文章摘要

文章信息/Info

Title:: Analysis and classification of brain network based on sub-period apnea and sleep stage

作者:: 赵今朝1 ; 刘铭1; 江秀全1; 史维友1; 娄宜泰1; 冷建材1; 徐舫舟1△; 冯超1; 杨清波2; 唐吉友3; 鲁珊珊3; （1．齐鲁工业大学（山东省科学院）光电工程国际化学院，济南 250353；2．齐鲁工业大学（山东省科学院）数学与统计学院，济南 250353；3．山东第一医科大学附属第一医院（山东省千佛山医院）神经内科，山东省免疫研究所，山东省风湿免疫与转化医学重点实验室，济南 250014）

Author(s):: ZHAO Jinzhao1; LIU Ming1; JIANG Xiuquan1; SHI Weiyou1; LOU Yitai1; LENG Jiancai1; XU Fangzhou1; FENG Chao1; YANG Qingbo2; TANG Jiyou3; LU Shanshan; （1. International School for Optoelectronic Engineering, Qilu University of Technology (Shandong Academy of Sciences), Jinan 250353,China；2.School of Mathematics and Statistics, Qilu University of Technology (Shandong Academy of Sciences), Jinan 250353；3. Department of Neurology, The First Affliated Hospital of Shandong First Medical University （Shandong Provincial Qianfoshan Hospital）, Institute of Neuroimmunology, Shandong Key Laboratory of Rheumatic Disease and Translational Medicine, Jinan 250014）

Keywords:: Electroencephalography ; Sleep stage; Classification; Brain functional connectivity; Phase-locked value; Sleep apnea

摘要:: 睡眠分期是评估睡眠质量的基础。然而，睡眠呼吸暂停(sleep apnea, SA)会改变测试者的睡眠结构，进而影响对睡眠分期的准确评估。因此，在评估睡眠质量时，准确检测睡眠呼吸暂停和睡眠分期至关重要。为准确评估睡眠分期，本研究通过研究脑区之间的功能连接，探讨了脑功能连接的相互作用关系。采用锁相值(phase locking value, PLV)在不同时间段上进行特征提取，构建功能连接网络；然后利用多个时间段的PLV进行特征融合，并通过LibSVM(library for support vector machines，LibSVM)结合分类性能优化策略的方法进行睡眠分期。同时，本研究还分析了睡眠呼吸暂停和正常呼吸对脑网络的影响。实验结果显示，睡眠呼吸暂停时的各脑区连通紧密程度大于正常呼吸时，并在子时段数为30时，睡眠分期的分类准确率达到了88.87％，呼吸暂停的检测准确率达到了93.64％。该算法在睡眠分类和呼吸暂停检测方面表现出良好性能，有助于推动脑电睡眠分类和呼吸暂停检测系统的开发和应用。

Abstract:: Sleep staging is the basic evaluation of sleep quality. However, sleep apnea (SA) changes the sleep structure of testers, which affects the accurate evaluation of sleep stages. therefore, the accurate detection of SA and sleep stages is very important in the evaluation of sleep quality. In order to accurately assess sleep stages, we discussed the interaction of functional connections between brain regions by studying the functional connections between brain regions. Using the phase-locked value (PLV) to extract features in sub-periods, and a functional connection network was constructed. Then, the PLV of multiple sub-periods was used for feature fusion, and the classification performance optimization strategy was used for sleep staging. At the same time, the brain network changes of SA and normal breathing were analyzed. The experimental results showed that when the number of sub-periods was 30, the classification result of sleep stage reached 88.87%, and the accuracy of detecting apnea reached 93.7%. The algorithm has good sleep classification and apnea detection performance, and can effectively promote the development and application of EEG sleep classification and apnea detection system.