Lung radiotherapy verification with a 3D printed thorax phantom and an ionisation chamber array

In this study, a 3D printing error inspired the development of a novel method for using a sagittally-sliced 3D printed thorax phantom to perform dosimetric verifications of lung radiotherapy treatment methods using a 2D ionization chamber array. A full-size thorax model was designed for 3D printing with multiple tissue densities including lung and bone and printed as a series of 2.4 cm sagittal slices using a Raise 3D Pro dual nozzle printer (Raise 3D Technologies Inc, Irvine, USA). An error introduced midway through printing resulted in one half of the phantom being printed at unrealistically high densities. A method was therefore devised whereby the entire phantom was used to plan two lung treatments, one conventionally fractionated and one hypo-fractionated, which were then verified via measurements using an Octavius 729 ionisation chamber array (PTW-Freiburg GmbH, Freiburg, Germany) in combination with several correctly-printed slices of the phantom. The measurements allowed dose distributions in planes through the target, adjacent to the target and at the location key of organs at risk to be verified, for both treatment plans. This method has the potential to be adapted for use with other phantoms and other dosimetry arrays to allow efficient evaluation of future treatment techniques.


Introduction
Dosimetric verification of radiotherapy treatment dose using a humanoid phantom containing lung-like heterogeneities is widely recommended when implementing new lung treatment modalities, and is a component of credentialling for several clinical trials of lung stereotactic ablative radiotherapy (SBRT/SABR) [1,2].These recommendations and requirements have arisen from the known fallibilities of radiotherapy treatment planning systems when calculating dose in low-density media [3,4].
While several generic commercial phantoms are available to facilitate these verification measurements, demands for adaptability and patient-specificity have recently led to a growth in the development of bespoke thorax phantoms [1,2] including phantoms produced using 3D printing techniques [5,6,7,8].Since several groups had produced 3D printed thorax phantoms that had empty space in the lung region, to allow moving components to be placed within [7,8], the initial aim of this study was to produce a 3D printed thorax phantom with a moving component where lung-equivalent material was included in both the moving component and the rest of the pulmonary cavity.However, when this work was interrupted by the COVID-19 pandemic and affected by a 3D printing fault [9], we took the opportunity to instead develop a novel method for using a sagittally-sliced 3D printed thorax phantom to perform dosimetric verifications of lung treatment methods using a 2D ionization chamber array.

