Font Size: a A A

A Study Of Sleep Detection Methods In Patients With Sleep-disordered Breathing

Posted on:2021-06-13Degree:MasterType:Thesis
Country:ChinaCandidate:Y L NiuFull Text:PDF
GTID:2514306038986929Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
The problem of sleep disorders is becoming more and more serious worldwide,which has a serious impact on people's daily life,physical health,and national economy.Sleep-disordered breathing(SDB)is the most common chronic disease in sleep disorders.It is necessary to research portable devices that can replace traditional PSG to achieve high-precision sleep monitoring of SDB patients.This paper proposes SDB sleep monitoring methods based on two strategies to provide a theoretical basis for future research and development of home-based sleep monitoring equipment for SDB.The main research contents of this article include:1.To improve the performance of the classification model,two strategies are proposed from the perspective of features and data volume.The first strategy is a feature selection algorithm that combines ANOVA with maximum correlation and minimum redundancy algorithms.The second strategy is a weight-based resampling algorithm.Combining the two strategies,the classification model framework of this paper is proposed.2.Use this framework to study the sleep staging algorithm of the SDB population.Firstly,it is explained that the research on sleep staging algorithm for SDB population is the necessary work to realize high-precision sleep monitoring of SDB population.The performance of the classification model before and after adopting the two strategies is analyzed.The classification results of SDB patients using the healthy person model and the SDB model are different.3.Research on classification and recognition algorithm of sleep apnea using this framework.The experimental conclusion is:1.The overall classification accuracy of the three classifications of SDB patients before and after the adoption of the two strategies increased by 8%on average,the Kappa coefficient increased by 0.300 on average,and the training time decreased by 68s on average.The overall classification accuracy of the three categories of healthy people increased by 9.3%on average,the Kappa coefficient increased by 0.353 on average,and the training time decreased by 36s on average;this shows that the two strategies proposed can effectively improve the performance of the classification model.2.The SDB sleep staging algorithm proposed in this paper uses 9 features for classification 1 to achieve an overall classification accuracy of 80%,classification 2 uses 8 features to achieve an overall classification accuracy of 83%,and classification 3 uses 8 features to achieve an overall The classification accuracy rate is 85%,which can meet the needs of portable sleep monitoring.However,when SDB patients used a healthy person model to identify sleep phases,they found that the accuracy rate was significantly lower than that of the SDB model.Category 1 fell by 121%,Category 2 fell by 31%,and Category 3 fell by 11%.The results further illustrate that the sleep staging model of healthy people is not suitable for the SDB population.Exploring the sleep staging algorithm for SDB is very important for portable devices that achieve high-precision sleep monitoring of SDB patients.3.The 9 features adopted by the sleep apnea recognition model proposed in this paper achieve a total classification accuracy rate of 80%and a Kappa coefficient of 0.61,which can meet the needs of portable sleep monitoring.
Keywords/Search Tags:Sleep-disordered breathing, sleep staging, sleep apnea, feature selection, data balance
PDF Full Text Request
Related items