Research On Sleep Staging Based On Extreme Gradient Boosting Algorithm

Various diseases related sleep seriously affect people’s health.Polysomnography is the main method for the diagnosis of sleep disorders.However,the accuracy of the existing polysomnography monitoring system’s automatic diagnosis is low,and the diagnosis process still mainly relies on expert manual operation,which is time-consuming and labor-consuming.Therefore,it is of great significance to develop a high accuracy polysomnography intelligent analysis algorithm suitable for clinical application.Based on electroencephalogram,electromyograeog and electrooculogram,this thesis implements a sleep monitoring system staging algorithm,the main research contents are as follows:1.Analyzing the relationship between EEG,EMG,EOG and sleep stage,45 features with strong correlation in temporal,frequency and nonlinearity domains are extracted as candidate sets,and 20 optimal features are selected as the input of the classifier through the maximum information coefficient filtering method,embedding method and wrapper method.2.Studying the sleep staging algorithms based on support vector machine(SVM)and extreme gradient boosting(XGBoost),this thesis uses the weight search method to optimize the sample imbalance problem,and the extreme gradient boosting algorithm with higher accuracy is used to build the model.The generalization error of the model is optimized by adjusting the parameters.At last,the classifier staging results are modified combined with the physiological characteristics of the sleep cycle,then the automatic sleep staging results are obtained.3.Based on the VS2008 MFC framework,the automatic staging algorithm of sleep monitoring system analysis software is completed by using C++.4.120 training sets and 80 test sets were randomly composed of the first 200 groups of data from the National Sleep Research Resource Database of the United States.The accuracy,precision,recall and F1 score are used to evaluate the algorithm.The test results are as follows: the average accuracy is 83.24%,the precision and recall of Wake and NREM2 are more than 80%,and the precision and recall of NREM3 and REM are more than 70%.Compared with the results before optimization,the F1 score of NREM1 increased by 10%.The results of clinical trials and staging analysis show that the algorithm can be applied to the clinical analysis of sleep researches.In this thesis,an automatic sleep staging algorithm and analysis software based on extreme gradient boosting tree are implemented.The performance of the algorithm is verified by the database and clinical trials,which proves that it has the advantages of high accuracy and fast running speed.The research results have passed the preliminary evaluation of China astronaut center,and are expected to be applied to manned space missions and clinical sleep disorders diagnosis system,which has good practical value.