Abstract
Background
Recently, sleep wearables have experienced a boom, particularly devices worn on the finger, such as smart rings, which have garnered growing attention. However, there currently exists no review that comprehensively and systematically evaluates the accuracy of these devices.
Main body
This scoping review evaluates 11 finger-worn devices against polysomnography (PSG) across 28 articles from inception to April, 2025, assessing sleep staging (primarily healthy participants) and obstructive sleep apnea (OSA) detection (clinical cohorts) via heterogeneous sensors and algorithms. Current evidence in our study demonstrates efficacy in sleep/wake classification (pooled accuracy = 87%, 95%CI = 86%~89%) and severe OSA screening, but reveals persistent challenges in multi-stage sleep analysis (pooled accuracy for light/deep/rapid eye movement (REM) sleep = 0.65, 0.81, 0.74, respectively) and mild OSA detection. All devices, except one, demonstrated higher accuracy at an apnea-hypopnea index (AHI) threshold of 30 than at thresholds of 5 or 15.
Conclusions
Finger-worn devices may be used for longitudinal sleep/wake tracking in healthy adults or as a triage tool for severe OSA, but should not replace PSG for diagnosis of mild OSA or detailed sleep architecture assessment. Achieving PSG-comparable accuracy in multi-stage sleep analysis and mild OSA detection requires addressing algorithmic transparency, population diversity, and standardized methodologies.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12967-026-08214-x.
Keywords: Finger-worn devices, Polysomnography, Sleep stage, Sleep apnea
Background
Sleep, a modifiable determinant of health, plays a pivotal role in maintaining physiological homeostasis and cognitive function. Robust evidence links adequate sleep duration and quality to metabolic regulation, immune competence, and cardiovascular health [1, 2], while sleep deficiencies are associated with impaired neurobehavioral performance, mood disorders, and increased risk of chronic disease [3]. Sleep can be assessed subjectively by using questionnaires or objectively through measurement of physiological sleep parameters. Although polysomnography (PSG) remains the gold standard for evaluating sleep architecture and respiratory events, its clinical utility is limited by technical complexity, high cost, accessibility constraints, and patient discomfort from multi-sensor attachments in unfamiliar environments [4]. These limitations necessitate multi-night recordings to counter the “first-night effect,” yet such approach remains impractical for longitudinal or community-based studies.
The advent of consumer-grade wearable technologies has revolutionized sleep monitoring, offering scalable, cost-effective alternatives for both clinical and ambulatory settings, including wrist-worn devices, rings, and headbands [5]. Wrist-worn actigraphy, historically the dominant form of wearable sleep tracking, utilizes accelerometry to infer sleep-wake patterns but exhibits limited sensitivity to subtle movements characteristic of light sleep or transient arousals. Recent innovations in miniaturized multi-sensor systems have shifted focus toward alternative anatomical placements, with finger-worn devices emerging as a promising platform. The wrist actigraphy devices typically only detect significant body movements involving the forearm, while a device placed on the finger may be able to detect much smaller movements from the hand and finger that occur during wakefulness and light stages of sleep. Therefore, this location provides better signal quality and is less affected by noise, resulting in more accurate detection of sleep stages. Traditional pulse oximeters use photoplethysmography (PPG) sensors, which are also commonly placed on fingertips, toes, or earlobes. These devices capitalize on the finger’s rich vascularity and fine motor activity, enabling simultaneous acquisition of PPG, accelerometry, and pulse oximetry data. Such multimodal integration enhances detection of sleep stages and respiratory events, particularly obstructive sleep apnea (OSA), a globally prevalent disorder with significant cardiometabolic morbidity [6].
Early-generation finger-worn devices demonstrated modest accuracy in sleep staging. For instance, the Oura Ring Gen1 achieved 65% agreement with PSG for light sleep, 51% for deep sleep, and 61% for REM sleep [7]. Subsequent iterations, such as the Oura Ring Gen2, reported marked improvements (80% light sleep, 91% deep/REM sleep accuracy) through advanced algorithms and sensor fusion [8]. The two-stage classification (wake/sleep) also demonstrated higher yet inconsistent accuracy, ranging from 0.85 [9] to 0.96 [8], depending on specific devices used. Furthermore, the identification of OSA severity varied across different devices, such as the Belun Ring [10], Circul Ring [11], and RingConn Smart Ring [12]. These disparities underscore the need for rigorous, standardized review for the performance of finger-worn technologies against PSG benchmarks.
Despite expanding research interest, no systematic synthesis exists that evaluates finger-worn devices in sleep staging and obstructive sleep apnea (OSA) detection. This review addresses this critical gap by comprehensively appraising recent technological advancements, examining their accuracy against polysomnography, methodological rigor of validation studies, and clinical implications.
Methods
The study was registered on PROSPERO prospectively (ID CRD42024605972) and was performed in compliance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses).
Search strategy
We performed a systematic search including the databases of PubMed, Web of Science, and Scopus to pinpoint relevant papers published to 8 April 2025. The search was conducted using both MeSH (medical subjects heading terms) and free-text words such as ‘Finger-worn device,’ ‘Wearable ring,’ ‘Sleep stages,’ ‘Sleep Apnea Syndromes,’ and ‘Detect,’ with appropriate use of Boolean operators (AND, OR) to combine search terms. We limited the search to English language articles. Additionally, the reference lists of relevant articles were searched to identify any additional articles. We also reviewed the official websites of major smart ring manufacturers for any available validation study reports and examined clinical trial for completed but potentially unpublished studies. The complete search words and search strategy are provided in Supplementary Table 1.
Selection process
Based on the search, the study selection process was conducted by two independent investigators (YXJ and JWX). Any disagreements were addressed by a third reviewer (JW). Supplementary Table 2 presents the inclusion and exclusion criteria for study selection. The inclusion criteria were the publications that involved population using any finger-worn device, that reported the performance of detecting sleep stages and sleep apnea, and that incorporated PSG as reference. Letters, comments, conference abstracts and review articles were excluded. If the study was published repetitively, the most recent study or the study with larger sample size was selected. Exclusion criteria also included studies that utilized non-ring devices (e.g., Watch, Fitbit, or WHOOP). The study selection process is shown in Fig. 1.
Fig. 1.
PRISMA flow diagram
Data extraction and analysis
Data extraction involved study characteristics (e.g., first author, year of publication, location, research setting, and duration), participants demographics (e.g., sample size, mean age, sex, and health status), and device-specific attributes (e.g., type, algorithm, feature selection, and epoch length). Validity refers to the extent to which a measurement tool reflects the actual characteristic it is intended to assess [13]. In this study, we mainly focused on criterion validity, which examines criterion-related validity by benchmarking finger-worn devices against the gold-standard polysomnography (PSG). Therefore, we documented the diagnostic performance outcomes in terms of accuracy, sensitivity, specificity, balanced accuracy and Cohen’s kappa. The detailed indices are as follows:
Accuracy: the proportion of all test results (both correct and incorrect) that are correctly classified.
![]() |
Sensitivity: the ability of a test to correctly identify individuals with the condition (true positives).
