Accuracy of Artificial Intelligence in Diagnosing Obstructive Sleep Apnea Using Photoplethysmography: Systematic Review and Meta-Analysis.
Brian Sheng Yep Yeo, Esther Yanxin Gao, Joan Ern Xin Tan et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic accuracy of AI-based methods using photoplethysmography (PPG) for obstructive sleep apnea (OSA) in adults. Analyzing 13 studies with 9983 participants, the AI models demonstrated a pooled sensitivity of 79.6% and specificity of 76.5% compared to conventional testing. The findings suggest that AI could serve as a cost-effective screening tool for OSA.
Key findings
- AI trained on PPG achieved a pooled sensitivity of 79.6% and specificity of 76.5%.
- Specificity increased with higher apnea-hypopnea index (AHI) cutoffs: AHI ≥30 had 85.1%.
- Deep learning models showed higher specificity (82.9%) than traditional machine learning (63.6%).
- Overall evidence quality was rated as moderate.
Why it matters
The findings indicate that AI-based PPG methods could provide a more accessible and cost-effective alternative for diagnosing obstructive sleep apnea, which is crucial given the limitations of traditional polysomnography.
What to keep in mind
The study noted limitations such as small sample sizes, potential bias, and underrepresentation of certain geographic regions.