End-to-end quality assurance for Volumetric Modulated Arc Therapy with Fricke-Xylenol orange-Gelatin gel dosimeters and dual-wavelength cone-beam optical CT scanner

The whole treatment process undergone by patients in clinics with Volumetric Modulated Arc Therapy (VMAT) can be tested by implementing 3D end-to-end (E2E) quality assurance with gel dosimetry. In this work, a 3D E2E test was performed in a head phantom for the verification of a VMAT treatment, using FXG (Fricke-Xylenol orange-Gelatin) gel dosimetry and a newly developed dual-wavelength reading method on a cone-beam optical CT scanner. This dosimetric method intends to enable accurate measurements in the out-of-field zone and in the tumor volume, with an effective dose range up to 10 Gy. CT images of the phantom with a gel flask were used to create a treatment plan with a brain tumor of complex shape. A very good agreement between 90 %, 80 %, 60 % and 40 % isodose curves and high 3D γ passing rates (2%/2mm) of 98.6 % and 96.7 % between measured and computed dose maps showed that E2E tests can be successfully implemented with this novel dosimetric method.


Introduction
Volumetric Modulated Arc Therapy (VMAT) is a technique developed to fit tightly the target volumes and spare as much as possible the surrounding healthy tissues during treatment.Used for stereotactic irradiations, VMAT can deliver complex dose distributions characterized by non-uniform doses and steep dose gradients.3D end-to-end (E2E) quality assurance is therefore a necessary tool to test the whole treatment process that patients undergo [1].Gel dosimeters are promising candidates to perform this test, as they are able to measure 3D dose distributions with a high spatial resolution.Among them, Fricke-Xylenol orange-Gelatin (FXG) gel with optical CT readout presents the advantage of being easy to implement.However, it is only accurate on a limited dose range up to 3 Gy [2,3].Therefore, a novel dual-wavelength reading method of FXG gel dosimeters has been developed to increase the applicability to a dose range up to 10 Gy, and become clinically representative of dynamic radiotherapy applications [4].
In this study, FXG gel dosimetry and the associated dual-wavelength reading method applied to the cone-beam optical CT scanner Vista16™ (ModusQA) have been used to implement a 3D E2E test in a head phantom on a Novalis TrueBeam STx accelerator (Varian).

FXG gel preparation
The FXG gel composition selected for this study provides a high-sensitive linear dose response up to 10 Gy [4].The FXG gel dosimeters are prepared with ultrapure deionized water, 5 %wt gelatin from porcine skin (gel strength 300, Type A), sulfuric acid to reach a pH of 1.6, 0.3 mM ferrous ammonium sulfate and 0.09 mM xylenol orange.They are poured into Teflon-FEP cylinders of 3.9 cm diameter and 6.3 cm height.One batch of gel is prepared for a whole experiment.The gel dosimeters are prepared one day before their irradiation and are placed in the irradiation room at least eight hours prior to irradiation, for thermal stabilization and threshold dose lowering [4].

FXG gel dose calibration
Ten FXG gel flasks were irradiated at known doses on the 0.25 -10 Gy dose range for calibration curves establishment.TRS 398 recommendations for irradiations in terms of absorbed dose to water under reference conditions were followed.A 67.5 x 64.5 x 56.0 cm 3 water tank and a 6 MV FFF (Flattening Filter Free) photon beam with a dose rate of 1400 MU/min were used with the Novalis accelerator.It was previously verified that the gel calibration coefficient was not dependent on the irradiation conditions (whether in a water tank or in an anthropomorphic head phantom).

Treatment planning and irradiations
The STEEV phantom (CIRS) was used for 3D E2E quality assurance tests (Figure 1 (a)).A waterequivalent plastic insert (outer dimensions 9.4 x 6.4 x 6.4 cm 3 ) was designed to fit a FXG gel flask into the phantom (Figure 1 (b)).CT images of the head phantom containing a gel flask, and covered with a thermoplastic mask (used for immobilization) were acquired using a CT scanner (Somatom, Siemens Healthineers).A 360° two-arc VMAT treatment plan with a calculation grid spacing of 1.25 x 1.25 x 1 mm 3 was created using Eclipse software (Varian) with the AcurosXB algorithm.A heterogeneous dose distribution was prescribed to a brain tumor of complex shape with three dose levels (4.9 Gy, 5.5 Gy and 6.5 Gy) to apply a dose gradient in the Planning Target Volume (PTV).The maximum dose delivered was 6.9 Gy (Figure 2).The phantom was precisely repositioned prior to irradiation using the ExacTrac system (Brainlab).Two gel flasks were irradiated with this treatment plan.Irradiations for FXG gel dose calibration and E2E tests were performed within one day to have the same thermal and spontaneous oxidation histories in the gel flasks.