![]() |
Specificity: the ability of a test to correctly exclude individuals without the condition (true negatives).
![]() |
Balanced accuracy: indicates the mean of wake epochs correctly classified and sleep epochs correctly classified, calculated as the mean of sensitivity and specificity.
Cohen’s kappa (κ): is calculated to evaluate the agreement values relative to that which could be expected due to chance.
We conducted a meta-analysis to provide summary measures for sleep stages. Given the heterogeneity across these studies, we chose random effects models in the meta-analysis. The analysis was conducted by implementing R statistical software (version 4.3.2) and the “metagen” R packages (version 4.3.2).
Evaluation of quality
The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool was used to evaluate the quality of the primary studies [14]. The risk of bias included four main domains: participants, index test, reference standard, and flow and timing. In every domain, risk was classified as ‘high’ if the criteria did not meet at least one of the signaling criteria, ‘unclear’ if information was unavailable for one or more criteria, and ‘low’ if all signaling criteria were fully satisfied. If a study is judged as “low” on all domains relating to bias then it is appropriate to have an overall judgment of “low risk of bias” for that study. If a study is judged “high” or “unclear” on one or more domains then it may be judged “at risk of bias”. Quality assessment was performed by two independent investigators (YXJ and JWX), with any differences resolved through discussion and consensus.
Results
Study characteristics
We involved 28 articles including 2426 participants in the review, with a subset analysis of 14 articles [7–9, 15–25] explored the performance of finger-worn devices in sleep stages, and 13 articles [10–12, 26–35] revealed the detection on sleep apnea. One article [36] examined the above two aspects (Tables 1 and 2; Fig. 2). Supplementary Table 3 shows the basic characteristics of included studies across 28 articles. The systematic analysis encompassed researches spanning 2018–2025, with 20 studies emerging in or after 2022. In terms of participant demographics, the sleep stage studies predominantly involved adults aged 18–50 years (12/15) and participants with normal weight (8/15). In contrast, the sleep apnea studies more frequently included participants over 50 years old (9/14) and those who were overweight or obese (12/14). For disease severity in apnea-focused studies, the majority of studies simultaneously reported the AHI of 5, 15, 30 (11/14). Notably, only three studies targeted children and adolescents in the above two aspects. 15 studies employed 30-second epochs for analysis.
Table 1.
Overview of articles related to sleep stage
| Author | Location | Number of participants | Age | Health status | Device | Used sensors | Environment | Time | Algorithm | Feature selection | Epoch length | Group | 2 stages accuracy (wake/ sleep) | 4 stages accuracy (wake/ light/ deep/ REM) | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Accuracy | Wake | Light sleep | Deep sleep | REM | ||||||||||||||
| de Zambotti et al. 2019 [7] | America | 41 (13 female) | 17.2 ± 2.4 years | Healthy | Oura Ring Generation 1 | ACC, PPG | Laboratory | 2 days | Proprietary algorithm (first version of Ōuraring algorithm) | N/A | 30s | N/A | N/A | N/A | N/A | 0.65 (0.60, 0.69) | 0.51 (0.43, 0.59) | 0.61 (0.54, 0.69) |
| Roberts et al. 2020 [15] | America | 8 (5 female) | 40.8 ± 4.8 years | Healthy | Oura Ring Generation 1 | ACC, PPG | Laboratory | 4 days | Proprietary algorithm (Sleep-wake classification algorithms) | Mean IBI, SD of IBI, and motion seconds | 30s | N/A | 0.90 (0.86, 0.94) | N/A | N/A | N/A | N/A | N/A |
| Chee et al. 2021 [16] | Singapore | 53 (28 female) | 15.0–19.0 years | Healthy | Oura Ring Generation 2 | ACC, PPG | Semi-laboratory | 15 days | Proprietary algorithm | N/A | 30s | Split sleep group | 0.89 (0.87,0.91) | N/A | N/A | 0.52 (0.50, 0.54) | 0.79 (0.74, 0.84) | 0.53 (0.45, 0.61) |
| Continuous group | 0.89 (0.87, 0.91) | N/A | N/A | 0.52 (0.49, 0.55) | 0.83 (0.79, 0.87) | 0.51 (0.44, 0.58) | ||||||||||||
| Adaptation group | 0.89 (0.88, 0.90) | N/A | N/A | 0.52 (0.50, 0.54) | 0.79 (0.76, 0.82) | 0.53 (0.48, 0.58) | ||||||||||||
| Altini and Kinnunen 2021 [8] | Singapore, Finland and America | Singapore:59 (30 female); Finland:19 (11 female); USA:40 (24 female) | 30.1 ± 16.4 years | Healthy | Oura Ring Generation 2 | ACC, PPG, temperature | Laboratory or home | Singapore: 1–8 days; Finland: 1–2 days; USA: 1–7 days | Proprietary algorithm | Mean, maximum, interquartile, mean amplitude deviation, arm angle, and HRV features | 30s | N/A | 0.96 | N/A | 0.96 (0.96, 0.97) | 0.80 (0.80, 0.81) | 0.91 (0.90, 0.91) | 0.91 (0.90, 0.91) |
