Computational fluid dynamics (CFD) simulations of airflow in the human airways have the potential to provide a great deal of information that can aid clinicians in case management and surgical decision making, such as airway resistance, energy expenditure, airflow distribution, heat and moisture transfer, and particle deposition, as well as the change in each of these due to surgical interventions. However, the clinical relevance of CFD simulations has been limited to date, as previous models either did not incorporate neuromuscular motion or any motion at all. Many common airway pathologies, such as obstructive sleep apnea (OSA) and tracheomalacia, involve large movements of the structures surrounding the airway, such as the tongue and soft palate. Airway wall motion may be due to many factors including neuromuscular motion, internal aerodynamic forces, and external forces such as gravity. Therefore, to realistically model these airway diseases, a method is required to derive the airway wall motion, whatever the cause, and apply it as a boundary condition to CFD simulations. This paper presents and validates a novel method of capturing in vivo motion of airway walls from magnetic resonance images with high spatiotemporal resolution, through a novel combination of non-rigid image, surface, and surface-normal-vector registration. Coupled with image-synchronous pneumotachography, this technique provides the necessary boundary conditions for dynamic CFD simulations of breathing, allowing the effect of the airway's complex motion to be calculated for the first time, in both normal subjects and those with conditions such as OSA.