Compatibility of 3D printing materials and printing techniques with PAGAT gel dosimetry

Polymer gel (PG) dosimetry enables three dimensional (3D) measurement of complex dose distributions. However, PGs are strongly reactive with oxygen and other contaminations, limiting their applicability by the need to use specific container materials. We investigate different 3D printing materials and printing techniques for their compatibility with PG. Suitable 3D printing materials may provide the possibility to perform PG dosimetry in complex-shaped phantoms. 3D printed and PG-filled test vials were irradiated homogenously. The signal response was evaluated with respect to homogeneity and compared to the signal in already validated reference vials. In addition, for the printing material VeroClear™ (StrataSys, Eden Prairie, USA) different methods to remove support material, which was required during the printing process, were investigated. We found that the support material should be used only on the outer side of the container wall with no direct contact to the PG. With the VeroClear™ material a homogenous signal response was achieved with a mean deviation of relative to the reference vials. In addition, the homogeneous irradiation of an irregularly-shaped gel container designed with the same printing material and technique also lead to a homogenous PG response. Furthermore, a small field irradiation of an additional test-vial showed an accurate representation of steep dose gradients with a deviation of the maximum position of relative to the reference vial.

arbitrary geometries. This may be of great advantage to test new radiotherapy treatment techniques (Schreiner 2015, Kamomae et al 2017, Oh et al 2017, Yea et al 2017.

Polymer gel
In this study, the PAGAT-(PolyAcrylamide Gelatin gel fabricated at ATmospheric conditions) PG was used as it can be produced in-house at low costs, under atmospheric conditions and has a small dose rate dependence (De Deene et al 2006). The gel consists of a gelatin matrix (6% w/w Gelatin, 300 bloom, SIGMA Aldrich), enriched with two different monomers (2, 5% w/w acrylamide and 2, 5% w/w N,N′-methylene-bis-acrylamide) as active components. Due to the high reactivity of the gel with oxygen the gel was flushed with nitrogen for 5 min to reduce the amount of dissolved oxygen in the gel (De Deene et al 2002). Directly afterwards, 5 mM bis[tetrakis(hydroxymethyl)phosphonium] chloride (THPC) was added as an antioxidant to further reduce interactions with oxygen. After production, the PG was filled into small vials made of different materials (see section 2.2). Before filling, the vials were flushed with nitrogen, and were sealed with Parafilm 'M' Laboratory film (Bemis, Neenah, USA) afterwards to reduce the influence of penetrating oxygen. Additionally, the vials were enwrapped in aluminum foil to protect the gel from light (Koeva et al 2009), placed in a desiccator, which was flushed with nitrogen for 10 min and stored in a refrigerator at 4 • C for 20 − 24 h. 4 h prior to irradiation, the vials were removed from the refrigerator to allow for adaption to room temperature.

3D printing material and printing techniques
To test the usability of different 3D printing materials, a first set of test vials (figure 1(a)) was designed having a similar size and shape as BAREX ™ containers (table 1). BAREX ™ (VELOX GmbH, Hamburg, Germany) vials were already verified for compatibility with PAGAT dosimetry (Mann et al 2017) and therefore used as reference in this study.

Support material
The PolyJet/MultiJet printing technique requires the use of support material if structures with an overhanging shape shall be printed since the subsequent layer is printed while the previous layer is still liquid. After the printing is completed, this support material has to be removed. For the VeroClear ™ material, different removal methods were tested: (i) purely mechanically by means of a water jet or (ii) by applying additionally a 2% sodium hydroxide (NaOH) lye for several hours to degrease the material and to remove material residues (Stratasys 2013), (iii) in addition, a printing method was tested that employs the support material on the outside rather than inside of the vials to avoid contact with the PG. In this case, the vials were printed in two separate parts, which were then glued together using the same printing material and by curing the interface of both parts with UV-light for 30 min. In case of the VisiJet M3 Crystal ™ material, removal of the support material required heating to 55 • C and residual support material was dissolved in a bath of sunflower oil (3DSystems Product 2012). Afterwards the vials were cleaned with a degreasing agent.

Irregular shapes
As the BAREX ™ vials are available only in a single size and shape the purpose of this study was to find a 3D printing material compatible with PG dosimetry that allows designing gel containers in arbitrary geometries. Based on the previous investigations, the most promising technique for 3D printing and support material handling was selected to design irregularly-shaped gel containers ( figure 1(b)).