| Ghorbani et al. 2022 [17] | Singapore | 58 (26 female) | 37.1 ± 13.0 years | Healthy | Oura Ring Generation 2 | ACC, PPG, temperature | Home | 3 days | Gen 3 algorithm, Gen 2 algorithm | N/A | 30s | Gen2 | 0.90 (0.88, 0.91) | 0.62 (0.60, 0.64) | N/A | N/A | N/A | N/A |
| Gen3 | 0.93 (0.92, 0.94) | 0.76 (0.74, 0.78) | N/A | N/A | N/A | N/A | ||||||||||||
| Miller et al. 2022 [18] | Australia | 53 (26 female) | 25.4 ± 5.9 years | Healthy | Oura Ring Generation 2 | ACC, PPG | Laboratory | 1 day | Proprietary algorithm | Sleep time, and HRV features | 5 min | N/A | 0.89 | 0.61 | N/A | N/A | N/A | N/A |
| Taeyoung Lee et al. 2023 [19] | Korea | 75 (36 female) | 43.6 ± 14.1 years | Subjective sleep discomfort | Oura Ring Generation 3 | ACC, PPG | Hospital | 1 day | Proprietary algorithm | N/A | 30s | N/A | N/A | 0.54 | 0.82 | 0.62 | 0.80 | 0.85 |
| Ju Lynn Ong et al. 2024 [24] | Singapore | 60 (34 female) | 38.5 ± 15.1 years | Healthy | Oura Ring Generation 3 | ACC, PPG, temperature | Laboratory | 1 day | Oura Sleep Staging Algorithm 2.0 (OSSA 2.0) | N/A | 30s | N/A | 0.91 (0.89, 0.93) | N/A | 0.74 (0.69, 0.79) | 0.76 (0.74, 0.78) | 0.74 (0.68, 0.80) | 0.82 (0.78, 0.86) |
| Thomas Svensson et al. 2024 [25] | Japan | 96 (49 female) | 41.9 ± 13.8 years | Healthy | Oura Ring Generation 3 | ACC, PPG, NTC | Hotel | 1-3days | Oura Sleep Staging Algorithm 2.0 (OSSA 2.0) | ACC, HR, HRV features, and temperature | 30s | Non-dominant hand | 0.92 (0.92, 0.92) | N/A | N/A | 0.76 (0.75, 0.76) | 0.89 (0.89, 0.89) | 0.91 (0.91, 0.91) |
| Dominant hand | 0.92 (0.92, 0.92) | N/A | N/A | 0.76 (0.76, 0.76) | 0.89 (0.89, 0.89) | 0.91 (0.91, 0.91) | ||||||||||||
| Rebecca Robbins et al. 2024 [20] | America | 35 (20 female) | 20–50 years | Healthy | Oura Ring Generation 3 | Hospital | 1 day | Oura Sleep Staging Algorithm 2.0 | N/A | 30s | N/A | 0.92 | 0.76 | N/A | N/A | N/A | N/A | |
| Sebastian Herberger et al. 2025 [21] | Europe | 45 (20 female) | 54.6 ± 12.2 years | Unhealthy | Oura Ring Generation 3 | ACC, PPG, NTC | Laboratory | 1 day | N/A | Blood oxygen, HR, HRV, respiration, skin temperature, and movement | 5 min | Oura ring | 0.85 (0.82, 0.88) | 0.53 | N/A | N/A | N/A | N/A |
| Circul ring | N/A | Laboratory | 1 day | N/A | Blood oxygen, HR, HRV, ECG, blood pressure, and finger temperature | 30s | Circul ring | 0.66 (0.62, 0.70) | 0.35 | N/A | N/A | N/A | N/A | |||||
| SleepOn | N/A | Laboratory | 1 day | N/A | Blood oxygen, HR, and movement | 60s | SleepOn | 0.84 (0.81, 0.87) | 0.50 | N/A | N/A | N/A | N/A | |||||
| Hannah Scott et al. 2021 [9] | Australia | Development group: 25 (15 female) | 25.4 ± 6.4 years | Healthy | THIM (firmware v 1.0.3) | ACC | Laboratory | 1 day | THIM sleep tracking algorithm | N/A | 30s | Development group | 0.85 (0.82, 0.88) | N/A | N/A | N/A | N/A | N/A |
| Test group: 20 (14 female) | 23.2 ± 5.0 years | Test group | 0.85 (0.82, 0.88) | N/A | N/A | N/A | N/A | N/A | ||||||||||
| MichaelA.Grandner et al. 2023 [22] | America | 36 (16 female) | 23.4 ± 3.8 years | Healthy | Happy Ring | ACC, PPG, EDA, temperature | Laboratory | 3 days | Happy ring: proprietary algorithm (a “generalized” algorithm; a “personalized” algorithm) | Mean, SD, range, frequency of values and variables | 30s | Personalized group | 0.92 (0.91, 0.93) | N/A | N/A | 0.81 (0.79, 0.83) | 0.95 (0.94, 0.96) | 0.92 (0.91, 0.93) |
| Generalized group | 0.91 (0.90, 0.92) | N/A | N/A | 0.67 (0.65, 0.69) | 0.85 (0.83, 0.87) | 0.85 (0.83, 0.87) | ||||||||||||
| Oura Ring Generation 2 | ACC, PPG | Laboratory | 3 days | N/A | N/A | 5 min | Oura Ring Generation 2 | 0.88 (0.86, 0.90) | N/A | N/A | 0.61 (0.58, 0.64) | 0.81 (0.78, 0.84) | 0.84 (0.82, 0.86) | |||||
| Zachary Strumpf et al. 2023 [36] | America | 84 (42 female) | 48.3 ± 14.4 years | Confirmed OSA and suspect OSA | Belun Ring(BSP2) | ACC, PPG | Laboratory | 1 day | DL-enabled BSP2 algorithms | SpO2, PR, and ACC signal | 30s | N/A | N/A | N/A | 0.88 (0.87, 0.88) | 0.82 (0.81, 0.82)* | 0.90 (0.90, 0.90) | |
| Adrian R. Willoughby et al. 2024 [23] | Singapore | 63 (36 female) | 39.6 ± 15.5 years | Healthy | Oura Ring Generation 3 | N/A | Laboratory | 2 days | Oura Sleep Staging Algorithm 2.0 | N/A | 30s | N/A | N/A | N/A | Sleep period 1: 0.82 | Interruption wake: 0.94 | Sleep period 2: 1.00 | Morning wake: 0.98 |
Note: *indicates the non-rapid eye movement (NREM), including light sleep and deep sleep
Abbreviations: ACC=accelerometer, PPG=photoplethysmography, NTC=negative temperature coefficient, IBI=interbeat interval, SD=standard deviation, HRV=heart rate variability, HR=heart rate, SpO2 = oxygen saturation, PR=pulse rate, EDA=electrodermal activity, REM=rapid eye movement, OSA=obstructive sleep apnea, N/A = not applicable
Table 2.
