摘要目的:评价Somnolyzer系统在多导睡眠图(PSG)自动评分中的临床应用价值。方法:本研究为诊断实验研究。采用非随机抽样的方法选取2017年3月至2019年10月在中日友好医院睡眠诊疗中心行整夜PSG监测的患者138例。分别由2名睡眠技师进行人工判读(M1、M2)，Somnolyzer分析系统自动判读(S)及同样2名睡眠技师分别对自动评分进行编辑修正，获得人机协作的判读分析(SM1、SM2)。应用同类相关系数(ICC)评价各判读评分方法的一致性。睡眠分期和呼吸事件的判读依据2007年美国睡眠医学学会标准。结果:5种方法对睡眠结构和呼吸事件判读一致性良好，其中睡眠结构相关指标的平均ICC及95% CI分别为：N1% 0.973(0.965~0.979)；N2% 0.898(0.869~0.923)；N3% 0.951(0.937~0.963)；R% 0.964(0.954~0.973)；总睡眠时间0.991(0.988~0.993)；睡眠潜伏期0.933(0.913~0.949)；睡眠效率0.972(0.964~0.979)；觉醒指数0.991(0.989~0.994)。呼吸事件各指标的平均ICC及95% CI分别为：阻塞性呼吸暂停指数0.989(0.986~0.992)；中枢性呼吸暂停指数0.994(0.992~0.995)；混合型呼吸暂停指数0.998(0.997~0.998)；低通气指数0.988(0.984~0.991)；呼吸暂停低通气指数0.992(0.990~0.994)；氧减指数0.999(0.998~0.999)；最大氧减持续时间0.913(0.888~0.934)；呼吸努力相关微觉醒指数0.964(0.953~0.972)。 结论:Somnolyzer自动评分系统对睡眠结构参数和呼吸事件的判读结果与人工评分判读无显著差异，二者一致性良好。鉴于人工评分相关的负担，自动评分系统在PSG评分的具有补充应用价值。

abstractsObjective:To explore the clinical value of Somnolyzer system in automated scoring of polysomnography (PSG).Methods:This is a diagnostic test study.A non-random sampling method was used to enroll 138 patients underwent overnight PSG monitoring in Sleep clinic, China-Japan Friendship Hospital from March 2017 to October 2019.Each original PSG was manually scored by two experienced technologists (M1, M2), and the same original PSG was scored by Somnolyzer system to obtain automatic score (S), and the two technologists edited the automatic score and obtained two modified polysomnography scores (SM1, SM2). The intraclass correlation coefficient (ICC) was used to assess the consistency between groups.Sleep staging and respiratory events scores were conducted according to the 2007 American Academy of Sleep Medicine (AASM) criteria.Results:There were good agreements of five methods for the main parameters in both sleep structure and respiratory events.For sleep structure parameters, average ICC and 95% CI within 5 groups were: percentage of sleep stage N1 (N1%): 0.973, 95% CI (0.965-0.979); N2%: 0.898, 95% CI (0.869-0.923); N3%: 0.951, 95% CI (0.937-0.963); R%: 0.964, 95% CI (0.954-0.973); total sleep time (TST): 0.991, 95% CI (0.988-0.993); sleep latency (SL): 0.933, 95% CI (0.913-0.949); sleep efficiency (SE): 0.972, 95% CI (0.964-0.979); arousal index (ArI): 0.991, 95% CI (0.989-0.994). For respiratory events, average ICC and 95% CI within 5 groups were: obstructive apnea index (OAI): 0.989, 95% CI (0.986-0.992); central apnea index (CAI): 0.994, 95% CI (0.992-0.995); mixed apnea index (MAI): 0.998, 95% CI (0.997-0.998); hypoventilation index: 0.988, 95% CI (0.984-0.991); apnea hypopnea index (AHI): 0.992, 95% CI (0.990-0.994); oxygen desaturation index (ODI): 0.999, 95% CI (0.998-0.999); duration of oxygen desaturation: 0.913, 95% CI (0.888-0.934); respiratory effort related arousal (RERA) index: 0.964, 95% CI (0.953-0.972). Conclusions:No difference was significant in the interpretation of sleep parameters and respiratory events between the Somnolyzer automated scoring system and manual scoring, with good consistency.Considering the burden imposed by manual scoring, automated scoring systems may serve as an alternative complementary means of PSG scoring.