Comparative Accuracy of Millimeter-Wave Radar and Polysomnography in Sleep Apnea Detection

Introduction

Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS) is a common sleep disorder characterized by apnea or hypoventilation resulting from complete or partial collapse of the upper airway during sleep [1]. According to existing research data, approximately 1 billion people worldwide are affected by obstructive sleep apnea syndrome. Among adults aged 30 to 69, up to 936 million (95% CI, 903–970) have varying degrees of mild to severe obstructive sleep apnea problems, with 425 million (95% CI, 399–450) adults in this age group suffering from moderate-to-severe obstructive sleep apnea. China has the highest prevalence rate of this disease [2]. The prevalence of OSAHS in China is 2% to 4%, and the prevalence among hypertensive patients is as high as 49.3%, with a prevalence of 83% among patients with persistent hypertension [3].

Patients with OSAHS may face serious cardiovascular disease problems due to the fact that OSAHS leads to disruptions in a variety of physiologic levels in the patient’s body. This includes blood gas abnormalities, respiratory arousal, and a surge in sympathetic nerve activity, which leads to a surge in blood pressure and heart rate [4]. Furthermore, each obstructive event is associated with reduced airflow, paradoxical abdominal and chest movements, and large fluctuations in intrathoracic pressure, with reduced ventilation leading to elevated carbon dioxide levels in the body and reduced oxygen saturation. Subsequently, compensatory mechanisms allow the patient to awaken from sleep, spontaneously opening the upper airway, which in turn terminates the apnea event. Oxygen saturation gradually returns to the baseline level prior to the apnea event; however, the upper airway may collapse again when the patient resumes sleep, and this process can continue throughout the sleep period [5]. Over time, patients with untreated OSAHS will be chronically exposed to nocturnal intermittent hypoxemia and hypercapnia, reoxygenation cycles, respiratory arousal, intrathoracic pressure fluctuations, and altered autonomic nervous system activity. Conversely, these exposures are thought to increase cardiovascular disease risk by several pathways, including blood pressure dysregulation, endothelial dysfunction, inflammation, oxidative stress, and altered metabolism [6]. In addition to this, OSAHS causes an increase in snoring, daytime sleepiness, disturbed sleep architecture, cognitive decline, and cardiovascular events [7,8]. Obesity, family history of genetic disorders, non-communicable diseases such as diabetes, and lifestyle changes such as sedentary lifestyle are also major factors contributing to OSAHS. Increased prevalence in these factors has contributed to the continued increase in the prevalence of OSAHS.

The main sleep monitoring device commonly used in clinical practice is polysomnography. Use of this device is based on guidelines published and continuously updated by the American Academy of Sleep Medicine (AASM) [9], which defines the technical requirements for obtaining and analyzing sleep study data, making polysomnography a decisive factor in the diagnosis of OSAHS. Therefore, polysomnography has become the gold standard for diagnosing OSAHS. The polysomnography sleep monitor is an instrument capable of recording a variety of indicators related to the patient’s sleep, including electroencephalography (EEG), electrooculography, electromyography, and other physiological signals. It provides detailed information on the sleep cycle, sleep staging, and sleep structure, enabling the analysis of sleep status and disorders. However, polysomnography has several limitations, such as complex equipment, cumbersome operation that may interfere with the subject’s sleep, high requirements for the testing environment, high cost, and poor patient compliance. These factors limit its widespread application in grassroots settings, such as at home or in community health centers. As a result, there has been growing interest in developing simpler and more accessible methods to address these issues, leading to the emergence of non-sensory millimeter-wave radar sleep monitors, which offer portable and user-friendly sleep apnea detection capabilities [10,11].

In contrast to polysomnography, the non-sensory millimeter-wave sleep monitor offers several advantages, including a lack of physical interference with sleep, no constraints, and ease of operation, which have gradually attracted increasing attention. This device utilizes millimeter-wave technology to monitor vital signs in a non-contact manner, enabling real-time monitoring and assessment of sleep quality [12]. The monitoring radar is capable of efficiently extracting cardiac motion signals and respiratory information from chest wall movements mixed with respiratory signals [13,14], allowing for the staging of rapid eye movement (REM) sleep and other sleep phases [15,16]. Additionally, the radar is specifically designed for sleep scenarios, with a mounting and antenna beam angle optimized for single-person measurements in single-bed settings, significantly reducing interference from the surrounding environment.

The purpose of this paper is to evaluate the accuracy and reliability of the non-sensory millimeter-wave sleep monitor by analyzing its monitoring principles, indices, and practical applications and comparing these parameters with those of polysomnography, with the aim of providing a reference for clinical sleep monitoring and diagnosis.

