一种仅使用呼吸信号检测非眼动睡眠的方法  

Detection of Non-rapid Eyes Movement Sleep only by Respiration Signals

作　　者：李延军[1] 仲崇发[1] 李琳[1] 祝瑞云[1] 

机构地区：[1]中国航天员科研训练中心,北京100094

出　　处：《航天医学与医学工程》2015年第4期253-258,共6页Space Medicine & Medical Engineering

基　　金：国家自然科学基金资助项目(61401417);中国航天员科研训练中心航天医学基础与应用国家重点实验室资助项目(SYFD130051865)

摘　　要：目的利用非眼动睡眠呼吸平稳的特征进行非眼动睡眠的检测。方法在呼吸信号时域定义幅度波动的变化率rA,在频域定义呼吸率的不规整度rf,以rA与rf为自变量建立线性分类判别方程。选择EDF睡眠扩展数据库的9位受试者(每人睡眠2晚)共18例睡眠口鼻气流记录,其中9例记录用于模型训练,其余9例记录用于测试。结果个体化模型的训练错误率为20.0%,测试错误率为28.3%;通用模型的训练错误率为22.7%,测试错误率为30.8%。结论非眼动睡眠中,口鼻气流比较规整;觉醒与眼动睡眠中,呼吸不规整且波动较大。仅使用呼吸幅度变异性与呼吸率规整度可粗略对非眼动睡眠进行识别。Objective To detect non-rapid eyes movement （NREM） sleep by the stationary feature of respiration during NREM sleep. Methods The amplitude variability ra in the time domain, and the irregular ratio r/of respiratory rate in the frequency domain were defined respectively for respiratory signal. A linear discriminant function with rA and rf as its independent variables was set up for classification. Eighteen ora-nasal airflow recordings from 9 subjects were selected from the expanded sleep database in european data format （ EDF）, with 9 recordings for the training of the classification model and the rest 9 recordings for testing. Results The error rate of individual-dependent model was 20.0% and 28.3% respectively during the training phase and testing phase, while it increased to 22.7% and 30.8% respectively for independent model during the training phase and testing phase. Conclusion Ora-nasal airflow presents regular patterns during NREM sleep, while it shows high variability and irregular patterns during wake and REM sleep. A rough detection of NREM sleep could be achieved by only using the respiration amplitude variability and irregular ratio of respiration rate.

关 键 词：自动睡眠分期 非眼动睡眠 呼吸信号 呼吸幅度变异性 呼吸率规整度 线性分类 

分 类 号：R318.04[医药卫生—生物医学工程]