Syed Doha Uddin;Md. Shafkat Hossain;Shekh M. M. Islam;Victor Lubecke
{"title":"Heart Rate Variability-Based Obstructive Sleep Apnea Events Classification Using Microwave Doppler Radar","authors":"Syed Doha Uddin;Md. Shafkat Hossain;Shekh M. M. Islam;Victor Lubecke","doi":"10.1109/JERM.2023.3317304","DOIUrl":null,"url":null,"abstract":"Obstructive Sleep Apnea (OSA) is the most common type of sleep disorder that consists of multiple episodes of partial or complete closure (apnea, hypopnea) of the upper airway during sleep and underdiagnosed problems as there is no reliable portable in-home sleep monitoring system. Doppler radar system is gaining attention as an in-home sleep monitoring system due to its non-contact and unobtrusive form of measurement. Prior research on Radar-based sleep monitoring systems mostly focused on distinguishing apnea and normal breathing patterns using radar-reflected signal amplitude that can't distinguish accurately apnea and hypopnea events. Apnea and hypopnea events were distinguished using effective radar cross-section (ERCS) for short-scale study and ERCS changes with sleeping postures and so on. In this work, we proposed a heart rate variability-based robust feature extraction technique to distinguish different sleep disorder events such as apnea, hypopnea, and normal breathing. HRV-based feature extraction technique was employed on ten consented OSA participants' clinical studies to find a distinguishable feature known as the power of the low-frequency band (0.04-0.15 Hz) and high-frequency band (HF) (0.15-0.4 Hz). The extracted hyper-feature (HF and LF) was then integrated with the traditional Machine learning classifiers (ML) including k-nearest neighbors (KNN), support vector machine (SVM), and random forest. SVM outperformed other classifiers with an accuracy of 97% for distinguishing different OSA events that also supersedes other reported results (ERCS). The proposed method has several potential applications including in-home sleep monitoring, OSA severity detection, respiratory disorder detection, and so on.","PeriodicalId":29955,"journal":{"name":"IEEE Journal of Electromagnetics RF and Microwaves in Medicine and Biology","volume":"7 4","pages":"416-424"},"PeriodicalIF":3.0000,"publicationDate":"2023-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Journal of Electromagnetics RF and Microwaves in Medicine and Biology","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10265037/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
Obstructive Sleep Apnea (OSA) is the most common type of sleep disorder that consists of multiple episodes of partial or complete closure (apnea, hypopnea) of the upper airway during sleep and underdiagnosed problems as there is no reliable portable in-home sleep monitoring system. Doppler radar system is gaining attention as an in-home sleep monitoring system due to its non-contact and unobtrusive form of measurement. Prior research on Radar-based sleep monitoring systems mostly focused on distinguishing apnea and normal breathing patterns using radar-reflected signal amplitude that can't distinguish accurately apnea and hypopnea events. Apnea and hypopnea events were distinguished using effective radar cross-section (ERCS) for short-scale study and ERCS changes with sleeping postures and so on. In this work, we proposed a heart rate variability-based robust feature extraction technique to distinguish different sleep disorder events such as apnea, hypopnea, and normal breathing. HRV-based feature extraction technique was employed on ten consented OSA participants' clinical studies to find a distinguishable feature known as the power of the low-frequency band (0.04-0.15 Hz) and high-frequency band (HF) (0.15-0.4 Hz). The extracted hyper-feature (HF and LF) was then integrated with the traditional Machine learning classifiers (ML) including k-nearest neighbors (KNN), support vector machine (SVM), and random forest. SVM outperformed other classifiers with an accuracy of 97% for distinguishing different OSA events that also supersedes other reported results (ERCS). The proposed method has several potential applications including in-home sleep monitoring, OSA severity detection, respiratory disorder detection, and so on.