Answers are grounded in this paper and cite their sources.
Study Objectives Intermittent hypoxemia is increasingly recognized as a stronger predictor of cardiovascular morbidity and mortality in obstructive sleep apnea (OSA) than the apnea–hypopnea index (AHI). We developed a sound-based AI model to detect oxygen desaturation events and estimate the oxygen desaturation index (ODI) from nocturnal breathing sounds in a fully contactless manner. Methods The model was trained using 2,973 nights of synchronized polysomnography (PSG) and audio recordings obtained from hospital- and home-based settings. Validation and independent testing were performed on 664 and 802 nights, respectively. Audio signals were converted into Mel spectrograms and processed using a Vision Transformer–based architecture. Performance was evaluated at the 30-second epoch level using sensitivity, specificity, precision, F1 score, accuracy, and Cohen’s κ. ODI estimation was assessed using Pearson correlation, Bland–Altman analysis, mean absolute error (MAE), and root mean square error (RMSE). Results For epoch-by-epoch desaturation detection, the model achieved a sensitivity of 0.79, specificity of 0.95, precision of 0.78, F1 score of 0.79, overall accuracy of 0.92, and κ of 0.74. Estimated ODI showed agreement with PSG-derived ODI (r = 0.93, R² = 0.87, p < 0.001). Bland–Altman analysis demonstrated a mean bias of −0.29 events/h with 95% limits of agreement from −14.73 to 14.14 events/h. Conclusions Breathing sound–based deep learning enables contactless detection of oxygen desaturation events and reliable ODI estimation. This approach may provide a scalable, non-contact method for sleep-disordered breathing assessment in real-world settings.
No takes yet. Share an insight, caveat, or question.
Choi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: