Recognition Of Low Back Pain Pain EEG Signals Based On Neural Network Algorithm

In the global burden of disease research in recent years,pain is the leading cause of disability in men and women.When a person experiences pain interference for a long time,a series of emotional diseases,such as depression,anxiety and other negative emotions,will occur.In severe cases,it will affect the patient's quality of life and mental state.The treatment of low back pain has not achieved good therapeutic effects.The main reasons are: first,the mechanism of pain generation is intricate,and most of the current effectiveness evaluations are based on subjective clinical evaluation tables,which are subjective;second,they have language For patients with communication disorders,doctors can only judge the degree of pain based on their physical response to pain and facial expressions.Therefore,the use of scientific auxiliary methods to achieve precise treatment has become a top priority.Among them,the objective and effective recognition of low back pain is the key to auxiliary methods.This article is based on the subject of the Minimally Invasive Center,Shenzhen Institute of Advanced Technology,Chinese Academy of Sciences-"Quantitative Research on the Dysfunction and Pain Level of Patients with Non-specific Low Back Pain Based on Electromyography/Acoustomuscular Image" to classify low back pain pain The recognition was studied in depth.main tasks as follows:1)Use machine learning algorithms(including support vector machines,logistic regression)and feature extraction algorithms(including sample entropy extraction,co-space pattern extraction)to study pain classification and recognition.2)First,the VGG13 network,optimized one-dimensional convolution,and two-dimensional convolutional neural network algorithms are applied to the two-class classification of pain and health EEG signals,and their accuracy and precision The degree,recall rate and F1 value were analyzed in detail.3)Then the algorithm combining the convolutional neural network and the recurrent neural network: CNN+LSTM network algorithm,applied to the pain and health EEG signal classification research,and its accuracy,The precision,recall rate and F1 value were analyzed in detail.4)Finally,this article combines convolutional neural network and recurrent neural network: CNN+Bi-LSTM network algorithm is applied to the research of pain and health EEG signal classification,and the corresponding indicators: accuracy,precision,recall The F1 value is compared with the previous methods.At the same time,the AUC value and ROC curve are used to evaluate the classification results of the above algorithms.The test results show that the accuracy of the two-dimensional convolutional neural network algorithm in the classification of pain and health reached about 90%,and other indicators such as accuracy,recall rate,and F1 value were all 89%;the accuracy of the CNN+LSTM network algorithm reached The accuracy rate of the CNN+Bi-LSTM algorithm is about 97%,and the other indicators are all 97%.Experiments show that the algorithm combining CNN and RNN has the best effect.Therefore,the established classification model can be used as a feasible method to distinguish whether a subject is in a "chronic low back pain state",and can complement the clinical evaluation methods of low back pain patients.