Developing a 3D-printing-based method to create anthropomorphic dosimeters for radiotherapy-delivery verification

Anthropomorphic three-dimensional (3D) dosimeters can be useful for verification of radiotherapy delivery. The aim of this study was to develop a 3D-printing-based method for creating anthropomorphic 3D dosimeters. Internal structures were created using water dissolvable 3D prints as negatives. External structures were 3D-printed, and a mould was produced using silicone rubber. Realistic liver and trachea dosimeters with both internal and external anthropomorphism were produced and subsequently irradiated with photons and protons. A 3%/3 mm gamma analysis resulted in 87% and 86% pass rates. The limiting factor to the resolution of the dosimeters was the 3D prints detail.


Purpose
Anthropomorphic phantoms play an important role for testing clinical equipment in conventional radiotherapy and proton therapy.They are usually constructed to imitate the head, thoracic, or abdominal region.They should ideally be compatible with both magnetic resonance (MR) and computed tomography (CT) scanners, be deformable and able to move in a respiratory manner [1][2][3][4][5][6].
Anthropomorphic phantoms cannot measure the dose themselves, but they can be combined with anthropomorphic 3D dosimeters.Other groups have developed head phantoms containing a 3D polymer gel dosimeter for proton therapy [7] and time-resolved 3D dosimetry using a combined MR linear accelerator system [8].However, the regions of the body experiencing the highest degree of motion are the thorax and abdomen.In this region, considerable and partly respiratory driven organ deformations are common.Creating an anthropomorphic phantom and dosimetry system that can move and deform as the abdominal or thoracic region has the potential to enhance our understanding of how intrafractional and interfractional motion and deformations perturbs dose distributions.
In previous studies, a radiochromic silicone-based 3D dosimeter was proven suitable for both conventional radiotherapy [9][10][11] and proton therapy [12][13][14] even while being deformed [15,16].The dosimeter can be cast into any shape with a high degree of detail [17] and has even been 3D printed directly [18].If the dosimeters also contain air cavities complex air-tissue interfaces could be studied experimentally.Therefore, the ideal 3D dosimeter is deformable, anthropomorphic, can contain air cavities, be embedded in anthropomorphic phantoms, and work for both photon and proton irradiations.
The first aim of this study was therefore to develop a 3D-printing-based method for producing anthropomorphic silicone-based 3D dosimeters which can be deformed during irradiation, with and without air cavities.Particularly relevant are the liver in the thorax and the mediastinal volume surrounding the trachea in the upper abdomen.The second aim of the study was to test the performance of the dosimeters by irradiating them and compare the measured dose with the calculated dose distributions.

Internal structures
Two different tracheas were 3D printed using a polyvinyl alcohol (PVA) water dissolvable filament (S1 in Figure 1a).The 3D prints were applied as "negatives" in their mould thereby creating internal structures.The first print was a simple trachea resembling a Y-shaped structure made to fit into a cylindrical 50 mm x 50 mm mould (Figure 1b).The second print was made from a 3D model of a trachea which was printed to fit into a 75 mm x 74 mm Ø cylindrical mould.The Y PVA print was immersed in boiling water for 20 seconds to smoothen its surface (S2).
The PVA prints were suspended in their moulds which was filled with the dosimeter mixture (S7).A weight was placed on top of the PVA 3D print to prevent it from floating.After curing, the PVA was punctured while submerging the dosimeter in warm water.Pliers were carefully applied to remove chunks of the dissolving PVA whereafter the PVA residue was rinsed off with water (step 8, Fig. 1a).

External structures
The external structures were 3D printed using a polylactic acid filament (S3).The liver dosimeter print was divided into three parts to ensure that the parts could be read-out by the optical CT scanner, which could scan objects up to about 10 cm in size.Each 3D printed part was sanded down, coated with a layer of primer, and spray-painted to provide a smooth surface (S4).A casting basin was made from sculpting clay and filled with silicone moulding rubber.The 3D print was suspended in the basin and the moulding rubber was left to cure for 24 hours and thereafter removed (S5).The mould was filled with the dosimeter mixture (S6).A specially designed nut was suspended in the mixture during curing for later attachment in the optical scanner.