Overview of articles related to sleep apnea
| Author | Location | Number of participants | Age | Health status | Device | Used sensors | Environment | Algorithm | Feature selection | Epoch length | AHI/OSA cut off | Accuracy |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Wenbo Gu et al. 2020 [26] | America | 50 (23 female) | 54.9 ± 16.3 years | Without major comorbidities or heart rate-affecting medications | Belun Ring | ACC, PPG | Laboratory | Neural network algorithm | SpO2, PR, HRV, ACC signal, and PPG waveform | 30s | 5 | 0.86 |
| 15 | 0.86 | |||||||||||
| 30 | 0.86 | |||||||||||
| Eric Yeh et al. 2021 [10] | America | 78 (51 female) | 51.5 ± 13.4 years | On heart-rate-affecting medications or with non-arrhythmic conditions | Belun Ring (BSP1) | ACC, PPG | Laboratory | BSP proprietary OSA detection algorithm | SpO2, PR, HRV, ACC signal, and PPG waveform | 5 min | 5 | 0.56 (0.45,0.68) |
| 15 | 0.81 (0.70,0.89) | |||||||||||
| 30 | 0.91 (0.82,0.96) | |||||||||||
| Ou, Yi-Hui et al. 2023 [35] | Singapore | 129 (15 female) | 60.5 ± 8.4 years | With hypertension and high cardiovascular risk | Belun ring | ACC, PPG | Laboratory | Nneural network algorithm | N/A | N/A | 5 | 0.84 |
| 15 | 0.87 | |||||||||||
| 30 | 0.92 | |||||||||||
| Zachary Strumpf et al. 2023 [36] | America | 84 (42 female) | 48.3 ± 14.4 years | Confirmed OSA and suspected OSA | Belun Ring (BSP2) | ACC, PPG | Laboratory | DL-enabled BSP2 algorithms | SpO2, PR, and ACC signal | 5 min | 5 | 0.85 (0.75,0.92) |
| 15 | 0.89 (0.81,0.95) | |||||||||||
| 30 | 0.91 (0.82,0.96) | |||||||||||
| Nuttida Panichapat et al. 2024 [27] | Thailand | 75 (21 female) | 9.2 ± 2.8 years | Suspected OSA | Belun Ring | ACC, PPG | Hospital | Neural network algorithm | SpO2, HRV, ACC, and PPG | N/A | 3 | 0.68 |
| 5 | 0.52 | |||||||||||
| 7 | 0.44 | |||||||||||
| Frederik Massie et al. 2018 [28] | Belgium | 101 (44 female) | 53.0 ± 13.0 years | Suspected OSA | NightOwl | ACC, PPG | Laboratory | N/A | Actigraphy, SpO2, PAT, PR, and other features | N/A | Overall | 0.81 |
| Bart Van Pee et al. 2022 [29] | Belgium, America | 167 (62 female) | 56.0 ± 15.0 years | Suspected OSA | NightOwl | ACC, PPG | Laboratory | N/A | N/A | N/A | 5 | 0.90 (0.86,0.95) |
| 15 | 0.86 (0.80,0.91) | |||||||||||
| 30 | 0.91 (0.87,0.95) | |||||||||||
| Christopher J. Lyne et al. 2023 [30] | Australia | 100 (43 female) | 48.0 ± 11.8 years | Suspected OSA | NightOwl | ACC, PPG | Hospital | Proprietary algorithm | Actigraphy, PAT | N/A | NightOwl Mini overall | 0.64 |
| NightOwl reusable overall | 0.68 | |||||||||||
| George do Lago Pinheiro et al. 2020 [31] | 304 (135 female) | 55.3 ± 13.8 years | Suspected OSA | Biologix™ | ACC | Laboratory | Automated cloud algorithm | N/A | 1s | Moderate-to-severe OSA, ODM-ODI cut off = 12, 3% oxygen desaturation | 0.90 | |
| Moderate-to-severe OSA, ODM-ODI cut off = 14, 4% oxygen desaturation | 0.91 | |||||||||||
| Rui Zhao et al. 2022 [11] | China | 207 (62 female) | 48.2 ± 14.7 years | Suspected sleep apnea | Circul Ring | ACC, PPG | Laboratory | Automated algorithm | SpO2, PR, and ACC signal | 1s | 5 | 0.86 |
| 15 | 0.76 | |||||||||||
| 30 | 0.91 | |||||||||||
| Mi Lu et al. 2023 [32] | German | 39 (10 female) | 56.8 ± 15.0 years | Suspected sleep apnea | SleepImage Ring | PPG | Laboratory | Cloud-based SleepImage system | PR variability, respiration coherence, cross spectral power, and SpO2 | N/A | 5 | 1.00 (1.00,1.00) |
| 15 | 0.79 (0.67,0.92) | |||||||||||
| 30 | 0.90 (0.80,0.99) | |||||||||||
| Zetong Wu et al. 2024 [12] | America | 58 | 54.0–93.0 years | Without clinical cardiovascular disease | RingConn smart ring | PPG | N/A | N/A | Trend components, periodic component, and SpO2 | 60s | 5 | 0.83 |
| 15 | 0.83 | |||||||||||
| 30 | 0.88 | |||||||||||
| Hongbo Guo et al. 2024 [34] | China | 58 (17 female) | 52.2 ± 12.9 years | N/A | RingConn smart ring | ACC, PPG, temperature | Hospital | N/A | SpO2, ACC, PPG, and temperature | 60s | 5 | 0.81 |
| 15 | 0.86 | |||||||||||
| 30 | 0.88 | |||||||||||
| James Tisyakorn et al. 2024 [33] | Thailand | 190 (89 female) | 43.0 ± 16.0 years | Suspected OSA | Wellue O2 ring | N/A | Laboratory | N/A | N/A | 20s | Moderate-to-severe OSA, 11% ODI cut off = 1.25 | N/A |
Abbreviations: ACC=accelerometer, PPG=photoplethysmography, OSA=obstructive sleep apnea, SpO2 = oxygen saturation, PR=pulse rate, HRV=heart rate variablity, PAT=peripheral arterial tone, ODM=overnight digital monitoring, ODI=oxygen desaturation index, N/A = not applicable
Fig. 2.
Distribution of studies and included participants based on health status. The figure presents the distribution of studies and relative participants included in the reviewed studies based on their health status. Abbreviations: n = the number of study, OSA=obstructive sleep apnea, CVD=cardiovascular disease
Figure 2 illustrates the distribution of included participants by health status. In studies utilizing finger-worn devices for sleep staging, 86% (12/14) predominantly involved healthy participants. Conversely, for OSA assessment, 77% (10/13) of the studies focused on clinical cohorts. Specifically, among the OSA-focused studies, eight involved participants with suspected sleep apnea, two focused on individuals with cardiovascular disease (CVD), and only two included healthy participants.
Algorithms for finger-worn devices
Eight (73%) finger-worn devices predominantly employed proprietary algorithms (Tables 1 and 2). Oura Ring’s algorithms performed iterative advancements: the first generation model combined the physiological signal with a gradient boosting classifier to evaluate classification performance [15]. The second generation model employed LightGBM with DART optimization [8]. The third generation OSSA 2.0 integrated multimodel data from a 3D accelerometer, seven skin temperature sensors, and infrared PPG to enhance staging precision [24]. Hannah et al. developed THIM sleep tracking algorithm scored the sleep or wakefulness every 30s, which incorporated a smoothing function [9]. The Happy Ring platform implemented a dual algorithmic framework: a “generalized” algorithm adopted leave-one-out cross-validation (LOOCV) for all users, and a “personalized” model utilized non-overlapping time window and a five-fold cross-validation to preserve sleep pattern continuity [22]. Belun ring integrated a medical-grade ring pulse oximeter with a proprietary neural-network-based artificial intelligence algorithm [10, 36]. Circul Ring automatically processed embedded rPPG and accelerometer (ACC) signals [11].
Types of devices and sensors
This comprehensive review critically evaluated 11 advanced wearable rings, with a particular focus on their specialized applications (Figs. 3 and Supplementary Fig. 1): the Oura ring, SleepOn, THIM, and Happy ring were meticulously examined for their capabilities in sleep stage monitoring and analysis, while the NightOwl, Biologix™, SleepImage, RingConn smart ring and Wellue O2 ring were rigorously evaluated for efficacy in sleep apnea detection. The Circul ring and Belun ring enabled assessment of both sleep stages and sleep apnea.
Fig. 3.
Types of finger-worn devices and sensors. Abbreviations: ACC=accelerometer, PPG=photoplethysmography, EDA=electrodermal activity, N/A = not applicable
The sensors of finger-worn device were shown in Fig. 3 and Supplementary Fig. 2. Eight devices utilized multi-sensor to collect physiological, movement, and other health-related metrics. Specifically, four devices [7, 11, 28, 35] utilized ACC and PPG to measure body signals. Three smart rings [8, 24, 34] incorporated data from ACC, PPG, and temperate sensors. Furthermore, Happy Ring [22] not only utilized the above three sensors, but also relied on electrodermal activity (EDA) sensors to facilitate multidimensional analyses of physical activity, peripheral blood flow, and distal thermoregulation. Two finger-worn devices sensors were unclear.
