Fabrication of a compliant phantom of the human aortic arch for use in Particle Image Velocimetry (PIV) experimentation

2.1 Phantom properties

A meta study of measurements of in vivo aortic geometry of healthy individuals was used to determine the dimensions of the phantom [12]. The aortic geometry was obtained from a meta-study of healthy individuals. The distensibility of the phantom was matched to reported values of aortic distensibility [10]. The model was scaled up by 50% to allow manufacture of thicker phantom walls (0.48 mm in the brachiocephalic artery and 0.99 mm in the aortic arch). A preliminary model of the ascending aorta and the brachiocephalic artery was produced (Figure 1).

Figure 1

Internal geometry of the in vivo aorta. The black region depicts test phantom geometry.

2.2 The internal mould

The model was divided into three parts and printed in low density on a Stratasys Dimension Elite using blue ABS plastic with a wall-thickness of 3 mm. The printing yielded a rough surface on the internal mould that was smoothed using 800-grit sandpaper and brief emersions in acetone (3 s followed by 20 s drying). The smoothed internal mould was scanned using an Artec Spider 3D scanner (published resolution of 100 μm) and processed using Artec Studio software. The software allows a comparison between the original CAD model and the manufactured internal mould (Figure 2). The internal mould had good horizontal fidelity but lower vertical fidelity due to the layering process involved in printing. The maximal deviation between the smoothed mould and the computational model was approximately 0.4 mm.

Figure 2

Deviation between CAD model and smoothed printed part (mm).

2.3 The external mould

Since the internal mould geometric deviation approached the proposed wall thickness in places, the external mould was modelled using offset 3D scan data from the internal mould. Geomagic Design X was used to produce STL files with the correct geometry offsets. During import of the 3D scan mesh file, the “Mesh Build Up Wizard” was run and the “Region in Groups” tool was used. On these groups reference planes were inserted to build up enough planes to cover the whole model. On each plane a “Mesh Sketch” with 200 points of intersection of plane and scan file was made and splines interpolated between the intersections curves (Figure 3). A loft feature was applied to these splines to generate the transition geometry from the aortic arch to the brachiocephalic artery. The result is a solid CAD model of the same shape as the mesh that captured the imperfections caused by manufacturing of the internal mould (Figure 6).

Figure 3

Planes and splines in Geometric Design X fitted to the scan data.

A CNC 3-axis machining canter milled the external mould geometry from Ebalta S-Model-board. The contact surfaces of the external mould were coated with an enamel spray paint to improve mould smoothness and limit silicone perfusion during curing.

Figure 4 shows the discrepancy between the finished external mould the computational model. While the greatest discrepancies are observed in the mounting plates and injection hole, the region of interest also contains discrepancies. In particular, there are significant changes in the wall of the aortic arch in one of the mould sections (0.3 mm). This error corresponds to a reduction of 30% of the intended wall thickness for the phantom. All regions of the brachiocephalic artery were well formed.

Figure 4

Difference between computational data and scan of the external mould in mm.

The internal mould was held in position by interference fits at each end of the internal mould (ascending aorta, brachiocephalic artery and peak of aortic arch). The internal mould was flexible enabling placement in the correct position relative to the external mould. Cyanoacrylate adhesive was used to fix the position of the internal mould. Figure 5 shows how the scanned model with the external mould compares to the manufactured parts.

Figure 5

Internal mould in external mould in Geomagic Design X (left) and printed/machined model (right).

2.4 Producing the phantom

The process described in Geoghegan et al. [10] was used to mitigate the incidence of bubble formation in the phantom. Dow Corning Sylgard 184 silicone was injected into the lower port of the mould until it flowed out of the overflow ports indicating full investment of the mould. After 48 h of curing at room temperature (22°C), the external mould was removed. To vacate the internal mould, the phantom and internal mould was submerged in acetone for 48 h.

After PIV experimentation, the mould was dissected and the wall thickness of the mould was measured at 80 points on the aortic arch and 56 points on the brachiocephalic arch. Measurements were made using digital callipers. The smallest wall thickness was effectively zero. This occurred in the brachiocephalic artery (Figure 6). This rendered this part of the phantom useless for experimentation. The greatest wall thickness was 1.47 mm. This was 0.48 mm greater than the desired 0.99 mm in the aortic arch. The median wall thickness of the aortic arch walls was 1.01 mm, the minimum was 0.70 mm and the quartiles were 0.89 mm and 1.20 mm, respectively. The median wall thickness of the brachiocephalic artery was 0.58 mm (IQR 0.22 mm to 0.74 mm). This outcome implies that the internal mould was not centred properly during the pouring process.

Figure 6

Deviation of the silicone mould from the desired wall thickness as measured at the black points.

The phantom refractive index matched the working fluid in Figure 7, which should effectively render the phantom invisible in the images [12]. The appearance of the phantom indicates that the surface finish on the phantom was not particularly smooth. Tactile inspection of the phantom indicated that the internal surface of the phantom was smooth, but the outer surface was rough.

Figure 7

Refractive index matching phantom in the working fluid.

The working fluid velocity field in the radial plane of the aortic arch was calculated using methods defined by Raffel et al. [13]. Figure 8 shows that at the lower edge the velocity is normal to the phantom. However, some vectors confound expected behaviour. In particular, it could be reasonably expected that the flow would be parallel to the phantom walls. It is expected that the surface roughness confounded the laser light plant and yielded erratic velocity fields that had insufficient fidelity to enable clinical research.

Figure 8

Velocity field in the aortic arch phantom.