Irradiation
The dosimeters were CT scanned for treatment planning in Eclipse.The liver and Y dosimeter were irradiated with conventional photon beams using a Varian TrueBeam accelerator.The Y dosimeter received two opposing 16 mm x 16 mm fields to a maximum dose level of 10 Gy.The liver treatment plan consisted of six conventional 6 MV photon fields (15 mm x 15 mm) evenly spaced on a 180degree arc with a prescribed maximum dose level of 5 Gy.A four-degrees-of-freedom couch correction was performed using a cone-beam CT.The trachea dosimeter received a two-field spot scanning proton plan using a Varian ProBeam system with a prescribed maximum dose level of 16 Gy.

Read-out and data analysis
The absorptivity of the dosimeters was read-out pre-and post-irradiation using an optical CT scanner with 1000 projections.The projections were reordered to have the same starting point in the pre-and post-read-out after which they were image registered in MATLAB.Using the ordered-subsets-convex total-variation reconstruction method led to 1 mm 3 voxels of the absorptivity.
The measured dose distributions (except the Y) were compared to the calculated distributions using a global 3%/3 mm 3D gamma analysis including all voxels in the datasets above 10% of the maximum calculated dose.

Results
Both methods developed to produce anthropomorphic dosimeters proved successful.The PVA filament was easy to remove without damaging the dosimeters.The detail of the dosimeters and silicone rubber moulds was fine enough that the layers of the 3D prints were visible on them.
Comparing the measured and calculated dose led to a gamma pass rate of 87% for part 2 of the liver dosimeter and 86% for the trachea dosimeter (Fig. 2).

Discussion
The developed method using 3D printing and silicone rubber was capable of producing anthropomorphic dosimeters and would work for structures with overhang of material and very intricate details only limited by the 3D prints.The dosimeters measured the expected dose distribution and could be useful in an anthropomorphic phantom.
For the photon irradiation the low gamma pass rate was primarily due to a low delivered dose.The read-out system has a limited dynamic range and if the dosimeter becomes too large then no light would reach the detector at the thickest projections which would impede the dose reconstruction.When using protons the measurements usually require a quenching correction due to dose-rate and linear energy transfer dependencies [13,14] which require a Monte Carlo simulation.For this the Hounsfield units of the dosimeter must be overwritten with -64 HU on the CT scans to get the correct stopping power [19].However, the overwritten dosimeter structure in the treatment planning system could not be transferred to the Monte Carlo simulation and the quenching correction was therefore not applied.Furthermore, aligning the pre and post projections lead to some artefacts near the boundary of the dosimeters (see Fig. 2).If the dosimeter formulation could be altered to be tissue equivalent it would improve the applicability and transferability into clinical use.Also, the advantage of anthropomorphic dosimeters is limited if the size cannot exceed 10 cm since many organs are larger than that.
If the dosimeter could be printed directly with the dosimeter material as investigated by Wheatley et al [18] it would greatly decrease the production time.However, the method presented in this work does not require a specialized 3D printer and, in the case of the external moulds, these are reusable.
In future work, the 3D liver dosimeters will be used in proton spot scanning to investigate a protocol-specific plan during deformation in an anthropomorphic motion phantom [20].Furthermore, replacing the radiochromic component with an optical stimulated luminescence component to make the dosimeters reusable is being investigated [21,22] which would likely increase clinical applicability.
In conclusion, we have successfully created a method to produce anthropomorphic dosimeters and shown they can measure the planned dose.

Figure 1
Figure 1.a) A flowchart depicting the steps included in producing the anthropomorp hic dosimeters.b) Silicone rubber moulds for the dosimeters (top row), the corresponding 3D prints or models (middle row) and the final 3D dosimeters (bottom row).