Performance in sleep stage detection versus PSG
For wake/sleep classification, as shown in Fig. 4, a meta-analysis of seven studies revealed a pooled accuracy of 0.87 (0.86,0.89). Specifically, the Oura Ring Generation 2 with Gen 3 algorithm reported by Ghorbani et al. demonstrated the highest accuracy at 0.93, while the Circul ring reported by Sebastian et al. exhibited the lowest accuracy at 0.66.
Fig. 4.
The accuracy of finger-worn devices for two sleep stages versus PSG. Note: Weights are from random-effects model. Mean and 95% confidence intervals are displayed. Abbreviations: DL=DerSimonian-Laird
For four-stage sleep classification, a meta-analysis of five studies (Fig. 5) yielded pooled accuracy of 0.65 (0.56,0.74) for light sleep, 0.81 (0.77,0.85) for deep sleep, and 0.74 (0.68,0.79) for REM sleep detection. Specifically, Michael A et al. reported that the Happy Ring with personalized algorithm, demonstrated the highest accuracy rates: 0.81 (0.79,0.83) for light sleep, 0.95 (0.94,0.96) for deep sleep, and 0.92 (0.91,0.93) for REM sleep. In contrast, the Oura Ring Generation 2 reported by Chee et al. showed the lowest light (0.52) and REM (0.51) sleep accuracy, while Generation 1 reported by de Zambotti showed the lowest deep sleep accuracy (0.51). The leave-one-out analysis confirmed the robustness of these results (Supplementary Figs. 3 and 4). Notably, only the Belun ring performed three-stage classification, which reported accuracy of 0.88 for wake, 0.82 for NREM, and 0.90 for REM [36] (Table 1).
Fig. 5.
The accuracy of finger-worn devices for four sleep stages versus PSG. Note: Weights are from random-effects model. Mean and 95% confidence intervals are displayed. Abbreviations: DL=DerSimonian-Laird
In two-stage sleep classification compared with PSG, finger-worn devices demonstrated sensitivity ranging from 0.74 to 0.97, specificity ranging from 0.35 to 0.98, and balanced accuracy ranging from 0.55 to 0.90. For four sleep stages, sensitivity ranges were 0.45–0.82 for light sleep, 0.17–0.89 for deep sleep, and 0.14–0.89 for REM sleep, with corresponding specificity values of 0.64–0.91, 0.80–0.95, 0.87–0.94 across finger-worn devices. Reported Cohen’s kappa ranged from 0.06 to 0.70, with a considerable number of studies not reporting the metric. Details are shown in Supplementary Tables 4–6.
Performance in sleep apnea detection versus PSG
Figure 6 shows that finger-worn devices predominantly utilize AHI thresholds of 5, 15, and 30 events/hour for severity classification in adults. At the AHI = 5 events/hour, the SleepImage Ring demonstrated the highest accuracy (1.00), whereas the Belun Ring in Eric et al.’s study exhibited the lowest accuracy (0.56). For using an AHI threshold of at least 15 events/hour, the Belun Ring reported by Zachary et al. achieved an accuracy of 0.89, but Circul Ring reported 0.76. At the severe threshold (AHI = 30 events/hour), the Belun Ring reported by Ou et al. maintained high accuracy (0.92), while the RingConn smart ring reported by Wenbo et al. showed the lowest performance (0.86). Notably, all of devices displayed higher accuracy at AHI thresholds of 30 compared to the cutoff of 5 and 15, except for the SleepImage Ring. Besides, Nuttida et al. [27] highlighted declining accuracy for Belun ring at low thresholds in children: 0.68, 0.52, 0.44 at AHI = 3, 5, and 7 events/hour.
Fig. 6.
The accuracy of finger-worn devices for sleep apnea versus PSG. Note: Error bars indicate 95% confidence intervals (95% CI). *Study participants were children. Abbreviations: AHI=apnea hypopnea index
Beyond overall accuracy, the sensitivity, specificity, and Cohen’s kappa of finger-worn devices for OSA detection across various AHI thresholds are detailed in Supplementary Table 7. At the threshold of AHI = 5, sensitivity ranged widely from 0.85 to 1.00. At the moderate threshold (AHI = 15), sensitivity ranged from 0.65 to 0.97, and at the severe threshold (AHI = 30), it ranged from 0.22 to 0.91. In contrast, specificity showed increase at the cutoff of 5, 15, and 30 events/hour, ranging from 0.03 to 1.00, 0.69-1.00, and 0.91-1.00 respectively. Belun ring demonstrated the highest kappa value at the threshold of 15 (κ = 0.79) and 30 (κ = 0.93). At the cutoff of 5, the RingConn smart ring performed highest kappa value (κ = 0.65).
Heterogeneity across studies
To explore the sources of heterogeneity, we conducted a subgroup analysis based on different groups (Supplementary Tables 8 and 9). For sleep stage, accuracy was higher with 30s epochs for two-stage sleep (0.88, 95% CI: 0.86–0.90, I2 = 94.4%) and light sleep (0.66, 95% CI: 0.56–0.75, I2 = 99.6%). Participants aged 18–50 years had high accuracy across most sleep stages, whereas those > 50 years showed low accuracy for two sleep stage (0.78, 95% CI: 0.68–0.89, I2 = 96.9%). Furthermore, healthy participants exhibited higher accuracy (0.89, 95% CI: 0.88–0.91, I2 = 83.9%) compared to unhealthy individuals (0.78, 95% CI: 0.68–0.89; I2 = 96.9%) in two sleep stage. For OSA detection, AHI threshold indicated the accuracy of 0.78, 0.85, and 0.91 for the threshold of 5, 15, and 30.
Bias assessment
As summarized in Supplementary Fig. 5, three (11%) studies demonstrated a high risk of bias in patient selection domain, while 21 (75%) studies were rated as having an unclear risk for this domain due to insufficient details on whether participants were enrolled consecutively or randomly. For the index test domain and reference standard domain, only one (4%) and two (7%) studies demonstrated a high risk of bias. In the flow and timing domain, 20 studies (71%) exhibited a high risk of bias, primarily due to participant exclusions (e.g., insufficient sleep duration or device malfunctions). Overall, only one study was judged to have a low risk of bias, while 27 were judged to have some risk of bias. To provide a comprehensive review of wearable ring technologies, studies with identified risks of bias were retained.
Discussion
Key findings and technological implications
Wearable technologies have expanded to include body-, wrist-, and finger-worn devices [37]. Finger-worn devices represent an innovative category within this field, offering the ability to monitor diverse sleep metrics, yet their accuracy in staging sleep and detecting apnea remains poorly quantified. This review systematically evaluates their potential for these two tasks.
