基于EEMD衍生的慢波活动指标在阻塞性睡眠呼吸暂停严重程度关联上优于传统N3分期
An EEMD-derived slow-wave activity metric improves association with obstructive sleep apnea severity beyond conventional N3 staging.
文献信息
| PMID | 42731337 |
|---|---|
| 原文 | 在 PubMed 查看原文 ↗ |
| 发表日期 | 2026 |
| 作者 | Huimin Sun |
| 作者单位 | Key Laboratory of Child Development and Learning Science, Ministry of Education, School of Biological Science & Medical Engineering, Southeast University, Nanjing, 211189, China; Center for Nonlinear Dynamics in Medicine, Southeast University, Nanjing, 211189, China. Electronic address: huiminsun@seu.edu.cn. |
| 期刊 | Sleep medicine |
| SCI 分区 | Q1 |
| IF | 4.5 |
| 研究类型 | AI/ML · 临床 |
| 所属专科 | 鼻科 |
中文摘要
研究目的： 传统睡眠分期依赖视觉标准，忽视了睡眠构型的连续转换以及脑波的个体差异。因此，传统N3睡眠指标是否足以捕捉与异质性睡眠障碍（如阻塞性睡眠呼吸暂停，OSA）相关的睡眠结构改变仍不确定。我们的目标是确定，基于慢波活动（SWA）升高方法衍生的慢波睡眠（SWS）是否比传统N3睡眠与呼吸暂停低通气指数（AHI）表现出更强的关联。
方法： 我们分析了基于社区的睡眠心脏健康研究（SHHS）中的多导睡眠图（PSG）数据，共纳入823名符合条件的参与者（年龄：45-90岁；412名男性和411名女性）。使用集合经验模态分解（EEMD）对睡眠脑电图（EEG）进行分解。识别慢波相关本征模态函数（IMF）以推导归一化SWA，并应用经验性阈值和数据驱动阈值检测SWS片段。将SWS指标与AHI的关联与传统N3睡眠的关联进行比较。根据心率反应（ΔHR）和缺氧负荷（HB）进一步将参与者分为高风险组和低风险组。在不同分层阈值下评估相关性趋势。
结果： 与传统睡眠评分相比，SWS量化与AHI的相关性显著更强，相关系数从传统N3指标的-0.27至-0.28显著改善至SWS指标的-0.36至-0.39。分层分析显示，这种相关性强度的增强在高风险组（ΔHR或HB升高）中明显更大，在某些分层阈值下相关系数高达-0.70。
结论： 本研究基于内在EEG动态波动而非经验性视觉阈值来识别SWS。EEMD衍生的SWA与OSA严重程度的关联比传统N3分期更强。
英文摘要
STUDY OBJECTIVES: Conventional sleep staging relies on visual criteria, overlooking the contiguous transitions of sleep configuration and individual variability of brainwaves. Consequently, whether conventional N3 sleep metrics sufficiently capture sleep architecture alterations associated with heterogeneous sleep disorders such as obstructive sleep apnea (OSA) remains uncertain. Our objective was to determine whether slow-wave sleep (SWS) derived from an elevated slow-wave activity (SWA)-based approach exhibits stronger associations with the apnea-hypopnea index (AHI) than conventional N3 sleep.
METHODS: We analyzed polysomnography (PSG) data from the community-based Sleep Heart Health Study (SHHS), comprising 823 eligible participants (age: 45-90 years; 412 males and 411 females). Sleep electroencephalogram (EEG) was decomposed using ensemble empirical mode decomposition (EEMD). Slow-wave-related intrinsic mode functions (IMFs) were identified to derive normalized SWA, and empirically informed and data-driven thresholds were applied to detect SWS episodes. Associations between SWS metrics and AHI were compared to those from traditional N3 sleep. Participants were further stratified into high- and low-risk groups based on heart rate response (ΔHR) and hypoxia burden (HB). Correlation trends were evaluated across different stratification thresholds.
RESULTS: Compared with traditional sleep scoring, SWS quantification yielded substantially stronger correlations with AHI, with correlation coefficients improving significantly from -0.27 to -0.28 for traditional N3-based metrics to -0.36 to -0.39 for SWS-based metrics. Stratification analyses revealed that this enhancement in correlation strength was markedly greater in the high-risk group (elevated ΔHR or HB), reaching correlation coefficients of up to -0.70 under certain stratification thresholds.
CONCLUSIONS: This study identifies SWS based on intrinsic EEG dynamical fluctuations rather than empirical visual thresholds. EEMD-derived SWA shows stronger associations with OSA severity than conventional N3 staging.