Dual-wavelength cone-beam optical CT readout
The new dual-wavelength reading method has been applied to the cone-beam optical CT scanner Vista16™ for the optical readout of the FXG gel flasks [4].For the E2E QA measurements, the low dose values measured at 590 nm have been combined to the high dose values measured at 633 nm with a dose threshold between the two dose maps set at 2.5 Gy.A pre-irradiation reference scan and a postirradiation data scan of the dosimeter were acquired at both wavelengths, each one with 2000 projections.The Feldkamp-Davis-Kress (FDK) algorithm and a Hamming filter were applied for the reconstruction with voxels of dimensions 0.5 x 0.5 x 0.5 mm 3 .The matching liquid in the tank was a mixture of 10 %wt propylene glycol -deionized water.

Dose distribution analysis
Linear calibration curves  =   × Δ +   ( = 590 nm or 633 nm) were established on the 0.25 -4 Gy and 2 -10 Gy dose ranges at 590 nm and 633 nm respectively.The optical attenuation coefficients Δ in the previous equation correspond to the average values of the voxels contained in cubic regions of interest (ROIs) of 1 cm 3 at the center of the gel calibration flasks.
For the gel flasks irradiated with the VMAT treatment plan, a marking was made at their top to facilitate registration of the measured dose maps to the calculated one.Measured and computed dose distributions were analyzed with in-house programs developed using Python software.Cylindrical ROIs of 34 mm diameter and 53.5 mm height, corresponding to the volumes of gel inside the flasks, were selected in the dose maps.The comparison between measured and calculated dose distributions was performed by displaying isodose curves, dose profiles and Dose-Volume Histograms (DVHs) in the gel flasks, and by conducting 3D and 2D local γ-index analysis with 2%/2mm passing criteria.The PyMedPhys library was used in Python to conduct the γ-index analysis [5].

Results and Discussion
90 %, 80 %, 60 % and 40 % isodose curves are displayed in Figure 3 in the axial, sagittal and coronal slices passing through the tumor volume.A very good agreement between the isodose curves is observed, with only a slight discrepancy between 40 % isodose curves of maximum 2 mm.In addition, excellent 3D γ passing rates of 98.6 % and 97.6 % between computed and measured dose maps when using both gel flasks have been obtained.High 2D γ passing rates in the axial, sagittal and coronal slices passing through the tumor volume are also found between 92.8 % and 99.3 %.The failing points are mainly located in the regions where doses are below 3.5 Gy.
Figure 4 (a) represents an example of the measured and computed dose profiles along the frontback axis passing through the tumor volume.Low dose differences, below 2 %, are reported in the tumor volume between computation and measurement.However, higher dose differences (up to 7 %) are found for doses below 3.5 Gy.Nevertheless, the measured and computed DVHs of the gel flasks display a close agreement in the whole 0 -7 Gy dose range (Figure 4 (b)).
The discrepancies found for doses below 3.5 Gy are not due to the dual-wavelength reading method as the change of reading wavelength was applied at 2.5 Gy, meaning that these discrepancies were observed at both wavelengths.They are not due to diffusion effects in the gel matrix either, as dose The most probable reason may be that the treatment cannot be exactly delivered as planned in these regions, as significant dose constraints were applied to display a complex dose distribution.However, these discrepancies would not have been observed if less strict γ passing criteria had been selected (i.e.3%/2mm with a global γ-index analysis), which are commonly used for QA tests [6].Therefore, these discrepancies remain lower than the requirements for validating this treatment plan in clinical routine.Nevertheless, their cause will be investigated further with 2D film-based measurements.

Conclusions
The high γ passing rates and very good agreement between isodose curves, dose profiles and DVHs between computed and measured dose maps show that 3D E2E quality assurance for VMAT based on FXG gel dosimetry and dual-wavelength optical CT readout can be effectively implemented.

Figure 1 .
Figure 1.(a) STEEV phantom with thermoplastic mask for 3D E2E quality assurance and (b) waterequivalent plastic insert designed to fit the gel flask into the head phantom.

Figure 2 .
Figure 2. Axial, sagittal and coronal views of the computed dose distribution overlaid with the CT images of the head phantom and a gel flask.

Figure 3 .
Figure 3. 90 %, 80 %, 60 % and 40 % isodose curves of the measured and computed dose maps in the axial, sagittal and coronal slices passing through the tumor volume in a gel flask (in background: measured dose distribution).In black: calculated isodose curves; in gray: measured isodose curves.

Figure 4 .
Figure 4. (a) Measured and computed dose profiles along the front-back axis passing through the tumor volume and corresponding dose differences between computation and measurement (in blue).(b) Measured and computed DVHs of a gel flask.Black lines: computation; red dashed lines: gel measurements.