Sleep stage classification
The American Academy of Sleep Medicine (AASM) calls for rigorous validation of consumer sleep technologies, yet no standardized protocol exists for wearables lacking EEG. Menghini et al. proposed a framework that distinguishes two assessment tiers: macro-level sleep metrics (total sleep time, sleep onset latency, wake after sleep onset, and sleep efficiency) and epoch-by-epoch agreement (accuracy, sensitivity, and specificity) [38]. To maximize information on classifier performance, only studies reporting epoch-level comparisons were included in this review. Sleep stages are derived from physiology either by expert scoring or machine-learning algorithms [39]. Accurate staging is the cornerstone of clinical and research sleep measurement. Finger-worn sensors now reach high two-class (wake vs. sleep) accuracy, but aligning with the AASM four-class scheme (N1, N2, N3, REM) remains unresolved, such findings consistent with a recent review of 62 wearable devices [5]. This four-stage classification more accurately reflects human physiological patterns, thereby necessitating a high degree of precision from finger-worn devices.
Sleep apnea detection
Although prior reviews have examined portable devices for OSA diagnosis, evidence on finger-worn devices is scarce [40, 41]. Our study indicates that smart rings perform adequate sensitivity for mild but high specificity and accuracy for severe OSA, echoing a recent review without meta-analysis [42]. Performance must therefore be interpreted by intended use. High-specificity devices (e.g., Belun Ring at AHI ≥ 15/≥30) may suited for triaging moderate-to-severe OSA, while high-sensitivity devices (at AHI ≥ 5) could enable initial screening in primary care or at home, with referrals for PSG confirmation.
Notably, devices such as the Belun ring demonstrate concurrent monitoring capabilities for both sleep staging and OSA. Its second-generation algorithm, trained on first-generation datasets, integrates dual proprietary neural networks dedicated to sleep stage classification and respiratory event detection [36]. This multi-functionality advances holistic sleep assessment through integrated physiological capture.
Methodological and population considerations
This systematic review identified accelerated research growth, with 70% of relevant studies published within the past quadrennium. We evaluated 11 finger-worn devices, among which Oura Ring and Belun Ring constitute the most extensively researched. Notably, technological evolution has shifted from single-sensor to multi-sensor configurations [8], predominantly integrating ACC and PPG to capture concurrent physiological, kinematic, and health parameters [43]. Contemporary iterations further incorporate temperature sensors, enabling comprehensive monitoring of sleep-associated physiological dynamics, which align with the requirements of core sleep medicine biomarkers.
Finger-worn devices enable cost-effective multi-night sleep monitoring in natural environments. Continuous longitudinal data acquisition mitigates single-night sampling bias while tracking dynamic trends in physiological and environmental covariates. This approach enhances clinical decision-making through pattern recognition of interrelated variables, improving ecological validity in real-world contexts [44, 45].
Despite prior validation of finger-worn devices’ sleep monitoring potential [21, 46], current evidence reveals inconsistent PSG-referenced accuracy. Most devices have shown improved accuracy over time, primarily due to algorithmic optimizations and integration of multi-sensor technologies. Notably, the lack of transparency regarding proprietary algorithms-functioning as black boxes, which obscures the specific mechanisms. This opacity also restricts end-users and clinicians from accessing raw data, limiting further validation and refinement of the technology [5]. Furthermore, a considerable number of studies fail to focus on the imbalanced datasets in sleep-stage classification, limiting a comprehensive assessment of device performance.
Methodological heterogeneity in algorithms and sensor configurations likely underlies inter-study discrepancies. For instance, personalized algorithms (e.g., Happy Ring [22]) demonstrate superior accuracy to generalized approaches, suggesting customization as a critical development pathway. Epoch length is also a significant source of heterogeneity and influences accuracy estimates. Using the standard 30-second epochs yielded higher pooled accuracy for two-stage and light sleep staging. However, this subgroup also exhibited high statistical heterogeneity, suggesting that even among studies using the same epoch length, other factors (device, population) cause substantial performance variation. In addition, population heterogeneity appears to be another contributing factor, as evidenced by studies reporting variable accuracy results for the same finger-worn device, although hand dominance shows negligible effect [25].
Limitations
This review systematically evaluates finger-worn devices against PSG standards for sleep staging and OSA diagnosis, providing critical developmental insights. However, several limitations should be acknowledged in this review. The primary limitation of this review is the scarcity of published validation studies for the available consumer devices. To address this, we sought additional studies by checking references or through snowball sampling. Despite these efforts, it is possible that some finger-worn devices were overlooked. Second, the strength of evidence is tempered by the methodological quality of the included studies. Most studies were rated as having high or unclear risk of bias, which reflects the current lack of standardized protocols and reporting frameworks in this rapidly evolving field. This may inflate or bias the reported accuracy estimates. Third, as ring technology (including sensor hardware, firmware, and proprietary algorithms), evolves rapidly, the performance reported in the current literature cannot be directly generalized to future versions or substantially revised versions. Another limitation is that costs were not extensively addressed. Prices and business models vary widely and fluctuate with market dynamics, making cost considerations important for prospective users. Future studies should therefore assess the cost-effectiveness and add value of these wearable and nearable technologies. Finally, the limited applicability of current findings to the general population stems from spectrum bias. For sleep staging, the predominant enrollment of healthy adults (to minimize comorbidity confounders) restricts generalizability to pathological populations, as the performance of finger-worn devices in patients with diverse sleep pathologies remains largely unexamined. Conversely, studies on OSA have largely focused on high-risk, clinically suspected populations, typically overweight adults over 50 years old with moderate to severe apnea. This leaves individuals with mild OSA or those with comorbid condition (e.g., cardiovascular or neurological diseases) underrepresented. A major evidence gap also exists for pediatric and adolescent populations, where finger-worn devices have been scarcely evaluated. These limitations constrain the broader applicability of our findings and highlight the need for future research in more diverse clinical and demographic cohorts.
Clinical relevance and future directions
Our study has significant clinical relevance. Current devices demonstrate proficiency in detecting severe OSA and differentiating basic sleep/wake states; however, their accuracy in identifying mild OSA and distinguishing the four sleep stages remains inadequate for clinical diagnosis. Furthermore, the applicability of findings is often limited by population homogeneity, which may not reflect the complexity encountered in diverse patient profiles.
Future research should prioritize several key directions:
Promoting open-source frameworks: The current dominance of proprietary algorithms (8/11 devices in review) hinders independent validation and innovation. Open-source approaches would accelerate progress in these areas.
Addressing pediatric/adolescent populations: Current evidence is scarce, with only three studies in this review involving children or adolescents. Developing age-appropriate sensor placements and algorithms is an urgent need.
Establishing methodological consensus: Standardization is critical for quality control (QC) steps of data processing, algorithm calibration, epoch length, and feature definitions across studies and devices.
Fostering interdisciplinary collaboration: Advancing finger-worn device technology requires close, sustained collaboration among medical, engineering, and other relevant disciplines to foster innovation and expand clinical applications.
