Epilepsy EEG Signal Based On Machine Learning Research On Classification Algorithm

Epilepsy,a refractory neurological disease caused by abnormal discharge of neurons in the brain,suffers from epilepsy patients,but the diagnosis of epilepsy is still a medical problem.In clinical medicine,there are two main ways to diagnose epilepsy:(1)Diagnosed by experienced medical workers by looking at the EEG waveform of patients with epilepsy.This method is highly susceptible to subjective factors and doctor experience It is time-consuming and laborintensive,and can easily lead to misdiagnosis;(2)use intelligent epilepsy assist system for assisted diagnosis.The method is to use the artificial intelligence model to automatically identify the EEG image information,and then further analyze the wave pattern features extracted from the image information,and then compare the characteristics of epilepsy to provide medical staff with diagnostic assistance.This method can not only reduce the misdiagnosis caused only by human diagnosis,but also speed up the detection and classification speed and improve the quality of epilepsy classification.To use the intelligent epilepsy assistant system for auxiliary diagnosis,it is necessary to automatically classify the epilepsy EEG signals.Although the current algorithm has a high recognition rate,due to the nonstationary and nonlinear characteristics of the epilepsy signal,the existing algorithm has three major problems:(1)In the traditional automatic detection algorithm,a feature extractor needs to be designed manually,which will affect the overall performance;(2)the training data set is too small,and often a small number of patients’ EEG data are used for experiments,so they cannot have Illustrative and practical;(3)The complexity and particularity of EEG signals need to be relearned for each patient,which will inevitably lead to poor robustness and generalization ability.In response to the above problems,the research work of this article is asfollows:(1)Aiming at the problem that artificially designed feature extractors affect performance,this paper applies multi-convolution neural networks(CNNs)for epilepsy classification algorithms.CNN can automatically learn the parameters of the model during the training process,so there is no need to manually design the feature extractor.In the trainingprocess,each CNN model separately trains each patient’s data.During the test,a small amount of data is used to adjust the parameters of each training model,and then the output of each adjusted model is formed into a discrimination vector,and then the final discrimination type is obtained.Experiments show that,compared with the single CNN network model,the classification effect is better,and the indicators such as sensitivity,specificity and recognition rate are significantly better.(2)For the problem that too few training data sets affect the performance of the model,this article does not use the traditional single-lead data experiment,but use single-Channel(C3-P3)and double-Channels(C3-P3,C4-P4)Combined,the data set uses the open source CHB-MIT data set of Boston Children’s Hospital,which consists of 24 patients with 700 hours of EEG signals during 198 episodes.Experiments show that the classification effect of the data processing method using double Channels is significantly higher than that of single Channel.(3)Aiming at the problem of poor robustness and generalization ability of the automatic classification algorithm for epilepsy,this paper combines a recurrent neural network(RNN)and a long-short-term memory network(LSTM)to form an RNN-LSTM network.First,the original data set needs to be normalized,segmented storage,and labeling.Secondly,the time domain analysis method is used to extract the characteristics of the epilepsy signal to join the model for training,and the model is optimized through continuous parameter adjustment.Finally Classification by soft Max classifier.Experimental results show that,compared with a single RNN network,therecognition effect is better,and various indicators such as sensitivity,specificity and recognition rate are significantly better.The innovation of this paper is to use CNNs network and RNN-LSTM network to compare and analyze epilepsy classification through two data processing methods: single-Channel(C3-P3)and double-Channels(C3-P3,C4-P4).Compared with the existing single data processing,single CNN and single RNN network,the problem of low recognition degree is proved.Experiments show that the method of this paper has achieved good classification results.