Spotlight on Sleep: Beyond the Polysomnogram—Modern Approaches to Diagnosing Sleep Apnea

Polysomnography (PSG), performed within a sleep laboratory and fully monitored by a sleep technician, has been established as the gold standard for diagnosing obstructive sleep apnea (OSA).¹ PSG captures a wide range of physiologic measures, including EEG, electro-oculography (EOG), EMG, respiratory effort, airflow, oxygen saturation, electrocardiography (ECG), and limb movements, enabling the identification of various sleep disorders in addition to OSA.² Over the past decade, technologic advancements and evolving clinical guidelines have contributed to the growing acceptance of home sleep apnea testing (HSAT) as a viable, convenient, and cost-effective alternative for diagnosing OSA in select populations.³ Most HSAT devices are portable, and they typically measure respiratory parameters such as airflow, respiratory effort, and oxygen saturation, with some models incorporating sleep staging capabilities.

History and Evolution of HSAT

The concept of portable sleep monitoring emerged in the 1980s with simplified devices that focused primarily on respiratory measures.⁴ In 1994, the American Sleep Disorders Association, the historical predecessor of the American Academy of Sleep Medicine (AASM), introduced a classification system for sleep apnea testing which remains in use today.^4-6 The early iterations of these portable sleep monitoring devices were limited by issues of reliability, accuracy, and lack of standardization, which hindered their widespread acceptance. The landscape began to shift with the introduction of modern HSAT devices which incorporated advanced sensor technologies and automated data analysis along with clinical validation trials. A major milestone occurred in 2007 when the AASM clinical practice guidelines endorsed HSAT as an appropriate diagnostic tool in the setting of moderate to high likelihood of OSA in the absence of substantial comorbidities.⁶ The Centers for Medicare & Medicaid Services decision in 2008 to cover the use of HSAT for OSA diagnosis further accelerated its use in clinical practice in the United States.⁷ Subsequent randomized clinical trials found that an integrated home-based portable monitoring strategy for diagnosing and treating moderate to severe OSA is noninferior to a laboratory-based approach in terms of key outcomes, such as therapy acceptance, adherence, and functional improvement.^8-10 These findings have contributed to the widespread adoption of flow-based HSAT over the past decade, particularly in uncomplicated cases of suspected OSA. 

Types of HSAT Devices and the Meaning of FDA Clearance

HSAT devices are typically categorized on the basis of the number and type of physiologic channels they monitor, as outlined by the AASM⁵:

• Type 2 devices measure a minimum of 7 channels, including EEG, EOG, chin EMG, ECG or heart rate, airflow, respiratory effort, and oxygen saturation. 
• Type 3 devices measure a minimum of 4 channels, including airflow, respiratory effort, heart rate, and oxygen saturation. 
• Type 4 devices record 1 or 2 channels, typically oxygen saturation or airflow.

The Food and Drug Administration (FDA) classifies HSAT devices as class 2 medical devices, indicating a moderate level of risk. FDA clearance of HSAT devices is commonly granted through the 510(k) premarket notification process, which requires manufacturers to demonstrate that their new device meets the FDA’s safety standards and its performance is “substantially equivalent” to an already legally marketed predicate device. FDA clearance does not necessarily mean the device has undergone rigorous clinical trials to prove its accuracy or efficacy. Therefore, clinicians must carefully evaluate the performance of each HSAT device and discern its limitations before integrating it into clinical diagnostic pathways.¹¹

Advancements in HSAT: PPG/PAT and Non-PPG Technologies

In recent years, HSAT has evolved beyond conventional flow-based type 3 devices, incorporating innovative, artificial intelligence–powered sleep technologies for diagnosing OSA in ambulatory settings.¹² Since 2019, the FDA has cleared 12 novel wearable HSAT devices or software as a medical device (SaMD) for this purpose. SaMD refers to software designed for medical purposes that can operate on general-purpose platforms. Examples include SleepImage (MyCardio, Denver, CO) and EnsoSleep PPG (photoplethysmography) (EnsoData, Madison, WI), which can be paired with FDA-cleared pulse oximeters to facilitate OSA diagnosis.^13,14

