Numerical Modeling of Nasal Cavities and Air Flow Simulation

Abstract

Computational fluid dynamics (CFD) has many applications in biomedical engineering, such as simulating air dynamics in nasal cavities and lungs, blood flow in vessels, and blood flow in hearts. To perform CFD simulations, numerical models of anatomic structures have to be constructed. The models may be developed from tomographic slices of anatomic structures acquired by medical imaging modalities, such as computed tomography (CT) and magnetic resonance imaging (MRI). However, anatomic structures usually are highly irregular in shape. A mesh with large number of elements is needed to construct an accurate model of an anatomic structure. Manually constructing models would be tedious and error prone. An automatic geometric modeling method is highly desired.

In this dissertation, an automatic numerical modeling technique for nasal cavities and a mathematical model for the shape of the electro-olfactogram (EOG) are developed. Two issues are addressed for numerical nasal cavity modeling. The first issue is that the slice thickness of CT or MRI is usually much larger than the imaging plane resolution,and significant differences are observed between adjacent slices, making it difficult to construct accurate 3D models directly from acquired image slices. This problem is addressed by introducing a hierarchical spline-based image registration method to perform slice interpolation. The second issue is how to automatically generate 3D finite element CFD mesh from the segmented data. This issue is addressed by the development of an automatic mesh generation algorithm, called marching volume elements (MVE). The algorithm is able to generate three-dimensional (3-D) finite element mesh from volume data. Six human nasal cavity models and a dog model were developed with the numerical modeling technique, and air flow simulations were conducted with the developed models. The mathematical model for modeling the shape of electrical responses of olfactory epithelium to odorant stimuli is a linear input-output model. The model is able to predict the shape of the responses to different odorant concentrations for a fixed duration of stimuli. This model has the potential to evaluate olfactory electrical responses and to estimate kinetics of G-protein cascade within the olfactory receptor neuron.