Obstructive sleep apnea (OSA) is primarily characterized by intermittent nocturnal hypoxia and sleep fragmentation. Arousals interrupt sleep continuity and lead to sleep fragmentation, which can lead to cognitive dysfunction, excessive daytime sleepiness, and adverse cardiovascular outcome events, making arousals important for diagnosing OSA and reducing the risk of complications, including heart disease and cognitive impairment. Traditional arousal interpretation requires sleep specialists to manually score PSG recordings throughout the night, which is time consuming and has low inter-specialist agreement, so the search for simple, efficient, and reliable arousal detection methods can be a powerful tool to clinicians. In this paper, we systematically reviewed different methods for recognizing arousal in OSA patients, including autonomic markers (pulse conduction time, pulse wave amplitude, peripheral arterial tone, heart rate, etc.) and machine learning-based automated arousal detection systems, and found that autonomic markers may be more beneficial in certain subgroups, and that deep artificial networks will remain the main research method for automated arousal detection in the future.
阻塞性睡眠呼吸暂停（OSA）的主要特点是夜间间歇低氧及睡眠片段化。觉醒中断睡眠连续性，导致睡眠破碎，这可能会导致认知功能障碍、白天过度嗜睡以及心血管不良结局事件的发生，觉醒对于诊断OSA和降低并发症（包括心脏病和认知障碍）的发生风险非常重要。传统的觉醒判读需要睡眠专家对整夜的PSG记录进行手动评分，耗时长且专家之间的一致性较低，因此寻找简便、高效、可靠的觉醒检测方法，可为临床医生提供有力的帮助。本文系统综述识别OSA患者觉醒的不同方法，包括自主神经标志物（脉搏传导时间、脉搏波波幅、外周动脉张力、心率等）及基于机器学习的自动觉醒检测系统，发现自主神经标志物可能在特定的亚群中更有益，深度人工网络仍然是未来自动检测觉醒的主要研究方法。.