Wearable photoplethysmography enables scalable detection of obstructive sleep apnoea in a large population cohort (ERS 2026 late-breaking abstract)

Rule-based PPG + SpO2 algorithm (PPG-derived respiration, movement/HRV sleep estimation, AASM-aligned event detection) estimates AHI. Developed on two internal clinical datasets (NCT03927547, NCT06930404; n=308), externally validated on 1,881 MESA recordings against expert scoring (age 69.3 ± 9.1, 54% female). AHI ≥5: Se 89.5%, Sp 90.5%.

• Form factor: finger (MESA PSG oximeter for validation)

Background: Obstructive sleep apnoea affects an estimated 1 billion people worldwide yet remains substantially underdiagnosed due to limited access to conventional diagnostic pathways. Photoplethysmography (PPG), widely available in consumer wearables, presents an opportunity for scalable, accessible screening beyond traditional sleep laboratories. We evaluated whether a PPG-based algorithm could accurately detect OSA and estimate apnoea–hypopnoea index (AHI) compared with expert-scored reference data. Aims: We evaluated whether a PPG-based algorithm could accurately detect OSA and estimate apnoea–hypopnoea index (AHI) compared with expert-scored reference data. Methods: A rule-based algorithm was developed to detect respiratory events from PPG and oxygen saturation signals and estimate AHI, incorporating PPG-derived respiratory extraction, movement- and heart rate variability-based sleep estimation, and AASM-aligned event detection. The algorithm was trained using two internal clinical datasets ( NCT03927547 and NCT06930404 ; n=308). Independent validation was performed on 1881 recordings from the publicly available Multi-Ethnic Study of Atherosclerosis (MESA) dataset using expert-scored annotations. Participants had a mean age of 69.3±9.1 years and BMI of 28.7±5.5 kg·m⁻²; 54% were female. Results: At AHI ≥5 events·h⁻¹, sensitivity and specificity were 89.5% and 90.5%, respectively. At AHI ≥15 events·h⁻¹, sensitivity and specificity were 773% and 97.0%, respectively. Total sleep time estimation demonstrated a mean difference of 0 min, mean absolute difference of 51 min. Conclusions: These findings support the potential of wearable-derived PPG signals to enable scalable, home-based sleep apnoea screening and broaden access to early detection of OSA at population scale.

Verbatim from OpenAlex (opens in a new tab) · Abstract via OpenAlex (publisher-deposited); © the authors/publisher

No results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.

• MESA SleepNSRR polysomnography + wrist actigraphy (Multi-Ethnic Study of Atherosclerosis)Evaluation2,237 subjects · PPG · ECG · IMU