Research Article
Development and Validation of Algorithms for Sleep Stage Classification and Sleep Apnea/Hypopnea Event Detection Using a Medical-Grade Wearable Physiological Monitoring System
@INPROCEEDINGS{10.1007/978-3-031-06368-8_12, author={Zhao Wang and Zhicheng Yang and Ke Lan and Peiyao Li and Yanli Hao and Ying Duan and Yingjia She and Yuzhu Li and Zhengbo Zhang}, title={Development and Validation of Algorithms for Sleep Stage Classification and Sleep Apnea/Hypopnea Event Detection Using a Medical-Grade Wearable Physiological Monitoring System}, proceedings={Wireless Mobile Communication and Healthcare. 10th EAI International Conference, MobiHealth 2021, Virtual Event, November 13--14, 2021, Proceedings}, proceedings_a={MOBIHEALTH}, year={2022}, month={6}, keywords={Sleep stage classification Sleep apnea/hypopnea event Apnea-hypopnea index Physiological monitoring Wearable system Polysomnography}, doi={10.1007/978-3-031-06368-8_12} }- Zhao Wang
Zhicheng Yang
Ke Lan
Peiyao Li
Yanli Hao
Ying Duan
Yingjia She
Yuzhu Li
Zhengbo Zhang
Year: 2022
Development and Validation of Algorithms for Sleep Stage Classification and Sleep Apnea/Hypopnea Event Detection Using a Medical-Grade Wearable Physiological Monitoring System
MOBIHEALTH
Springer
DOI: 10.1007/978-3-031-06368-8_12
Abstract
Sleep is critical to the overall health of humans. Polysomnography (PSG) is the current gold standard for measuring sleep and diagnosing sleep-related breathing disorders. However, this method is labor-intensive, time-consuming, and confined to a sleep laboratory. In this paper, we leverage algorithms for sleep stage classification and sleep apnea/hypopnea event detection by using signals from single-lead electrocardiograph (ECG) and respiration. To validate the accuracy of the above two algorithms, two independent validation studies were conducted using a medical-grade wearable monitoring system to collect physiological data from patients in both clinical and home settings. In the validation study of sleep stage classification, the average accuracy of our four-class stage classification using the bi-directional long short-term memory (BLSTM) method is 77.83% on our in-house dataset of 30 enrolled patients. In the experiments of sleep apnea screening, the two-level apnea-hypopnea index (AHI) classification reports the overall accuracies of 96.67% and 91.43% in clinical and home environments, respectively. The results showed that the sleep analysis algorithms presented in this paper have good performance in both sleep stage classification and sleep event detection, either in clinical scenario and home settings, indicating that our device can be used along with the two algorithms for sleep analysis.
