Related Experiment Video
Updated: Jan 16, 2026
05:51
Assessing the Accuracy of Fitness Smartwatch Data for Cardiovascular and Physical Activity Monitoring: A Validation Study in Digital Health
Published on: February 21, 2025
1.3K
AI‑Enhanced Smartwatch AHI Estimation and AI‑Scored Polysomnography for Obstructive Sleep Apnea: Real‑World
Donghyeok Kim1, Jeong Yup Han2, Hyunjun Jung2
1Department of Otorhinolaryngology-Head and Neck Surgery, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea.
Nature and Science of Sleep
|September 29, 2025
Summary
This study found an AI smartwatch algorithm accurately estimates the apnea-hypopnea index (AHI), showing high accuracy for moderate-to-severe obstructive sleep apnea (OSA). While effective for detecting severe OSA, the algorithm tends to underestimate AHI in mild cases.
Area of Science:
- Sleep Medicine
- Medical Technology
- Artificial Intelligence
Background:
- Obstructive sleep apnea (OSA) diagnosis relies on polysomnography (PSG), a resource-intensive method.
- Wearable technology offers potential for accessible sleep apnea monitoring.
Purpose of the Study:
- To validate an AI-powered smartwatch algorithm for estimating the apnea-hypopnea index (AHI).
- To compare smartwatch AHI estimations against AI-scored and expert-scored Level 1 PSG in Korean adults.
- To assess the algorithm's inter-ethnic accuracy, as it was trained on South-American cohorts.
Main Methods:
- Simultaneous Level 1 PSG and smartwatch recordings were performed on 90 adults.
- Analysis included 53 datasets with at least 3 hours of valid smartwatch data.
- Agreement was evaluated using Spearman correlation, intraclass correlation coefficients, and receiver-operating-characteristic curves.
Main Results:
- The smartwatch AHI (eAHI) showed strong correlation with AI-scored (aiAHI) and expert-scored (pAHI) PSG (ρ=0.88, ICC=0.87 for aiAHI; ρ=0.85, ICC=0.82 for pAHI).
- For moderate-to-severe OSA (aiAHI ≥ 15 events/h), the algorithm achieved 92.3% sensitivity, 92.6% specificity, and 92.5% overall accuracy.
- Bland-Altman analysis indicated a systematic underestimation of AHI by the smartwatch, particularly in mild OSA cases.
Conclusions:
- The smartwatch algorithm demonstrates high concordance with PSG for estimating AHI, especially in detecting moderate-to-severe OSA.
- The technology shows promise for practical, scalable early identification and monitoring of OSA in real-world settings.
- Further optimization is needed to improve AHI accuracy for mild OSA cases due to scoring and duration calculation limitations.