Material and Methods

EQUIPMENT:

The device used in the control group of this experiment was the simple polysomnography device, ApneaLink^TM Air, a commonly used polysomnography monitoring device. It is capable of recording the patient’s breathing pattern, oxygen saturation, heart rate, pulse pressure, and other key indicators through multiple sensors. These data provide a comprehensive record of the patient’s sleep status, which helps physicians make an accurate OSAHS diagnosis and assess its severity.

As shown in Figure 1, the sensorless millimeter-wave sleep monitor used in the experimental group is a device being tested in the laboratory that utilizes millimeter-wave radar technology. It enables sleep monitoring by capturing data such as the patient’s respiratory rate, heartbeat, and chest contour fluctuations, all without direct contact with the patient’s body [15]. The device emits high-frequency radar waves [18], and the receiver uses a novel parametric respiratory filter algorithm to accurately isolate the respiratory signal. This allows the device to obtain the subject’s respiratory data by monitoring the undulating movements of the chest contour, and subsequently calculate the respiratory rate. Figure 2 compares the normal and abnormal respiratory waveforms monitored by the radar.

SUBJECT SCREENING CRITERIA:

Subjects were included in the trial if they met all of the following criteria in the inpatient cardiovascular medicine ward at our center:

(1) Age between 30 and 60 years. (2) Symptoms of suspected OSAHS meeting any of the following 3 criteria: a. Men aged ≤60 with a BMI ≥30. b. Patients with hypertension characterized by significantly elevated diastolic blood pressure, refractory hypertension, non-arteritic/anti-arteritic hypertension, and elevated and refractory early morning blood pressure. c. Subjects with symptoms related to OSAHS, such as nocturnal snoring, irregular snoring, limb twitching, waking up from spasms and suffocation, vivid dreams, increased nocturnal enuresis, morning headache, dry mouth, daytime lethargy, fatigue, and cognitive symptoms such as reduced intelligence, memory impairment, and personality changes.

The Sleep Questionnaire* survey was used to assess the following screening criteria for OSAHS:

Additionally, patients with unstable conditions, such as severe chronic lung disease, severe heart disease, extreme obesity (BMI >45 kg/m2), neuromuscular disorders, and those deemed unsuitable for inclusion by the investigators (eg, patients with conditions too severe to cooperate, pregnant women, and children), were excluded. This was done to ensure that the study results were not influenced by the patients’ subjective feelings. Furthermore, patients with prior experience of polysomnography were excluded to maintain the accuracy of the experimental results. With the consent of the human trials Ethics Review Committee of Shanghai First People’s Hospital and the patients, 68 patients suspected of OSAHS were selected as experimental subjects. The baseline characteristics of the subjects, including gender, age, BMI, total sleep time, risk index, and oxygen desaturation index, are shown in Table 1.

SLEEP ASSESSMENT:

From 8: 00 PM Beijing time to 8: 00 AM the following day (a total of 12 hours), each subject’s sleep was monitored using both the ApneaLink™ Air polysomnography device and the millimeter-wave radar monitoring device. The entire monitoring process took place in the cardiovascular ward of our center, with professional staff responsible for the installation, calibration, and placement of the devices. The results from both the ApneaLink™ Air polysomnography and the millimeter-wave radar monitoring devices were recorded independently by 2 professional staff members, ensuring that each party was unaware of the other’s data during the recording process. The recorded results underwent automated analysis by machine, followed by manual refinement and corrective analysis.

STATISTICAL METHODS:

Sensitivity and specificity were calculated to evaluate the ability of the millimeter – wave radar sleep monitor to correctly identify true – positive and true negative cases, respectively.

The formula used for sensitivity calculation was as follows:

Specificity was calculated to evaluate the ability of the millimeter-wave radar sleep monitor to correctly identify true negative cases. The formula used for specificity calculation was as follows:

The specificity was then compared between the 2 devices using the chi-square test, with significance set at P<0.05.

Results

Sixty-eight subjects who met the inclusion criteria at our center participated in this experiment. Each subject underwent sleep monitoring using both polysomnography (ApneaLink™ Air) and the millimeter-wave radar sleep monitor. The apnea-hypopnea index (AHI) was chosen as the primary observational index to assess the sensitivity and specificity of the monitoring devices. The severity of OSAHS was classified according to the criteria shown in Table 2, and based on this classification, subjects were divided into 3 subgroups according to their results.

Participants were instructed to avoid caffeinated beverages, daytime napping, and alcohol for the 24 hours prior to data collection. Subsequently, all-night polysomnography monitoring was conducted using the ApneaLink™ Air device to assess nasal airflow, blood pressure, body position, oxygen saturation (SpO2), and thoracic and abdominal movements. The sleep cycles were analyzed and scored by a professional sleep physician. In the sleep laboratory, subjects underwent both radar and polysomnography monitoring simultaneously. Apnea was defined as a reduction in airflow of at least 90% lasting for at least 10 seconds, while hypoventilation was defined as a reduction in airflow of at least 30% lasting for at least 10 seconds, accompanied by a decrease in SpO2 of at least 3%, or microarousals [17]. At the conclusion of the assessment, all data were reviewed by a professional sleep specialist, and the results were manually interpreted by 2 certified specialists according to the AASM Sleep and Related Events Scoring Manual. The final monitoring results are provided in Table 3.

