Assessing the severity of sleep apnea syndrome based on ballistocardiogram.
Assessing the severity of sleep apnea syndrome based on ballistocardiogram.
复制标题
根据心冲击图评估睡眠呼吸暂停综合征的严重程度
DOI：
10.1371/journal.pone.0175351
复制
发表时间：
2017
期刊：
影响因子：
3.7
通讯作者：
Yu Z
中科院分区：
文献类型：
--
作者：
Wang Z;Zhou X;Zhao W;Liu F;Ni H;Yu Z
背景睡眠呼吸暂停综合征是一种常见的与睡眠相关的呼吸紊乱，全世界约有4%-7%的男性和2%-4%的女性受到影响。目前已有多种方法用于SAS的诊断和病情评估，包括睡眠研究领域的金标准多导睡眠图(PSG)以及单通道心电、脉搏血氧饱和度等替代技术。然而，许多缺点仍然限制了它们在家庭环境中的推广。在这项研究中，我们的目的是提出一种基于BCG信号的自动评估睡眠呼吸暂停综合征严重程度的有效方法，该方法是非侵入性的，适用于家庭环境。方法开发一套非侵入性睡眠监测系统采集BCG信号，在此基础上提出睡眠呼吸暂停综合征严重程度评估的三阶段框架，即数据预处理、睡眠呼吸事件检测和睡眠呼吸暂停综合征严重程度评估。首先，在数据预处理阶段，为了克服BCG信号的局限性(如精度和可靠性低)，利用小波分解提取心跳信号的轮廓信息，并应用RR校正算法来处理丢失或虚假的RR间期。然后，在事件检测阶段，我们提出了一种基于迭代累积平方和的睡眠呼吸事件自动检测算法Physio_ICSS(即ICSS算法)，该算法最初用于检测时间序列中的结构性断点。特别是，为了在获得的RR间隔时间序列中有效地检测与睡眠相关的呼吸事件，该算法不仅探索了与睡眠相关的呼吸事件的实际因素(例如持续时间的限制和可能发生的睡眠阶段)，而且克服了现有方法的事件分割问题(例如，等长分割方法可能会将一个与睡眠相关的呼吸事件分割成不同的片段而导致错误的结果)。最后，通过融合从多个领域提取的特征，我们可以有效地识别与睡眠相关的呼吸事件，并评估睡眠呼吸暂停综合征的严重程度。结论136例不同程度睡眠呼吸暂停综合征患者的实验结果验证了该框架的有效性，准确率为94.12%(128/136)。
Background Sleep Apnea Syndrome (SAS) is a common sleep-related breathing disorder, which affects about 4-7% males and 2-4% females all around the world. Different approaches have been adopted to diagnose SAS and measure its severity, including the gold standard Polysomnography (PSG) in sleep study field as well as several alternative techniques such as single-channel ECG, pulse oximeter and so on. However, many shortcomings still limit their generalization in home environment. In this study, we aim to propose an efficient approach to automatically assess the severity of sleep apnea syndrome based on the ballistocardiogram (BCG) signal, which is non-intrusive and suitable for in home environment. Methods We develop an unobtrusive sleep monitoring system to capture the BCG signals, based on which we put forward a three-stage sleep apnea syndrome severity assessment framework, i.e., data preprocessing, sleep-related breathing events (SBEs) detection, and sleep apnea syndrome severity evaluation. First, in the data preprocessing stage, to overcome the limits of BCG signals (e.g., low precision and reliability), we utilize wavelet decomposition to obtain the outline information of heartbeats, and apply a RR correction algorithm to handle missing or spurious RR intervals. Afterwards, in the event detection stage, we propose an automatic sleep-related breathing event detection algorithm named Physio_ICSS based on the iterative cumulative sums of squares (i.e., the ICSS algorithm), which is originally used to detect structural breakpoints in a time series. In particular, to efficiently detect sleep-related breathing events in the obtained time series of RR intervals, the proposed algorithm not only explores the practical factors of sleep-related breathing events (e.g., the limit of lasting duration and possible occurrence sleep stages) but also overcomes the event segmentation issue (e.g., equal-length segmentation method might divide one sleep-related breathing event into different fragments and lead to incorrect results) of existing approaches. Finally, by fusing features extracted from multiple domains, we can identify sleep-related breathing events and assess the severity level of sleep apnea syndrome effectively. Conclusions Experimental results on 136 individuals of different sleep apnea syndrome severities validate the effectiveness of the proposed framework, with the accuracy of 94.12% (128/136).