Abstract
Background: Tracheal breathing sounds (TBS) have demonstrated strong potential as a non-invasive, wakefulness-based diagnostic tool for obstructive sleep apnea (OSA); yet the relationship between specific upper airway anatomical features and the resulting TBS spectra remains insufficiently understood. This study aims to enhance the diagnostic utility of TBS in OSA by investigating how the upper airway anatomy influences TBS spectral characteristics. Method: Patient-specific computational models of the upper airway were reconstructed from high-resolution CT scans of a healthy subject and an individual with OSA. Additional variants were generated with targeted constrictions at the velopharynx, oropharynx, and trachea, based on clinically reported anatomical ranges. Airflow dynamics were simulated using Large Eddy Simulation (LES), and the resulting acoustic responses were computed via Lighthill's acoustic analogy within a hybrid aero-acoustic framework. Results: Oropharyngeal constriction generated the most spatially concentrated vorticity patterns among single-region constricted models. Airway Resistance analysis revealed that severe velopharyngeal and oropharyngeal constrictions contributed most to regional airway resistance. Spectral analysis showed that velopharyngeal narrowing produced a progressive downward shift in the third resonance peak (1000-1700 Hz), while oropharyngeal narrowing induced an upward shift of the third peak and a downward shift of the fourth peak (1700-2500 Hz). These frequency shifts were attributed to the effective role of acoustic mass and airway compliance. Conclusions: Anatomical modifications of the upper airway produce region-specific changes in both flow and acoustic responses. These findings support the use of TBS spectral analysis for non-invasive localization of airway obstructions in OSA.