Conclusion
Finger-worn devices may be used for longitudinal sleep/wake tracking in healthy adults or as a triage tool for severe OSA, but should not replace PSG for diagnosis of mild OSA or detailed sleep architecture assessment. Achieving PSG-comparable accuracy in multi-stage sleep analysis and mild OSA detection requires addressing algorithmic transparency, population diversity, and methodological standardization. Therefore, future innovation should prioritize clinically actionable insights through interdisciplinary collaboration.
Electronic Supplementary Material
Below is the link to the electronic supplementary material.
Acknowledgements
We thank all the investigators of the studies included in this systematic review and meta-analysis.
Abbreviations
- ACC
Accelerometer
- PPG
Photoplethysmography
- NTC
Negative temperature coefficient
- IBI
Interbeat interval
- SD
Standard deviation
- HRV
Heart rate variability
- HR
Heart rate
- SpO2
Oxygen saturation
- PR
Pulse rate
- EDA
Electrodermal activity
- REM
Rapid eye movement
- OSA
Obstructive sleep apnea
- PR
Pulse rate
- PAT
Peripheral arterial tone
- ODM
Overnight digital monitoring
- ODI
Oxygen desaturation index
- CVD
Cardiovascular disease
- DL
DerSimonian-Laird
- AHI
Apnea hypopnea index
- N/A
Not applicable
Author contributions
YXJ, JWX, HJY, YZH, WJM, XMF, HMW, LX and JW have substantial contributions to conception and design, acquisition of funding, data and interpretation of data; YXJ and JWX analyzed the data and drafted the article, HMW, LX and JW revised it critically for important intellectual content. YXJ, JWX and HJY contributed equally to this work. JW is the guarantors. All authors contributed to final approval of the paper.
Funding
This work was funded by the Natural Science Foundation of Guangdong (2025A1515011659) and the fundamental research funds for the central universities, Sun yat-sen university (24qnpy211).
Data availability
All data generated or analysed during this study are included in this published article and its supplementary information files.
Declarations
Ethical approval
Ethical approval is not required because this study is retrieved and synthesized data from already published studies.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Yexin Jin, Jiawen Xu and Huijun Yue contributed Co-first author: equal contribution.
References
- 1.Wang J, Li AM, Lam H. Sleep duration and adiposity in children and adults: observational and mendelian randomization studies. 2019;27(6):1013–22. [DOI] [PubMed]
- 2.Wang J, Kwok MK, Au Yeung SL, Li AM, Lam HS, Leung JYY, et al. Sleep duration and risk of diabetes: Observational and Mendelian randomization studies. Prev Med. 2019;119:24–30. [DOI] [PubMed] [Google Scholar]
- 3.De Crescenzo F, D’Alò GL, Ostinelli EG, Ciabattini M, Di Franco V, Watanabe N, et al. Comparative effects of pharmacological interventions for the acute and long-term management of insomnia disorder in adults: a systematic review and network meta-analysis. Lancet. 2022;400(10347):170–84. [DOI] [PubMed] [Google Scholar]
- 4.Berry RB, Budhiraja R, Gottlieb DJ, Gozal D, Iber C, Kapur VK, et al. Rules for Scoring Respiratory Events in Sleep: Update of the 2007 AASM Manual for the Scoring of Sleep and Associated Events. J Clin Sleep Med. 2012;08(05):597–619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Birrer V, Elgendi M, Lambercy O, Menon C. Evaluating reliability in wearable devices for sleep staging. NPJ Digit Med. 2024;7(1):74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Benjafield AV, Ayas NT, Eastwood PR, Heinzer R, Ip MSM, Morrell MJ, et al. Estimation of the global prevalence and burden of obstructive sleep apnoea: a literature-based analysis. Lancet Respir Med. 2019;7(8):687–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.de Zambotti M, Rosas L, Colrain IM, Baker FC. The Sleep of the Ring: Comparison of the ŌURA Sleep Tracker Against Polysomnography. Behav Sleep Med. 2019;17(2):124–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Altini M, Kinnunen H. The Promise of Sleep: A Multi-Sensor Approach for Accurate Sleep Stage Detection Using the Oura Ring. Sens (Basel). 2021;21:13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Scott H, Lovato N, Lack L. The Development and Accuracy of the THIM Wearable Device for Estimating Sleep and Wakefulness. Nat Sci Sleep. 2021;13:39–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yeh E, Wong E, Tsai CW, Gu W, Chen PL, Leung L, et al. Detection of obstructive sleep apnea using Belun Sleep Platform wearable with neural network-based algorithm and its combined use with STOP-Bang questionnaire. PLoS ONE. 2021;16(10):e0258040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Zhao R, Xue J, Zhang X, Peng M, Li J, Zhou B, et al. Comparison of Ring Pulse Oximetry Using Reflective Photoplethysmography and PSG in the Detection of OSA in Chinese Adults: A Pilot Study. Nat Sci Sleep. 2022;14:1427–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wu ZT, Wu H, Fang KQ, Sze KSF, Feng QJ. A Transformer-based deep learning model for sleep apnea detection and application on ringconn smart ring. Ieee Int Symp Circ S. 2024.
