The first step to preparing the models for printing was to extract the airway surface from CT scan data. A CT scan takes multiple sequential two-dimensional images of the body density at each image
Airway No Reference Name Age Gender Printing Method Material Volume [ml]
06B Airway 06B1 44 M FDM ABS 53.84
Airway 06B2 FDM ABS
Airway 06B Jetted Photopolymer Acrylic
07B Airway 07B 43 M FDM ABS 60.16
09 Airway 09 41 F FDM ABS 48.27
10 Airway 10 51 M FDM ABS 57.57
slice across the scan volume. The colour of each pixel, ranging in greyscale from black to white, denotes the density measured at the pixel position. Together, the collection of image slices forms a three-dimensional density map. Each of the five airway CT scan datasets chosen had an image slice separation of 0.6 mm, thus preserving a large amount of airway detail. To extract the airway surface from these images the process of segmentation was used. Segmentation is an operation which separates pixels into two groups either side of a nominated pixel density value, known as the threshold. The interface of these two sets of pixels is then extracted as a surface. The selection of the threshold value used to extract each airway was made by trial and error iterating the process until the surface extracted was as complete and correct as possible, when compared to the known upper airway anatomy, without omitting any significant features.
Before the airway surface was edited and developed into a physical model it was first scrutinised to ensure it met a number of important criteria. One of the criteria imposed was that the airway should be healthy and representative of the general population, free of any geometric defects which may adversely affect experimental results. A consequence of this is that most scans which contain the full upper airway are not likely usable as they are taken with the express purpose of mapping the airway to detect irregularities. As there are health risks associated with excess exposure to the CT scanning process, it is uncommon for the entire upper airway to be captured in scans taken of other anatomical features within the same area. These two factors made acquisition of appropriate scan data a difficult task and thus the first check of segmented airways was to ensure all relevant geometric features, from the nares to the top of the trachea, were present and not truncated during the scanning process. A total of five airways of the twelve unique available scans met these criteria.
It was necessary then to edit each extracted airway surface to remove certain extraneous features such as the paranasal sinuses, auditory tubes and the tear ducts (Tortora and Derrickson, 2006). As the sinuses and the auditory tubes are air reservoirs which are only open to the exterior via small ostium into the upper airway, they do not experience significant air movement but rather pressure equalisation. Although the tear ducts are open to the exterior at both ends, the opening of the tear ducts into the nasal cavity is protected by a thin mucous membrane, Hasner's membrane, which helps to prevent flow back into the duct (Müller et al., 1978). These features therefore do not participate in the flow of air through the upper airway during breathing and were removed to reduce the model size and allow greater access to the airway. Each feature was removed as close to the airway volume as possible, where the diameters of the connecting orifices were at minimum. The small holes which remained in the airway surface meshes were capped with domes to provide a smooth continuous surface which would not adversely affect the flow pattern.
Further editing was required to crop each surface so that only the upper airway and some of the facial geometry was present. A small region of the face around the nose was retained as during inhalation air is drawn from a hemispherical region around the nares, the flow profile within which may be affected
by the surrounding geometry (Doorly et al., 2008b). Additionally, a certain amount of the facial geometry below the nares had to be retained to allow the nasal cannulae to be properly fitted to the final models. The inferior portion of the airway surface was truncated so that the passage terminated at the transition from larynx to trachea. Each mesh was therefore an open surface encompassing only the complex upper airway and a portion of the face. The open mesh of the first upper airway, airway 06B, is shown in Figure 7.
Figure 7: Open mesh of upper airway 06B.
Before these meshes could be materialised into physical models, the open surfaces had to be closed to provide a three dimensional volume. This took the form of an irregular box fitted to the shape of the airway, with the facial features forming one face and the inferior outlet of the upper airway exiting on another. The airway formed a cavity through the box which could be manufactured by 3D printing. As the trachea was to be represented in the experimental system as a circular tube with a 15 mm diameter, a smooth transition from the larynx-trachea cross section of the airway mesh was accomplished by forming a lofted surface to a 15 mm diameter orifice on the base of the box.
Because the airways were to be 3D printed it was deemed necessary to make each so the majority of the airway could be easily accessed for inspection and removal of any remaining support material deposited during printing. This lead to the sectioning of each airway model into three parts, a left and right side plus a small central piece containing the nasal septum. An example of these is shown in the form of airway 06B2 in Figure 8.
Figure 8: Three pieces of the airway 06B2 model.
Prior to manufacture, each airway model mesh was prepared for the measurement of static pressures in various locations by the introduction of pilot holes for pressure taps. In the case of the first airway 06B1 model pressure taps were drilled post manufacture without the use of pilot holes and, on inspection, it was found that a number of the taps did not enter the model normal to the airway surface. This introduced error to the static pressure measurements from dynamic pressure effects. Establishing pilot holes for the pressure taps in the model mesh allowed for accurate placement, ensuring each tap entered the correct anatomical region perpendicular to the airway surface. Five to six bolt holes for locating and securing the model pieces together were also introduced to each airway model at this stage.
Some of the later models to be manufactured underwent a final editing process before printing. As large portion of the production costs is for the print material, reduction of material volume was valuable. It was possible to achieve this by performing a shelling operation in which the solid volume was reduced to a shell with a minimum wall thickness of 3 mm. In the case of the airway 06B model further editing was carried out to remove as much of the non-functional material as possible. This left the airway surface, the lower face of the original box and circular bosses for the bolt and pressure tap holes. A comparison of this airway 06B model version to the earlier airway 06B2 version is given in Figure 9 to show the volume reduction.
Figure 9: Reduced material airway 06B model compared to the earlier airway 06B2 model.