1% of the world’s population.Epilepsy is caused by abnormal electrical activity in the local brain area,which can reduce the life quality of patients,and even threaten the life safety of patients for its characteristics of repeatability and paroxysmal.At present,the treatment of epilepsy is generally assisted by the long-term monitoring of the patient’s pathogenesis.Electroencephalogram is the common tool for monitoring the brain’s electrical signals.However,because the monitoring time is very long and the seizure of epilepsy is sudden and transient,the long-term observation and monitoring will consume a lot of doctor’s energy and waste a lot of medical resources.With the development of signal processing technology and artificial intelligence technology,epilepsy detection and prediction as an important means of computer-aided diagnosis and computer-aided rehabilitation prevention has attracted more and more attention,which are playing an increasingly important role in automatic diagnosis,lesion location and disease prevention of epilepsy.Based on the above background,this paper carries out the following research work:(1)In this paper,a cascade filtering method for epileptic EEG data preprocessing is proposed.The method analyzes the time-domain and frequency-domain characteristics of the signal.And a two-stage cascade filter is designed to eliminate the effects of noise.The experimental results show that the relatively pure EEG signals can be obtained by the preprocessing method of epileptic EEG data,which verifies the effectiveness of the proposed method.(2)An automatic epilepsy detection algorithm based on feature matrix fusion is proposed.In view of the existing methods of multi-channel epileptic EEG signal channel spatial information considered insufficient,this paper proposes a method to construct a multi-domain and multi-mixed feature matrix,which not only comprehensively considers the time domain,frequency domain,linear and nonlinear characteristics of signals,but also focuses on the spatial information among multi-electrode channels.Then the sparse automatic encoder is used to complete the feature fusion and learning,and the hidden information in the feature is further mined.Finally,the extreme learning machine is used for classification and detection.The results show that the proposed detection algorithm improves 5.4% on the basis of the existing advanced methods,which not only has remarkable detection accuracy,but also greatly improves the detection efficiency.(3)An automatic epilepsy prediction algorithm based on convolutional neural network is proposed.Aiming at the problem that manual features can easily cause the loss of EEG signal information,the data tensor representation form of multi-domain and multi-frequency band of EEG signal is constructed through the signal decomposition method,which comprehensively represents the multi-aspect characteristics of the signal.Then a convolutional neural network is designed to learn features automatically,and the features of epileptic EEG signals that are easy to be ignored are focused on by combining the attention mechanism.The epilepsy prediction algorithm proposed in this paper obtained better prediction results under the experiment based on the non-specificity of patients.And the post-processing technology proposed in this paper further effectively reduced the false prediction rate of the model,reaching the prediction accuracy of 86% on the CHB-MIT dataset.(4)A prototype BCI system for online epilepsy monitoring and alarm is designed and implemented.The system proposed has realized three modules,including EEG signal acquisition,EEG data transmission and EEG data processing.The EEG data processing module has realized real-time display,model detection,monitoring and alarm functions.On the basis of single-channel BCI system,multi-channel epilepsy monitoring and alarm system is designed,which is of great significance in assisting doctors in diagnosis and treatment and avoiding waste of medical resources.The experimental results in this paper show that the EEG signals of different channels are different.So the spatial characteristics between channels play an important role in epilepsy detection and prediction.Moreover,due to the complexity of EEG signals in epilepsy,a comprehensive characteristic analysis of EEG can get a better effect.The results of this paper effectively improve the accuracy and efficiency compared with the existing methods,which contribute to the development of computer-aided diagnosis technology. |
