智能手表睡眠检测的一致性和准确性分析  

Consistency and accuracy analysis of smartwatch sleep detection

作　　者：黄章辉 姚方来 张福军 刁姝尹 陈希 HUANG Zhanghui;YAO Fanglai;ZHANG Fujun;DIAO Shuyin;CHEN Xi(De Rucci Healthy Sleep Co.,Ltd.,Shenzhen,Guangdong 518000,China)

机构地区：[1]慕思健康睡眠股份有限公司,广东深圳518000

出　　处：《中国医学工程》2025年第2期1-7,共7页China Medical Engineering

摘　　要：目的 探讨智能手表的睡眠检测准确性,以及和多导睡眠监测(PSG)结果的一致性。方法 采用Alice PDx便携式多导睡眠监测仪和智能手表同时对15例健康成年人进行睡眠监测,统计组内相关系数(ICC)、Bland-Altman图和t检验等指标进行分析。结果 PSG和智能手表睡眠分期占比[浅睡、深睡和快速眼动睡眠(REM)阶段]的ICC都小于0.2,睡眠分期时长(夜间清醒时长、浅睡时长、深睡时长、REM时长和睡眠时长)的ICC差异较大,其中睡眠时长的一致性较高,深睡时长没有表示出明显的一致性;浅睡占比、深睡占比和REM占比的差值均值分别为0.16、-0.12和-0.05,睡眠时长、清醒时长、浅睡时长、深睡时长和REM时长的差值均值分别为-0.44、0.35、0.81、-0.89和-0.37,入睡时间点和清醒时间点的差值均值分别为0.01和-0.08;智能手表与PSG多导的浅睡占比、深睡占比、REM占比、清醒时长、浅睡时长、深睡时长、REM时长和睡眠时长比较差异有统计学意义(P<0.05),入睡时间点和清醒时间点比较差异无统计学意义(P>0.05);智能手表在清醒、浅睡、深睡和REM阶段的准确性分别为12.82%、51.69%、57.61%和33.93%,总体准确率为46.91%。结论 智能手表与PSG的入睡点和清醒点有良好的一致性,睡眠分期占比的一致性较差,浅睡和深睡阶段准确率较高,可以基本满足家庭睡眠分期监测的需求,为智能手表的使用和性能提升提供参考。【Objective】 To explore the accuracy of sleep detection by smartwatch and the consistency with polysomnography(PSG) results.【Methods】 The Alice PDx portable polysomnography and smartwatch were used to monitor the sleep of 15 healthy adults at the same time,and the statistical intraclass correlation efficient(ICC),Bland-Altman plots and t-tests were analyzed.【Results】 The ICC of sleep staging percentage(light sleep,deep sleep and REM stage) of both PSG and smartwatch was less than 0.2,and the ICC of sleep staging duration(nocturnal wakefulness,light sleep,deep sleep,REM and sleep duration) varied greatly,with the consistency of sleep duration being higher and deep sleep duration not indicating significant consistency.The light sleep percentage,deep sleep and REM percentage of difference means were 0.16,-0.12,and-0.05,respectively;the difference means of sleep duration,wakefulness duration,light sleep duration,deep sleep duration,and REM duration were-0.44,0.35,0.81,-0.89,and-0.37,respectively;the difference means of the time point of falling asleep and wakefulness were 0.01 and-0.08,respectively.The differences of smartwatch versus PSG polysomnography in the light-sleep duty ratio,deep sleep percentage,REM percentage,wakefulness duration,light sleep duration,deep sleep duration,REM duration and sleep duration were statistically significant(P<0.05),and the differences at the time point of falling asleep and wakefulness were not statistically significant(P>0.05).The accuracy of the smartwatch in the wakefulness,light sleep,deep sleep and REM phases were 12.82%,51.69%,57.61% and 33.93%,respectively,with an overall accuracy of 46.91%.【Conclusion】 The smartwatch has good consistency with PSG at the point of sleep onset and wakefulness,poor consistency in sleep staging percentage,high accuracy in light and deep sleep stages,which can basically satisfy the needs of home sleep staging monitoring,and provide a reference for the use of smartwatches and performance enhancement.

关 键 词：睡眠医学 智能手表 睡眠检测 睡眠分期 多导 

分 类 号：R74[医药卫生—神经病学与精神病学]