These emerging wearable technologies can be broadly categorized into PPG/peripheral arterial tonometry (PAT)–based and non–PPG-based technologies. Novel PPG/PAT-based wearables have a rapidly expanding role in OSA diagnostics. These systems capture physiologic signals, such as PPG/PAT, oxygen saturation, pulse rate, and accelerometry, to detect respiratory events, identify arousals, and derive key sleep metrics (eg, apnea-hypopnea index [AHI]). Novel PAT-based wearables leverage proprietary signal processing algorithms to derive PAT signals from PPG, eliminating the need for a dedicated PAT probe. On the other hand, non–PPG-based devices use alternative sensing modalities, including mandibular movement, chest and abdominal respiratory effort, and acoustic biosignals, to detect OSA. Whereas these novel non–PPG-based approaches offer valuable physiologic insights, including derived flow, respiratory effort, and respiratory arousals, they often need to be paired with pulse oximeters to provide essential oxygenation information. These advancements have expanded the capabilities of home-based OSA diagnostics, enhancing accessibility and broadening the scope of ambulatory sleep assessment.

Figure. Images of photoplethysmography (PPG)/peripheral arterial tonometry (PAT)–based and non-PPG–based at-home sleep apnea testing devices. SleepImage Ring (MyCardio, Denver, CO) (A); EnsoSleep PPG (EnsoData, Madison, WI) (B); NightOwl (ResMed, San Diego, CA) (C); Somfit (Compumedics Limited, Abotsford, Australia) (D); Belun Ring (Belun Technology Company Limited, Hong Kong) (E); SANSA (Huxley Medical, Atlanta, GA) (F); TipTraQ (PranaQ, Singapore) (G); Sunrise (Sunrise, Namur, Belgium) (H); Wesper Lab (Wesper, New York, NY) (I). 

PPG/PAT-Based Sleep Technologies for OSA Diagnosis

Table 1 presents a comparison overview of PPG/PAT-based devices, highlighting their sensors, channels, features, limitations, and performance at an AHI cutoff of 15 events/h (using AHI3%, AHI4%, or both). The performance statistics of these SaMD or devices were compiled into a single table for reference, but direct comparison of their performance based on these statistics should be avoided, because validation studies were conducted in diverse research environments with varying sample sizes, populations, and eligibility criteria. Directly comparing the performance metrics could lead to misleading conclusions.

SleepImage SaMD

The SleepImage SaMD (Figure, A) uses PPG sensors on the proximal phalanx or fingertip with a rule-based algorithm for derivation of AHI from cardiopulmonary coupling.^12,13 The SleepImage SaMD scores sleep stages in a nonconventional fashion based on the cardiopulmonary coupling frequency. Despite prior publications validating its performance by utilizing ECG or PPG from PSG, few validation studies using an independent SleepImage sensor exist. The SleepImage SaMD is FDA-cleared for individuals aged >2 years and uses 3-dimensional spectrogram technology to detect central sleep apnea (CSA). Its performance in children and adolescents and CSA detection accuracy warrant further validation. 

EnsoSleep PPG SaMD 

The EnsoSleep PPG SaMD (formerly, Aurora) (Figure, B) uses machine learning (ML) and deep learning (DL) algorithms to analyze PPG signals and can pair with FDA-cleared pulse oximeters to derive AHI and 4-stage classification (wake, REM, and light and deep sleep).¹⁴ EnsoSleep was cleared by the FDA in 2024, but no peer-reviewed studies evaluating its performance have been published. This SaMD was validated in a white paper using the Checkme O₂ pulse oximeter (Shenzhen Viatom Technology Co., Ltd., Shenzhen, China). Further performance validation may be warranted if EnsoSleep PPG is to be coupled with other FDA-cleared ring or fingertip pulse oximeters, because potential biases may arise when pulse oximeters with different hardware configurations are used in conjunction with the software. 