Irradiation experiments
The gel-filled test vials were irradiated with a clinical 6 MV linear accelerator (Linac) (Artiste, Siemens Healthineers, Erlangen, Germany) using a dose rate of 3 Gy min −1 measured under reference conditions at 5 cm depth and a source-axis-distance of 100 cm. Dose calculation was performed with the Raystation treatment planning system (RaySearch Laboratories, Stockholm, Sweden) and the dose delivery has an accuracy of about 0.5% for the setup of our experiment. For irradiation, the printed test vials were inserted into a water-filled cylinder phantom (Mann et al 2017). The centre of the test vials was positioned to the isocentre marked by the in-room laser system (LAP GmbH Laser Applikationen, Lüneburg, Germany). After irradiation, the vials were wrapped in Aluminum foil and stored at room temperature. As a reference, all irradiations were repeated with a gel-filled BAREX ™ container under identical conditions. Two different irradiation field geometries were applied: (a) Homogenous irradiation. Two opposing and equally-weighted beams (90 • and 270 • ) with a field size of 10.0 × 10.0 cm 2 were used to prescribe a total dose of 4 Gy to the centre of the container leading to a homogeneous dose distribution over the whole volume of the BAREX ™ reference vial. The homogenous irradiation was performed for all vials printed with different materials (table 1) and removal techniques of support material (section 2.2.1). The irregularly-shaped container was irradiated under identical conditions. (b) Small-field irradiation. Based on the results in (a), the most promising printing technique and support material handling was further investigated. For this, three equally-spaced beams (0 • , 120 • and 240 • ) with a field size of 1.0 × 1.0 cm 2 were applied, prescribing a maximum dose of 5 Gy to the centre of the PG within the BAREX ™ reference vial. The high dose gradients were located within the PG.

MR imaging
Approximately 48 h after irradiation, the gel containers were imaged on a 3 T Biograph mMR (Siemens Healthineers, Erlangen, Germany). To avoid influences of temperature differences on quantitative R 2 measurements, the containers were scanned within a water-flow phantom allowing for temperature constancy within ±0.1 • C (Mann et al 2017). The phantom was placed inside a 16-channel head/neck coil and scanned using a multi spin-echo sequence with 32 equidistant echoes with echo times TE = 27.5 − 880.0 ms and an echo spacing of 27.5 ms. The signal-to-noise ratio (SNR) was optimized to SNR ≈ 290 (SNR = R 2 /σ with mean R 2 value R 2 in an exemplary region of interest within the BAREX ™ reference and the corresponding standard deviation σ).The scans were performed with a resolution of 1.0 × 1.0 × 1.0 mm 3 , band width of BW = 130 Hz/pixel and a repetition time TR > 4000 ms to exclude influences of T 1 -relaxation. For comparison of the different MR images, an additional high-resolution (0.5 × 0.5 × 0.5 mm 3 ) 3D-image of the gel containers was acquired, which was used for registration purposes. For this, a standard true fast imaging sequence with steady state precession (TrueFISP) (Scheffler andHennig 2003, Chavhan et al 2008) as implemented by the MRI vendor was applied with the following imaging parameters: TR = 5.43 ms, TE = 2.72 ms, number of averages = 2, and a flip angle of 30 • .  (table 1). The VeroClear ™ vial (a) was printed in two separate parts and glued together afterwards without using support material on the inner side of the vial. 100% refers to the average R 2 -signal in the BAREX ™ vial.

Figure 3.
Relative transversal profiles for the homogenous irradiation of the VeroClear ™ material when using different techniques for the support material. The support material was removed (a) mechanically using a water jet, (b) by applying additionally NaOH, (c) by using the support material only on the outer side of the vial. 100% refers to the average R 2 -signal in the BAREX ™ vial.

Post processing
The MR data was processed on a personal computer using an in-house developed Matlab (The Mathworks Inc., Natick, USA)-based PG evaluation tool (Mann et al 2017) to calculate the spin-spin relaxation rate R 2 = 1/T 2 . To compare the R 2 -profile between the different materials, MR images were co-registered by means of a pointbased 3rd order B-Spline interpolation algorithm using three uniquely defined points as indicated by external markers (Beekly Medical, Bristol, USA). This was done with the image processing platform MITK (Nolden et al 2013).

Printing material
The relative R 2 -profiles of the homogenous irradiation are displayed for the tested materials using a representative transversal slice (figure 2). The VeroClear material showed a homogenous profile with a mean deviation of (−1.2 ± 0.4) % relative to the BAREX ™ reference material (n = 279 voxel). The whole evaluated volume (15 slices) within the PG revealed a mean deviation of (−1.4 ± 0.6) % (n = 4072 voxel). The maximum difference between voxels in the two vials was < 3%. In contrast, the signal for PLA ™ , PVB ™ and VisiJet ™ decreased in the regions close to the walls of the vial (figures 2(b), (c) and (f)). In the vials printed with the stereolithographic technique (Clear ™ & High Temp ™ , figures 2(d) and (e)), perforations of the container wall were found in the vial during the filling with PG (see discussion).

