ObjectivesThis study aims to identify characteristics in image‐based computational fluid dynamics (CFD) in children with obstructive sleep apnea (OSA).DesignDiagnostic study.SettingHospital‐based cohort.ParticipantsChildren with symptoms suggestive of OSA were recruited and underwent polysomnography.Main outcome measuresThree‐dimensional models of computational fluid dynamics were derived from cone‐beam computed tomography.ResultsA total of 68 children participated in the study (44 boys; mean age: 7.8 years), including 34 participants having moderate‐to‐severe OSA (apnea‐hypopnea index [AHI] greater than 5 events/h), and 34 age, gender, and body mass index percentile matched participants having primary snoring (AHI less than 1). Children with moderate‐to‐severe OSA had a significantly higher total airway pressure (166.3 vs. 39.1 Pa, p = .009), total airway resistance (9851 vs. 2060 Newton‐metre, p = .004) and velocity at a minimal cross‐sectional area (65.7 vs. 8.8 metre per second, p = .017) than those with primary snoring. The optimal cut‐off points for moderate‐to‐severe OSA were 46.2 Pa in the total airway pressure (area under the curve [AUC] = 73.2%), 2373 Newton‐metre in the total airway resistance (AUC = 72.5%) and 12.6 metres per second in the velocity at a minimal cross‐sectional area (AUC = 70.5%). The conditional logistic regression model revealed that total airway pressure, total airway resistance and velocity at minimal cross‐sectional area were significantly associated with an increased risk of moderate‐to‐severe OSA.ConclusionsThis study demonstrates that CFD could be a useful tool for evaluating upper airway patency in children with OSA.