Artificial intelligence in surgical planning and outcome prediction for obstructive sleep apnea: emerging hype or the future standard?

STUDY

objectivesTo evaluate the emerging role of artificial intelligence (AI) in diagnosis, risk stratification, and adult surgical planning for patients with obstructive sleep apnea (OSA), and to assess its potential clinical value and limitations.

methodsA narrative literature review was conducted A targeted systematic search elements to synthesize recent developments in AI applications across the OSA care continuum. Studies were selected based on relevance to diagnostic accuracy, wearable and home-based sleep monitoring, outcome prediction, and integration into surgical workflows. Special attention was given to evidence involving drug-induced sleep endoscopy, predictive modeling for surgical response, and AI-driven tools validated in real-world or telehealth settings.

resultsAI-powered models demonstrated high concordance with manual scoring of in-laboratory sleep studies, improved accuracy in event detection using wearable data, and effective classification of OSA severity from reduced physiological signals. Predictive algorithms integrating clinical and imaging data enhanced risk stratification and surgical candidate selection. In particular, deep learning models outperformed traditional clinical predictors in forecasting responses to hypoglossal nerve stimulation. However, variability in data quality, lack of pediatric-specific validation, and concerns regarding algorithm bias and transparency remain significant barriers. We emphasize drug-induced sleep endoscopy (DISE) analytics and hypoglossal nerve stimulation (HNS)/MMA outcome prediction, presenting diagnostic AI only insofar as it feeds pre-operative decision support.

conclusionsArtificial intelligence offers powerful tools to support individualized, efficient, and scalable OSA management. Its integration into clinical pathways could optimize diagnosis and treatment decision-making, especially in surgical contexts. Clinical translation will depend on external/temporal validation, calibrated probability outputs, decision-curve/net-benefit, and prospective decision-impact (target selection change, OR time, postoperative outcomes), with equity audits across subgroups. Future efforts should prioritize robust validation, interpretability, and equitable deployment to ensure safe and effective implementation.