Support material
Based on the promising results in section 3.1.1, VeroClear ™ was evaluated in more detail. Figure 3 shows a comparison of the transversal profiles for the homogeneously irradiated VeroClear ™ material when using different techniques for the removal of the support material. Only the gluing technique without the use of support material on the inner side of the vial showed a good agreement with the BAREX ™ reference material, while the other techniques exhibit a lower signal, which decreased further towards the wall of the vials. Based on these results, the VeroClear ™ material without the use of support material on the inner side of the vials was investigated in further experiments.

Irregular shape
The homogeneous irradiation of an irregularly-shaped container printed with the VeroClear ™ material revealed a homogenous signal response and showed a similar profile as the BAREX ™ reference vial (figure 4), whereas the absolute signal was about 3% smaller in the 3D printed container compared to the reference (see discussion). Figure 5 shows the transversal and sagittal profiles of the small-field irradiation for the BAREX ™ and VeroClear ™ material without the use of support material on the inner side of the vial. The profiles are well comparable and the maximum position shows only minor deviations of < 1 mm.

Discussion
In this work, it has been shown that the VeroClear ™ material in combination with the Objet30 Pro 3D printer (StrataSys) can be used to produce containers, which are compatible with the PAGAT polymer gel. With this material, a homogenous irradiation lead to a uniform signal response with only small deviations (< 3%, figure 2(a)) relative to the BAREX ™ reference vials. All other tested materials showed a signal decrease in the vicinity of the container wall and can therefore considered as incompatible with the use of PG. This 'walleffect' may be explained by an oxygen-permeability of the materials, which leads to a partial inactivation of the PG. In case of the stereolithographic printing technique (Clear ™ & High Temp ™ ) tiny holes with a diameter of approx. 1 mm were found in the material. These holes may origin from the mechanical removal of support struts required for the printing process.
Using the VeroClear ™ material allows printing of arbitrary gel containers. However, this was only possible, if no support material was used on the inner side of the container wall during the printing process as direct contact of the support material with the gel lead to a change of the signal response (figure 3). Most likely, this change is a result of chemical reaction of the PG with either residual support material or with remaining contaminations of the NaOH lye used for the removal of the support material (Baldock et al 2010).
For containers with varying cross section it is necessary to use support material and for PG container production this is still possible as long as the support material is used only on the outer side of the gel containers. This, however, requires printing of the containers in two separate parts, which have to be glued together afterwards (see section 2.2.1). The gluing uses the same printing material and after curing the interface by UV-light, a highly homogenous signal response similar to that of the BAREX ™ reference vial was obtained (figure 3). The advantage of 3D printing is the generation of arbitrarily-shaped gel containers as demonstrated by the irregular container ( figure 1(b)), for which a homogeneous irradiation still leads to a homogeneous signal response (figure 4).
Compared to the BAREX ™ reference vial, the homogenous irradiations of the regularly-shaped test vials revealed a mean signal difference of (−1.4 ± 0.6) % (figures 2(a) and 3(c)). In the irregularly-shaped container a slightly larger difference of 3% was found (figure 4). This larger deviation may result from differences in the volume and shape of the containers, leading to a different temperature equalization and as a consequence to a difference in the chemical polymerization rate (Sedaghat et al 2009). Especially, the temperature during and after irradiation may influence the polymerization rate (De Deene and Vandecasteele 2013). Depending on the gel container size, a small temperature difference may equalize differently leading to a small offset in the gel response. However, for relative dosimetry performed in this work, an offset in R 2 is of not critical.
In addition, using the VeroClear ™ test vials, steep gradients could be measured with high accuracy (figure 5) and the position of the maximum signal agreed well with that in the BAREX ™ reference vials with deviations of < 1 mm.
Having identified a combination of a 3D printing technique and a printing material, which is compatible with PG, it is now feasible to design almost arbitrarily-shaped gel containers. This is a very important feature to perform 3D dose verification in various geometric or anthropomorphic phantoms. These phantoms can be used for QA measurements and end-to-end tests, especially when new treatment techniques are introduced. One important example is MR-guided radiotherapy (MRgRT) (Lagendijk et al 2008, Fallone et al 2009, where it is intended to adapt the treatment plan at each fraction to compensate for changes of the patient anatomy. Validating the adapted treatment plan with PG dosimetry in a clinically relevant setting could be an important application of the proposed method.

Conclusion
In this study, the compatibility of 3D printing materials and printing techniques with PAGAT gel dosimetry was investigated. The VeroClear ™ material has been identified as a suitable material, when the support material is used only on the outer side of the container during the printing process. For this, the container has to be printed in two parts and glued together afterwards. Relative PG measurements in homogeneous irradiation fields revealed an agreement of the gel response of < 3% when compared to measurements in the BAREX ™ reference container, if similarly shaped test vials are used. Using this method, also steep dose gradients can be measured accurately.