NightOwl

NightOwl (ResMed, San Diego, CA) (Figure, C) is the first FDA-cleared probeless PAT-based device. Using a disposable compact sensor, comparable in size to a fingertip, NightOwl uses proprietary ML algorithms to estimate AHI and sleep metrics, and provides both AHI3% and AHI4% values with consistent accuracy. Its performance in detecting central events and sleep stages requires further validation.^12,15

ANNE Sleep

The ANNE Sleep (Sibel Inc., Niles, IL) system consists of 2 integral units: a chest piece at the suprasternal notch and a limb unit at the fingertip. It was cleared by the FDA as a PAT device and is integrated with DL-powered algorithms for sleep–wake classification. The device incorporates multiple channels, including PAT, body position, snoring, chest movement, SpO2, ECG, and pulse transit time measurements. Despite these capabilities and solid performance at the AHI4% cutoff of 15 using manual scoring, it lacks validated performance data for AHI3% and automatic sleep staging outcomes.^12,16

Somfit

Somfit (Compumedics Limited, Abbotsford, Australia) (Figure, D) captures EEG/EOG, accelerometry, snoring, and PAT signals from the forehead and uses rule-based algorithms for AHI estimation and DL algorithms for 5-stage sleep staging (wake, REM, N1, N2, and N3). According to its FDA 510(k) premarket notification, it offers superior performance compared with the WatchPAT-300 (ZOLL Itamar, Caesarea, Israel). However, no peer-reviewed publications are available for its AHI performance, and it lacks the capability to monitor respiratory effort and detect CSA.^12,17

Belun Ring

The Belun Ring (Belun Technology Company Limited, Hong Kong) (Figure, E), also known as the Belun Sleep System BLS-100, is an FDA-cleared PPG-based device worn on the index finger that uses dual proprietary DL algorithms to assess AHI and 3-stage sleep classification (wake, REM, and non-REM sleep). It also offers comprehensive pulse rate variability metrics, including both time- and frequency-domain pulse rate variability measures. The second-generation Belun Ring algorithms exhibit satisfactory performance in identifying moderate to severe OSA using AHI4% across individuals with different races and ethnicities and in populations with high body mass index. Its current version does not provide AHI3%, body position, or CSA detection.^12,18

SANSA

SANSA (Huxley Medical, Atlanta, GA) (Figure, F) incorporates PPG, accelerometry, and a single-lead ECG sensor within a chest-based system, using ML and rule-based algorithms for sleep–wake detection and respiratory event analysis. The all-in-one chest sensor suite captures PPG, SpO2, and ECG signals, but may require chest shaving for some users. The device was cleared by the FDA in 2024. It offers both AHI3% and AHI4% at cutoffs of 5, 15, and 30.^19,20

TipTraQ

The TipTraQ (PranaQ, Singapore) (Figure, G) fingertip PPG-based system, the latest addition to the novel OSA-detecting wearables, uses dual DL algorithms for sleep staging and respiratory event detection. It offers AHI3% and AHI4% at cutoffs of 5, 15, and 30, but does not provide CSA detection.^21,22

Non-PPG-Based Sleep Technologies for OSA Diagnosis

Table 2 compares non–PPG-based devices, emphasizing their unique sensing methods, channels, features, limitations, and performance.

Sunrise

Sunrise (Sunrise, Namur, Belgium) (Figure, H) is an innovative mandibular movement–based system that uses accelerometry and gyroscope sensors placed on the chin in the mentolabial sulcus to monitor jaw movements during sleep, which correlate with respiratory effort and arousals. The FDA-cleared second-generation Sunrise device also integrated thermistor flow and PPG into the chin sensor. However, its ML algorithms only extract data from accelerometry and the gyroscope, without incorporating the thermistor and SpO2 data. In the published Sunrise studies, researchers primarily assessed its respiratory disturbance index with customized optimal cutoffs based on post hoc analysis (not using conventional cutoffs of 5, 15, and 30), and did not present details on the Sunrise AHI3% or AHI4% data using the standard AHI cutoffs of 5, 15, and 30. Shaving of some facial hair over the mentolabial sulcus may be needed for people with beards.^12,23

