Numerical Simulation of Air Flow in the Human Nasal Cavity Kezhou Wang, 1 Thomas S. Denney Jr., 1 Edward E. Morrison, 2 Vitaly J. Vodyanoy 2 1 Department of Electrical and Computer Engineering/Auburn University/200 Broun Hall, Auburn, Alabama, 36849-5201 USA 2 Department of Anatomy, Physiology and Pharmacology/Auburn University/109 Greene Hall, Auburn, Alabama, 36849-5518 USA dennets@eng.auburn.edu Abstract — A highly automated method was used to con- struct numerical models of six 3-D human right nasal cavi- ties from computed tomography data. Steady state airflow simulations were performed with computational fluid dy- namics software for quiet breath on the six models. The method is validated with particle simulations. Simulation results from each of the six studies are compared. I. Introduction Knowledge of the airflow field in nasal cavities is essen- tial to understand the basic functions of the nose, such as air transportation and odorant sensation. Over the years, the airflow field has been studied with both experimental models and numerical models. The nasal cavities are im- aged with either computed tomography (CT) or magnetic resonance imaging (MRI). Although anatomically accurate nasal models can be obtained, it is extremely difficult to obtain measurements from the small and complicated nasal models. To overcome these difficulties, some researchers constructed scaled-up physical models [1], [2]. Schreck et al. [1] studied airflow with a three times enlarged plastic model of a half-nasal cavity based on MRI data. Hahn et al.[2] used a 20 times enlarged model of a healthy right human adult nasal cavity constructed from CT scans to study airflow patterns. Recently, numerical models [2], [3], [4], [5] were devel- oped to study the fluid flow in the nasal cavities. A 2-D steady laminar flow in the nasal valve was simulated in [4]. Elad et al. [5] simulated steady laminar flow with a 3-D simplified model. An anatomically accurate numerical model of flow in the human nose was constructed by Key- hani et al. [3]. Their numerical model based on the same CT slices used to construct the 20X model in [2], and a high degree of similarity was observed between the two models. But ,the 3D numerical model in [3] was constructed man- ually, which is a tedious and error prone process. In this paper, we present a highly automated technique for constructing numerical models of a human nasal cavity from CT data. This method includes following steps: 1) The CT image data is interpolated in the slice direction to match slice thickness with the imaging plane resolution. 2) The nasal airway is segmented using a region-growing algorithm. 3) A 3D volume mesh is constructed from the segmented airway. 4) Elements in the mesh are adjusted or deleted to improve the numerical accuracy of the simu- lation. Human interaction is only needed in segmentation step to correct the region-growing algorithm near narrow boundaries. Six numerical models of human nasal cavi- ties are developed with this automatic technique, and air flow was simulated with a commercially-available computa- tional fluid dynamics (CFD) software package. We discuss differences between airflow in the six studies which, to our knowledge, has not been practical before now. The nomenclature of the nasal anatomy used in this pa- per is same as in [2], [3], [6]. The nose is separated me- dially by the nasal septum into two equal size cavities. The external opening is called naris or nostril. Just be- yond the external naris is a funnel shaped dilated region called vestibule. The funnel leads to a region referred to as the nasal valve, where the airway is shaped like a nar- row slit. The main nasal passages begin from the end of the nasal valve where the cross-sectional area of the airway increases, and extend about 5 cm to the posterior end of the cavity. Beyond the vestibule, the main airway can be divided into different parts, the main channel and three wing-like channels called the inferior, middle, and superior meatuses (see Fig.1). The slit-shaped region in the supe- rior part between the nasal septum and the lateral wall of the main nasal passage is the olfactory airway, where the surface is covered with olfactory epithelium. II. Methods A. Image Acquisition The nasal cavities of six human subjects were imaged with CT. The slice thickness was 3mm, and the in-plane resolution was either 0.27mm×0.27mm (studies 1, 5, and 6) or 0.31mm×0.31mm (studies 2, 3, and 4). B. Image Interpolation and Segmentation In imaging modalities such as CT, a 3D sample is im- aged with a sequence of thick slices, where slice thickness is usually three or more times of the in-plane resolution. To obtain a data set with approximately isotropic resolu- tion, slices were interpolated between acquired slices us- ing the hierarchical spline interpolation (HSI) algorithm [7]. The interpolated image data is then segmented with a region-growing algorithm to extract the airway. The Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference Shanghai, China, September 1-4, 2005 0-7803-8740-6/05/$20.00 ©2005 IEEE. 5607