Development of a three-dimensional model of the human respiratory system for dosimetric use
© Rosati Rowe et al.; licensee BioMed Central Ltd. 2013
Received: 7 December 2012
Accepted: 25 April 2013
Published: 1 May 2013
Skip to main content
© Rosati Rowe et al.; licensee BioMed Central Ltd. 2013
Received: 7 December 2012
Accepted: 25 April 2013
Published: 1 May 2013
Determining the fate of inhaled contaminants in the human respiratory system has challenged scientists for years. Human and animal studies have provided some data, but there is a paucity of data for toxic contaminants and sensitive populations (such as children, elderly, diseased).
Three-dimensional modeling programs and publicly available human physiology data have been used to develop a comprehensive model of the human respiratory system.
The in silico human respiratory system model, which includes the extrathoracic region (nasal, oral, pharyngeal, and laryngeal passages), the upper airways (trachea and main bronchi), the tracheobronchial tree, and branching networks through alveolar region, allows for virtually any variation of airway geometries and disease states. The model allows for parameterization of variables that define the subject’s airways by integrating morphological changes created by disease, age, etc. with a dynamic morphology.
The model can be used for studies of sensitive populations and the homeland security community, in cases where inhalation studies on humans cannot be conducted with toxic contaminants of interest.
Morphologically realistic models of human organ systems have become critical to research advancement in a changing health research world of dwindling research funding, the push to phase out animal toxicology studies , and the need for immediate answers to protect human health. Research is also challenging in such areas as the homeland security community, where inhalation studies on humans cannot be conducted with contaminants of interest (e.g., B. anthracis, ricin, etc.) because of toxicity. An ideal solution is thus provided by a morphologically realistic computational model of the human respiratory system which can respond to the dynamic changes of respiratory mechanics and abnormal pathologies.
Precursors to this morphologically realistic computer model included numerous mathematical and flow models that looked at the deposition of particles in the human lung . There have also been a few nasal and oropharyngeal models investigating the movement of particles in the nasal passages and mouth/pharynx region [2, 3]. Research has even recently advanced to produce a three-dimensional model from computed tomography data that included a simulated oral cavity, pharynx and larynx, and seven generations into the airways . The latter model used computational fluid dynamics (CFD), including large-eddy simulations, to determine regional particle deposition, but was limited due to its lack of morphological realism, e.g., the lack of nasal passages, oral passage with tongue and teeth, uvula, five lobe lung, 23 branching generations, etc.). Recent work by Corley et al.  has developed a 3-D CFD airway model that extends from the external nares or mouth to the bronchiolar region of the human lung based on CT imaging of the head and torso of a female volunteer, although the airway only extends to the 9th generation due to the limitation posed by the resolution of the CT scanner.
Until now, no model has combined a physiologically based nose, larynx, pharynx, mouth, and lungs to create a three-dimensional morphologically realistic model of the human respiratory tract from the nares to the alveoli. The current model can be used to simulate inhalation, deposition, and exhalation of contaminants, and is progressing towards the consideration of age, race, gender, and health. This virtual human respiratory system includes the extrathoracic (ET) region (nasal, oral, pharyngeal, and laryngeal passages), the upper airways (trachea and main bronchi), the trachea bronchial tree, and branching networks down through alveolar region.
The current morphology modeling of the upper airway system began with the Visible Human Project’s (VHP) Brigham and Women’s Hospital (BWH) head image data. The acquisition of transverse MRI and Computed Tomography (CT) image data of the head of a 72-year-old male subject, with cryosections at 0.174 mm intervals and photographed at a resolution of 1056 × 1528 pixels, was performed at BWH, Harvard Medical School, under contract to the National Library of Medicine (NLM). This dataset is available to Virtual Human Project license holders and the images can be found in the directory BWH_Harvard when logged on to the NLM image server.
The structure of the nasal and oral cavities, pharynx, epiglottis, larynx, and esophagus of this subject were created separately using the marching cubes algorithm  and extracting threshold isosurfaces.
The nasal cavity is composed of two narrow passages that are separated by the nasal septum. Each nasal passage features three curved fin-like airway protrusions known as the superior, middle, and inferior meatus. Extending further into the airways, the two passages merge at the distal end of the nasal cavity and form the beginning of the nasopharynx. The floor of the nasal cavity is formed by the hard palate, which is also the roof of the mouth.
The final stitching of the nasal passage surface segments was completed by rebuilding the profile curves, hand manipulating the curve degree and number of edit points on adjacent surfaces segments, and essentially combining two shapes or segments creating a surface between them manually lofting the surfaces to form a closed watertight topology free from intersecting and twisted surfaces (Figure 6).
The resulting model was free of surface defects and was sufficiently smooth such that the resulting model is accurate to the original isosurface mesh obtained from the marching cubes algorithm . To combine the models effectively, we use merging, sewing, and hole-filling tools that provide flexibility in the joining of the surface segments.
To create a watertight extrathoracic model suitable for CFD studies, additional surface anomalies were corrected for all segments. These included non-manifold surfaces; different details, structures, and resolution of surfaces; matching surfaces; holes in the mesh at critical junction points; and other issues that occurred throughout the stages of combining the separate parts into a functional and morphologically-realistic model. A non manifold surface occurs when more than two surface elements (triangles) intersect on one edge.
To develop a robust method of generating lung morphologies for CFD studies, we have integrated the most promising aspects from numerous modeling techniques. The dynamic surface modeling technique we have developed uses data from idealized models to construct three-dimensional computer simulations of tubular airway structures within lungs. These anatomically accurate computer representations of human lungs are in a format such that their graphical displays may be used in the medical arena, and the resulting airway geometries can be used in CFD analysis. This robust morphology generation system can generate three-dimensional tubular airway structures using surface modeling techniques from existing anatomical data.