Wesper Lab

Wesper Lab (Wesper, New York, NY) (Figure, I) measures respiratory effort through dual wireless patches placed on the chest and abdomen to derive respiratory effort–derived flow. This FDA-cleared wearable device can also record body position and can be paired with the Checkme O₂ Max pulse oximeter to provide SpO2 data. Published studies have involved small cohorts. A recent study reported the correlation coefficient for AHI3%, but lacks detailed statistical data, including sensitivity and specificity. Further investigation is needed to validate AHI4% and total sleep time (TST) accuracy. Although Wesper Lab can offer PPG and SpO2, these data are not incorporated into its algorithms.^12,24

AcuPebble

The AcuPebble (Acurable, London, UK) uses a proprietary piezoelectric acoustic sensor placed on the anterior neck to detect breathing, heart sounds, movements, and snoring. It estimates respiratory flow and derives AHI via acoustic signals. Although the manufacturers claim the device has the capability to estimate “SpO2 drops” through acoustic signals, its physiologic plausibility remains poorly understood. Further validation is required for TST, sleep staging, and CSA detection. The SA100 model without a pulse oximeter was cleared by the FDA for OSA home screening; the Ox100, coupled with a fingertip pulse oximeter, was cleared by the FDA for adult OSA diagnosis.^12,25

BresoDX1

BresoDX1 (BresoTEC, Toronto, Canada), powered by DL algorithms, incorporates multiple sensing modalities, including acoustic signals, PPG, and tracheal motion analysis. It estimates airflow through tracheal breath sounds and captures respiratory effort from tracheo-sternal motion patterns. The algorithm integrates pulse oximetry data for enhanced respiratory event detection. Published data on TST, sleep stages, and AHI4% are lacking, and the device does not distinguish between central and obstructive events. BresoDX1 is not yet commercially available.^12,26

Application of HSAT in Neurology Practices

HSAT is increasingly recognized as a valuable diagnostic tool within neurology practices, where sleep disorders frequently coexist with neurologic conditions, such as stroke, epilepsy, and neurodegenerative diseases.²⁷ The high prevalence of OSA among individuals with these conditions underscores the need for efficient screening and timely diagnosis. Studies have shown that early identification and treatment of OSA in people with stroke or transient ischemic attack can improve functional outcomes and reduce the risk of recurrent vascular events.²⁸ A recent study by Sindorf et al²⁹ evaluated the use of the ANNE Sleep device to detect sleep apnea in patients with stroke undergoing acute inpatient rehabilitation. This highlights the potential of OSA-detecting wearables for early OSA identification in poststroke individuals which could improve their recovery and long-term outcomes. However, HSAT may not be suitable for people with complex sleep disorders, substantial comorbidities (eg, nocturnal seizures, severe underlying lung disease, severe dementia, severe cognitive issues), or suspected CSA, as seen in conditions like congestive heart failure and atrial fibrillation.³⁰

To integrate HSAT effectively into patient care, neurology practices can integrate the following strategies:

• Screening: Use validated tools such as the STOP-Bang questionnaire to identify individuals at risk for OSA.^31,32 The STOP-Bang questionnaire is availbale at www.stopbang.ca.
• HSAT: Use HSAT for individuals with a high pretest probability of moderate to severe OSA and no major comorbidities that would otherwise require in-laboratory PSG.
• PSG: Refer individuals with inconclusive or negative HSAT results or suspected CSA for in-laboratory PSG.

Some novel PPG-based devices can provide valuable data, such as hypoxic burden and heart rate variability data, that enhance OSA severity stratification and improve outcome prediction. The AASM launched a focused updated task force on OSA diagnostic testing in 2023. A new clinical practice guideline, expected by the end of 2025, will likely offer updated guidance on the integration of advanced HSAT wearable technologies into clinical practice. 

Conclusion

The evolution from in-laboratory PSG alone to the inclusion of flow-based HSAT and, more recently, wearables reflects considerable advancements in sleep technologies and a move toward patient-centered care. HSAT has proven to be a valuable diagnostic tool for OSA, especially within neurology practices, where sleep disorders are common comorbidities. HSAT increases accessibility, enables testing in a familiar home environment, and offers cost-effectiveness. Proper patient selection and interpretation of HSAT results by a sleep specialist remain essential for maintaining diagnostic accuracy. As technology continues to evolve, the role of HSAT in sleep medicine is poised to expand, further improving access to care and enhancing outcomes.