Method
A full-size female thorax model was created from CT images of the XCAT computational human phantom [10] by using thresholding to delineate tissues of sufficiently different density to warrant 3D printing with different in-fill percentages.The phantom was divided into 2.4 cm sagittal slices, as suggested by Craft and Howell to minimise thermal warping during fabrication [8].The phantom was also designed with an elliptical cavity and matching insert, which was originally intended to become the moving component.
All 3D printing was completed using a Raise 3D Pro dual nozzle printer (Raise 3D Technologies Inc, Irvine, USA), using different densities of poly-lactic acid (PLA) filament (Shenzhen Esun Industrial Co Ltd, Shenzhen, China) to model all soft tissues including lung and using StoneFil Terracotta filament (Formfutura BV, Nijmegen, Netherlands) to model all bone.A gyroid in-fill pattern was used for lung as recommended by Tino et al [11], while a grid pattern was used to maximise print speed for all denser tissues.In-fill percentages required to replicate the densities of the identified tissue types were derived by 3D printing test cubes at different in-fill percentages and performing transmission measurements with the nominal 6 MV treatment beam, via the method described by Dancewicz et al [12].Complete printing of the 20 cm long thorax section was expected to require 11 kg of PLA and 1 kg of StoneFil (costing approximately € 250) and take 600 h total printing time.
The printing process was interrupted for several months at approximately the half-way point, due to the local impacts of the COVID-19 pandemic, and then recommenced after a firmware update had reset the extruder calibration (e-step) to a default value, resulting in an increase filament extrusion per mm of print [9].In order to salvage this faulty print and produce useful results, one new slice was printed with a faux tumour close to the right chest wall and the phantom was CT scanned in a left lateral decubitus position to allow a novel dosimetry study to be undertaken.Treatments were planned for the decubitus phantom and an experimental method was devised where measurements were performed with an Octavius 729 ionisation chamber array and lower part of an Octavius 2 phantom (PTW-Freiburg GmbH, Freiburg, Germany) replacing the left hand (incorrectly printed) parts of the phantom.
Treatment planning followed standard local procedures, with the unusual orientation of the phantom handled by assuming the "patient" was lying supine and simply rotating the arc start and stop positions by 90 degrees, so that the contralateral lung was avoided as usual.The treatment plans and verification plans are shown in Figure 1, alongside photographs showing the setup of the phantom during CT scanning in its treatment planning configuration and an example verification configuration.
Two treatment plans were created.One was a conventionally fractionated treatment (60 Gy in 30 fractions) to a PTV generated by applying a 10 mm margin to the visible tumour in all directions except superior-inferior where a 20 mm margin was used, as though planning with a 3D CT scan of a breathing patient.The other plan was a SBRT/SABR treatment (48 Gy in 4 fractions) to the visible tumour plus a 5 mm margin on all sides, as though the visible tumour represented a maximum intensity projection image from a 4D CT series.Both treatments were planned using a volumetric modulated arc therapy (VMAT) method, using the Varian Eclipse treatment planning system (Varian Medical Systems, Palo Alto, USA).Inverse planning objectives differed between the two plans, taking into account the two different prescriptions as well as the dose tolerances of the heart, lung, spinal canal, rib and liver OARs.
The verification measurement method consisted of using the Octavius 729 ionisation chamber array to measure dose planes in various sagittal slices.In order to compare the array measurements to the planned dose, the Octavius 729 array was CT scanned three times, with different numbers of phantom slices placed on top.To achieve measurements indicative of the dose to the centre of the target, the left periphery of the target, and the heart and spinal canal OARs, the rightmost four, five and six slices of the phantom were carefully placed on top of the array and aligned using lasers to duplicate their arrangement during the phantom treatment planning scan (see Figure 1).The two lung treatment plans were each copied onto these three array scans and all doses were recalculated and exported for measurement on a Varian iX linac.

Results and discussion
Figure 2 shows the results of the verification measurements for both arcs from the two treatments that were planned for the faux tumour in the 3D printed phantom.The results from measuring with four phantom slices on top of the ionization chamber array confirm the dose calculation accuracy within the target and also illustrate the difference in relative dose falloff between the conventionally fractionated treatment and the SBRT/SABR treatment.Other than the larger size of the conventionally fractionated target, the main contributor to this is the comparative priority of sparing the surrounding lung from the very high (12 Gy) prescribed dose per fraction for the SBRT/SABR target (note the different vertical scales in Figures 2(a)-(b) and (c)-(d)).
The effect of target size is also apparent in the results for five phantom slices, also shown in Figure 2, where dose from the SBRT/SABR treatment plan has started to decrease, while the central region of the array is still encompassed in the prescription dose for the conventionally fractionated treatment.
Beneath six slices of 3D printed phantom, the results in Figure 2 show that the dose from the two treatment plans have substantially decreased relative to their prescriptions, although in this region the Octavius 2 phantom beneath the ionization chamber array is evidently producing increased out-of-field scatter that would not be present in the patient or in a fully realised 3D printed phantom.

Conclusion
By using a 3D printed phantom, a novel method for verifying lung radiotherapy treatment plans using an ionization chamber array was developed.This method has the advantage of allowing instantaneous dose measurements of beams largely delivered though the right-hand slices of heterogeneous humanoid phantom, after being planned with reference to a more complete phantom scan that included relevant OARs.This method has the potential to be adapted for use with other phantoms and other dosimetry arrays to allow efficient evaluation of future radiotherapy treatment techniques.

Figure 1 .
Figure 1.(a) and (b) photographs of CT setup for planning scan and verification scans.(c) and (d) sagittal views of conventionally fractionated and SBRT/SABR treatment plans.(e) and (f) transverse views of conventionally fractionated and SBRT/SABR treatment plans.(g), (h) and (i) transverse views of verification plans for conventionally fractionated treatment.(j), (k) and (l) transverse views of verification plans for SBRT/SABR treatment.

Figure 2 :
Figure 2: Verification results from (a) first arc and (b) second arc, from conventionally fractionated plan, and (c) first arc and (d) second arc, from SBRT/SABR plan, showing results for verification plans with four, five and six slices of phantom on top of the dosimeter array.