To describe the three-dimensional structure within lungs, each airway in the branching network must be uniquely defined in terms of its spatial coordinates. Therefore, each airway is represented by its parameterization into length, diameter, and angles of orientation. The orientation angles consist of the bifurcation angle between two branching airways, and the rotation angle, which is a measure of the rotation of the plane containing the branching airways. We use the morphology to describe a symmetric, dichotomously branching system in which a parent airway branches into two daughter airways. This morphology was developed based on work by Weibel , Soong , Yeh and Schum , Phalen et al. , and Lovelace Biomedical & Environmental Research Institute . It accounts for the inter-subject variability present in a typical adult population by examining probability distributions of the airway data. It is divided into 24 generations beginning with the trachea (generation 0), and ending at the alveolar sacs (generation 23). A mean value for each airway parameter is assigned for each generation. In our model, the first three generations are asymmetric representing the actual geometry of the trachea and large bronchii. Beyond these generations, the symmetric branching data is used.
The airway lengths and diameters are taken from a typical adult lung. To be anatomically realistic, we have used the branching angles, which are based on previous measurements of morphometric data [7–11], and we assumed a constant rotation angle of 90° throughout the airway network. Simply stated, the morphology is generated automatically from anatomical data, and can be controlled and modified in a scientific context within the bounds of that data.
Generating surface models of airway morphologies using anatomic data involves abstracting the task of creating a complex lung model, containing over 16 million airways, to the knowledge-based parameterization of its constituent airways. The tubular (right circular cylinder) structure of an airway is defined by its length and diameter, with its position defined by the orientation angles. The dichotomously branching system of airways that comprise the lung can be treated as a contiguous series of cylindrical tubes connecting at Y-shaped bifurcations. In creating the surface models of the airways, notable complexities arise in the formations of the bifurcation shapes (i.e., where the airways intersect). In this implementation, the models are complex surfaces composed of high-level NURBS surface patches. These surface patches are then stitched together both along the opposing lengths and around the ends where they connect with their parent or daughter airways. Each surface patch is built from a relatively coarse mesh of control points generated by algorithms using the basic length, diameter, and angle input parameters. The density of these control points is higher at the junction of the bifurcations so that the transition region between airways can be adjusted to be either smoother or sharper than shown.
The surface modeling techniques we have employed to generate the morphologies allowed us to create smooth connecting airways and realistic carinal ridge shapes, thus obtaining anatomically accurate representations of the bifurcating systems. This has been shown to be very important to lung airstreams and particle deposition , and were necessary to obtain realistic geometries that could then be further refined and exported to CFD software to study airflow properties and deposition patterns of inhaled aerosols in the lung. The NURBS surface model was used as a reference mesh for polygonal meshes built in Maya.
The smooth mesh preview steps only changes the display of the mesh; this enables the user to visualize how the mesh will appear when the mesh is smoothed. The original mesh remains in its original cubic form until the final mesh is converted to a fully smoothed equivalent in a subdivision proxy form. This is done using the Smooth Mesh Preview to Polygons tool which converts a Smooth Mesh Preview version of a polygon mesh to an actual polygon mesh. The attribute settings for the Smooth Mesh Preview are used when converting to the polygon mesh.
The polygon surface mesh obtained from the marching cubes algorithm  resulted in a highly dense mesh of 498577 vertices, 1411868 edges and 921622 faces. As our ultimate goal was to have a deformable mesh, it was desirable to have a less dense mesh that would allow us to morph the model to vary race, gender and age.
Current development of the model includes the addition of physiologically realistic cartilaginous rings; a parameterized implementation of automatic generation of bifurcations; and the addition of FaceGen (Singular Inversions, Toronto, ON) software to create realistic three-dimensional human faces adjusted for age, race, gender, and other controls, or fit specifically to a photograph. These faces can be applied to the developed polygonal meshes; internal oral and nasal structures can be attached and manipulated based on facial characteristics. These advances are expected to be fully incorporated into the current model by September of 2013. In addition to the current male model, an adult female model and a child/infant model are under development using the same techniques presented in this paper.
A state-of-the-art, morphologically realistic model of the human respiratory tract from the nares to the alveoli has been developed that simulates inhalation, deposition, and exhalation of contaminants. It includes the upper respiratory tract, the extrathoracic region (nasal, oral, pharyngeal, and laryngeal passages), the upper airways (trachea and main bronchi), the tracheobronchial tree, and branching networks down through alveolar region.
We are currently parameterizing variables that define the subject’s airways so that the model can integrate morphological changes created by respiratory disease, exposure to toxins or stressors, and age, along with a dynamic morphology that mimics the changes in the structures during a typical breathing cycle. The model will allow for virtually any variation of airway geometries and disease states. Such flexibility will be critical to investigating sensitive populations, such as children, the diseased, and the elderly. The model will also be critical to the homeland security community where inhalation studies on humans cannot be conducted with contaminants of interest (e.g., B. anthracis, ricin, etc.) because of toxicity.
The model may be used to predict dose from exposure to hazardous particulate-based contaminants, such as anthrax, ricin or particulate-based hazards. It may also prove useful in targeting pharmaceuticals to the appropriate location in a specific individual’s respiratory system.
The authors would like to acknowledge the contributions of Dr. Ted B. Martonen and Dr. James S. Brown. Throughout his esteemed modeling career, Dr. Martonen developed ground-breaking models of the human lung and extrathoracic regions. His mentorship and invaluable guidance fostered the development of the current model. As an expert in aerosol science and respiratory physiology, Dr. Brown provided great insight into the path that the current model should take, and was a valuable advisor during model scoping activities.
This research and manuscript was supported and funded by the US Environmental Protection Agency.
The views expressed in this paper are those of the author[s] and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.