- 13.Fitzner K. Reliability and validity: a quick review. Diabetes Educ. 2007;33(5):775–6. [DOI] [PubMed] [Google Scholar]
- 14.Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529–36. [DOI] [PubMed] [Google Scholar]
- 15.Roberts DM, Schade MM, Mathew GM, Gartenberg D, Buxton OM. Detecting sleep using heart rate and motion data from multisensor consumer-grade wearables, relative to wrist actigraphy and polysomnography. Sleep. 2020;43(7). [DOI] [PMC free article] [PubMed]
- 16.Chee N, Ghorbani S, Golkashani HA, Leong RLF, Ong JL, Chee MWL. Multi-Night Validation of a Sleep Tracking Ring in Adolescents Compared with a Research Actigraph and Polysomnography. Nat Sci Sleep. 2021;13:177–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ghorbani S, Golkashani HA, Chee N, Teo TB, Dicom AR, Yilmaz G, et al. Multi-Night at-Home Evaluation of Improved Sleep Detection and Classification with a Memory-Enhanced Consumer Sleep Tracker. Nat Sci Sleep. 2022;14:645–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Miller DJ, Sargent C, Roach GD. A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults. Sens (Basel). 2022;22(16). [DOI] [PMC free article] [PubMed]
- 19.Lee T, Cho Y, Cha KS, Jung J, Cho J, Kim H, et al. Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study. JMIR Mhealth Uhealth. 2023;11:e50983. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Robbins R, Weaver MD, Sullivan JP, Quan SF, Gilmore K, Shaw S, et al. Accuracy of three commercial wearable devices for sleep tracking in healthy adults. Sens (Basel). 2024;24(20). [DOI] [PMC free article] [PubMed]
- 21.Herberger S, Aurnhammer C, Bauerfeind S, Bothe T, Penzel T, Fietze I. Performance of wearable finger ring trackers for diagnostic sleep measurement in the clinical context. Sci Rep. 2025;15(1):9461. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Grandner MA, Bromberg Z, Hadley A, Morrell Z, Graf A, Hutchison S, et al. Performance of a multisensor smart ring to evaluate sleep: in-lab and home-based evaluation of generalized and personalized algorithms. Sleep. 2023;46(1). [DOI] [PMC free article] [PubMed]
- 23.Willoughby AR, Golkashani HA, Ghorbani S, Wong KF, Chee N, Ong JL, et al. Performance of wearable sleep trackers during nocturnal sleep and periods of simulated real-world smartphone use. Sleep Health. 2024;10(3):356–68. [DOI] [PubMed] [Google Scholar]
- 24.Ong JL, Golkashani HA, Ghorbani S, Wong KF, Chee N, Willoughby AR, et al. Selecting a sleep tracker from EEG-based, iteratively improved, low-cost multisensor, and actigraphy-only devices. Sleep Health. 2024;10(1):9–23. [DOI] [PubMed] [Google Scholar]
- 25.Svensson T, Madhawa K, Nt H, Chung UI, Svensson AK. Validity and reliability of the Oura Ring Generation 3 (Gen3) with Oura sleep staging algorithm 2.0 (OSSA 2.0) when compared to multi-night ambulatory polysomnography: A validation study of 96 participants and 421,045 epochs. Sleep Med. 2024;115:251–63. [DOI] [PubMed] [Google Scholar]
- 26.Gu W, Leung L, Kwok KC, Wu IC, Folz RJ, Chiang AA. Belun Ring Platform: A novel home sleep apnea testing system for assessment of obstructive sleep apnea. J Clin Sleep Med. 2020;16(9):1611–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Panichapat N, Niyomkarn W, Boonjindasup W, Thiamrakij P, Sritippayawan S, Deerojanawong J. Diagnostic accuracy of the belun ring in children at risk of obstructive sleep apnea. J Clin Sleep Med. 2024. [DOI] [PMC free article] [PubMed]
- 28.Massie F, De Almeida DM, Dreesen P, Thijs I, Vranken J, Klerkx S. An evaluation of the Night Owl home sleep apnea testing system. J Clin Sleep Med. 2018;14(10):1791–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Van Pee B, Massie F, Vits S, Dreesen P, Klerkx S, Bijwadia J, et al. A multicentric validation study of a novel home sleep apnea test based on peripheral arterial tonometry. Sleep. 2022;45(5). [DOI] [PMC free article] [PubMed]
- 30.Lyne CJ, Hamilton GS, Turton ARE, Stupar D, Mansfield DR. Validation of a single-use and reusable home sleep apnea test based on peripheral arterial tonometry compared to laboratory polysomnography for the diagnosis of obstructive sleep apnea. J Clin Sleep Med. 2023;19(8):1429–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Pinheiro GDL, Cruz AF, Domingues DM, Genta PR, Drager LF, Strollo PJ, et al. Validation of an Overnight Wireless High-Resolution Oximeter plus Cloud-Based Algorithm for the Diagnosis of Obstructive Sleep Apnea. Clin (Sao Paulo). 2020;75:e2414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Lu M, Brenzinger L, Rosenblum L, Salanitro M, Fietze I, Glos M, et al. Comparative study of the SleepImage ring device and polysomnography for diagnosing obstructive sleep apnea. Biomed Eng Lett. 2023;13(3):343–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Tisyakorn J, Saiphoklang N, Sapankaew T, Thapa K, Anutariya C, Sujarae A, et al. Screening moderate to severe obstructive sleep apnea with wearable device. Sleep Breath. 2025;29(1). [DOI] [PubMed]
- 34.Guo HB, Wu H, Xia JH, Cheng YB, Guo QH, Chen Y, et al. OSAHS detection capabilities of ringconn smart ring: a feasibility study. Arti Intel Circ Sys. 2024:597–601.
- 35.Ou YH, Ong J, Thant AT, Koo CY, Leung L, Sia CH, et al. The Belun sleep platform to diagnose obstructive sleep apnea in patients with hypertension and high cardiovascular risk. J Hypertens. 2023;41(6):1011–7. [DOI] [PubMed] [Google Scholar]
- 36.Strumpf Z, Gu W, Tsai CW, Chen PL, Yeh E, Leung L, et al. Belun Ring (Belun Sleep System BLS-100): Deep learning-facilitated wearable enables obstructive sleep apnea detection, apnea severity categorization, and sleep stage classification in patients suspected of obstructive sleep apnea. Sleep Health. 2023;9(4):430–40. [DOI] [PubMed] [Google Scholar]
- 37.Pinilla L, Chai-Coetzer CL, Eckert DJ. Diagnostic Modalities in Sleep Disordered Breathing: Current and Emerging Technology and Its Potential to Transform Diagnostics. Respirology. 2025;30(4):286–302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Menghini L, Cellini N, Goldstone A, Baker FC, de Zambotti M. A standardized framework for testing the performance of sleep-tracking technology: step-by-step guidelines and open-source code. Sleep. 2021;44(2). [DOI] [PMC free article] [PubMed]
- 39.Roomkham S, Lovell D, Cheung J, Perrin D. Promises and Challenges in the Use of Consumer-Grade Devices for Sleep Monitoring. IEEE Rev Biomed Eng. 2018;11:53–67. [DOI] [PubMed] [Google Scholar]
- 40.Chiang AA, Jerkins E, Holfinger S, Schutte-Rodin S, Chandrakantan A, Mong L, et al. OSA diagnosis goes wearable: are the latest devices ready to shine? J Clin Sleep Med. 2024;20(11):1823–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Abd-Alrazaq A, Aslam H, AlSaad R, Alsahli M, Ahmed A, Damseh R, et al. Detection of Sleep Apnea Using Wearable AI: Systematic Review and Meta-Analysis. J Med Internet Res. 2024;26:e58187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Khor YH, Khung SW, Ruehland WR, Jiao Y, Lew J, Munsif M, et al. Portable evaluation of obstructive sleep apnea in adults: A systematic review. Sleep Med Rev. 2023;68:101743. [DOI] [PubMed] [Google Scholar]
- 43.Baumert M, Cowie MR, Redline S, Mehra R, Arzt M, Pépin JL, et al. Sleep characterization with smart wearable devices: a call for standardization and consensus recommendations. Sleep. 2022;45(12). [DOI] [PubMed]
- 44.de Zambotti M, Goldstein C, Cook J, Menghini L, Altini M, Cheng P, et al. State of the science and recommendations for using wearable technology in sleep and circadian research. Sleep. 2024;47(4). [DOI] [PubMed]
- 45.Kamaldeep Singh AJ, Ishita Panchal H, Madan S, Chaturvedi A, Kostojchin A, Shahzadi MM, Khan. Shobhit Piplani. Smartphones and Consumer Devices in Management of Obstructive Sleep Apnea. Indian J Sleep Med. 2022;17(4):103–7. [Google Scholar]
- 46.de Zambotti M, Cellini N, Goldstone A, Colrain IM, Baker FC. Wearable Sleep Technology in Clinical and Research Settings. Med Sci Sports Exerc. 2019;51(7):1538–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analysed during this study are included in this published article and its supplementary information files.