The sensitivity of the millimeter-wave radar sleep monitor in diagnosing OSAHS was 92.7%, and the specificity was 84.6%, with an accuracy rate of 91.18%. The P-value for these results was greater than 0.05. The Kappa coefficient was 0.731 (P<0.001), indicating substantial agreement between the 2 devices. For moderate-to-severe OSAHS, the Kappa coefficient was 0.831 (P<0.001), demonstrating a high level of agreement in diagnosing this subgroup of patients. These results suggest that the millimeter-wave radar sleep monitor demonstrated a high level of consistency and reliability in diagnosing moderate-to-severe OSAHS.

Figure 3 illustrates the receiver operating characteristic (ROC) curves of the millimeter-wave radar device for diagnosing patients with varying severities of OSAHS, using AHI diagnostic thresholds of 5 and 15 events/hour, respectively. The areas under the curves (AUC) were 0.89 (95% CI: 0.88 to 0.90) and 0.97 (95% CI: 0.94 to 0.96), respectively. These findings indicate that the millimeter-wave radar sleep monitor performs robustly in diagnosing OSAHS, with excellent diagnostic performance for moderate-to-severe cases.

To better contextualize these findings, Table 4 compares the features and performance of this study’s millimeter-wave radar sleep monitor with other recent technological advances in apnea detection, including integrated Micro-Electro-Mechanical Systems (iMEMS)-based approaches. As shown in the table, the millimeter-wave radar sleep monitor offers several advantages over traditional polysomnography and emerging technologies, including non-contact monitoring, high accuracy in detecting OSAHS, and the ability to perform in real-world settings outside of sleep laboratories. These superior qualities highlight the potential of the radar-based monitor as a promising tool for widespread OSAHS screening and diagnosis.

Discussion

The millimeter-wave radar sleep detector demonstrates high agreement with polysomnography in monitoring sleep breathing (Figure 4), with sensitivity and specificity of 92.7% and 84.6%, respectively. These results align with previous studies on millimeter-wave radar for vital sign monitoring [18], confirming its reliability in reflecting sleep breathing patterns.

A key advantage of this technology is its suitability for primary care settings, where specialized equipment and trained personnel are often unavailable. Unlike polysomnography, which requires controlled environments and is prone to data loss due to sensor displacement (Figure 5), the radar system enables non-contact, continuous monitoring. It is unaffected by environmental factors (eg, light, temperature) and can penetrate bedding, making it ideal for home-based or long-term monitoring [19]. Additionally, its real-time data upload to cloud platforms addresses challenges related to medical resource distribution and professional shortages [20,21].

Compared with sensory devices, millimeter-wave radar offers significant non-invasive benefits, including reducing cross-infection risks in clinical settings such as ICUs [22]. Its portability and convenience also make it suitable for home use, as demonstrated by studies comparing ultra-wideband bioradar with polysomnography [23,24]. These studies confirmed high patient acceptance and diagnostic consistency, supporting its use for initial OSAHS screening.

However, the present study has limitations. First, we evaluated only 1 specific radar device, and results may not generalize to all market products. Second, the radar does not monitor EEG signals, relying instead on thoracic and abdominal movements to infer respiratory effort-related arousals [25]. Future research should explore combining sensory and non-sensory methods for more comprehensive sleep assessments.

Conclusions

This study demonstrates that the sensorless millimeter-wave radar sleep monitor achieves high diagnostic accuracy for OSAHS, with a sensitivity of 92.7%, specificity of 84.6%, and overall accuracy of 91.18%. The device shows excellent agreement with polysomnography, particularly in diagnosing moderate-to-severe OSAHS (Kappa coefficient: 0.831, P<0.001). These results highlight its potential as a reliable alternative to polysomnography for OSAHS screening and monitoring.

The radar’s non-contact design, portability, and ability to operate in home environments address key limitations of traditional polysomnography, such as patient discomfort and data loss due to sensor displacement. Additionally, its real-time data upload capability supports remote diagnosis, making it particularly valuable in resource-limited settings.

Future studies should focus on validating these findings in larger, more diverse cohorts and exploring the integration of multi-modal data (eg, combining radar with EEG or iMEMS sensors) to further enhance diagnostic accuracy. Additionally, long-term home-based monitoring studies are needed to evaluate the device’s performance in real-world scenarios and its impact on patient outcomes.