Surgical success for obstructive sleep apnea (OSA) depends on identifying sites of obstruction in the upper airway. In this study, we investigated sites of obstruction by evaluating dynamic changes in the upper airway using drug-induced sleep computed tomography (DI-SCT) in patients with OSA. Thirty-five adult patients with OSA were prospectively enrolled. Sleep was induced with propofol under light sedation (bispectral index 70–75), and low-dose 320-detector row CT was performed for 10 seconds over a span of 2–3 respiratory cycles with supporting a continuous positive airway pressure model. Most (89%) of the patients had multi-level obstructions. Total obstruction most commonly occurred in the velum (86%), followed by the tongue (57%), oropharyngeal lateral wall (49%), and epiglottis (26%). There were two types of anterior-posterior obstruction of the soft palate, uvular (94%) and velar (6%), and three types of tongue obstruction, upper (30%), lower (37%), and upper plus lower obstruction (33%). DI-SCT is a fast and safe tool to identify simulated sleep airway obstruction in patients with OSA. It provides data on dynamic airway movement in the sagittal view which can be used to differentiate palate and tongue obstructions, and this can be helpful when planning surgery for patients with OSA.