Numerical simulations and experimental measurements on flow features in a patient-specific upper airway model with obstructive sleep apnea

Abstract

In this article, the turbulent flow features in a patient-specific upper airway model with obstructive sleep apnea before and after surgery are numerically and experimentally studied by applying the large eddy simulation and laser Doppler anemometry. Two three-dimensional anatomically accurate numerical models are reconstructed based on the CT-scan of one severe obstructive sleep apnea patient for both pre- and post-surgery. Two experimental models of the same upper airway are established by using the three-dimensional printing technique according to the numerical models. The first direct comparison of flow velocity profiles between measurement and simulation at different locations and cross-sections is conducted in two upper airway models. The simulated result indicates that the especially large axial velocity and negative pressure appear in the narrowest cross-section before surgery, but they either reduce or vanish after surgery. Both measurement and simulation can capture the “pharyngeal jet” flow and the reversed flow in two upper airway models, and the agreement between the measured and calculated velocity profiles at different locations and cross-sections is excellent, which gives us great confidence that the large eddy simulation can be a reliable method for predicting the flow features associated with obstructive sleep apnea airway. Although this work represents significant advance, there still exists some limitations, such as the discrepancies between measurement and simulation in the interface of main stream velocity and reverse velocity due to the flow complexity, in the reversed flow region owing to the uncertainty of laser Doppler anemometry measurement at low velocity range, and the incapability of capturing flow near the wall due to the intrinsic feature of laser Doppler anemometry measurement. Besides, it is very difficult to find the corresponding locations exactly between experimental and numerical models, which could also cause the deviation between experiment and simulation.