Nassi Thijs, Amidi Yalda, Oppersma Eline, Donker Dirk W, Redeker Nancy S, Westover M Brandon, Thomas Robert J
Beth Israel Deaconess Medical Centre, Harvard Medical School, Boston, MA, United States.
Cardiovascular and Respiratory Physiology, TechMed Centre, University of Twente, Enschede, The Netherlands.
Sleep. 2025 Jul 24. doi: 10.1093/sleep/zsaf213.
Loop gain (LG) is a critical parameter for assessing ventilatory control stability in sleep apnea, with implications for personalized treatment. Existing LG estimation methods are hindered by complex processing and specialized equipment, limiting clinical applicability. This study aims to develop an automated method to quantify LG from respiratory inductance plethysmography (RIP) signals to enhance precision management of sleep apnea.
Polysomnography data from Massachusetts General Hospital, high-altitude studies at Beth Israel Deaconess Medical Centre, and heart failure patients were analysed. Cases included an apnea-hypopnea index >15 and ≥ 4 hours of recorded sleep. RIP signals were filtered, normalized, and segmented into 8-minute windows. LG estimation employed an augmented Mackey-Glass equation and an expectation-maximization algorithm. Simulation experiments on synthetic breathing data with known parameter values quantified the accuracy of our parameter estimates.
Data from 465 patients were analysed, including 400 patients from the Massachusetts General Hospital dataset and 65 heart failure patients. The method accurately estimated LG across diverse apnea phenotypes. Patients with a higher central apnea index, high self-similarity or heart failure exhibited significantly higher median LG values (0.19, 0.27 and 0.41 respectively) compared to those with obstructive apnea (median LG = 0.11-0.14; p<.001). In addition, LG was significantly elevated during non-rapid eye movement sleep and at higher altitudes.
The automated LG estimation method developed in this study provides a scalable, non-invasive tool for endotyping in sleep apnea. By accurately modelling patient-specific ventilatory control, this approach supports personalized management strategies in apnea and broader clinical contexts.
环路增益（LG）是评估睡眠呼吸暂停通气控制稳定性的关键参数，对个性化治疗具有重要意义。现有的LG估计方法因处理复杂和设备专用而受到阻碍，限制了临床应用。本研究旨在开发一种自动方法，从呼吸感应体积描记法（RIP）信号中量化LG，以加强睡眠呼吸暂停的精准管理。
分析了来自麻省总医院的多导睡眠图数据、贝斯以色列女执事医疗中心的高原研究数据以及心力衰竭患者的数据。病例纳入标准为呼吸暂停低通气指数>15且记录睡眠时长≥4小时。对RIP信号进行滤波、归一化处理，并分割为8分钟的窗口。LG估计采用增强型Mackey-Glass方程和期望最大化算法。对具有已知参数值的合成呼吸数据进行模拟实验，以量化我们参数估计的准确性。
分析了465例患者的数据，其中包括来自麻省总医院数据集的400例患者和65例心力衰竭患者。该方法能够准确估计不同呼吸暂停表型的LG。与阻塞性呼吸暂停患者（LG中位数=0.11 - 0.14；p<0.001）相比，中枢性呼吸暂停指数较高、自相似性高或患有心力衰竭的患者表现出显著更高的LG中位数（分别为0.19、0.27和0.41）。此外，在非快速眼动睡眠期间和海拔较高时，LG显著升高。
本研究开发的自动LG估计方法为睡眠呼吸暂停的内型分类提供了一种可扩展的非侵入性工具。通过准确模拟患者特异性通气控制，该方法支持呼吸暂停及更广泛临床背景下的个性化管理策略。
