Feasibility of a GATE Monte Carlo platform in a clinical pretreatment QA system for VMAT treatment plans using TrueBeam with an HD120 multileaf collimator

Abstract Purpose To evaluate the quality of patient‐specific complicated treatment plans, including commercialized treatment planning systems (TPS) and commissioned beam data, we developed a process of quality assurance (QA) using a Monte Carlo (MC) platform. Specifically, we constructed an interface system that automatically converts treatment plan and dose matrix data in digital imaging and communications in medicine to an MC dose‐calculation engine. The clinical feasibility of the system was evaluated. Materials and Methods A dose‐calculation engine based on GATE v8.1 was embedded in our QA system and in a parallel computing system to significantly reduce the computation time. The QA system automatically converts parameters in volumetric‐modulated arc therapy (VMAT) plans to files for dose calculation using GATE. The system then calculates dose maps. Energies of 6 MV, 10 MV, 6 MV flattening filter free (FFF), and 10 MV FFF from a TrueBeam with HD120 were modeled and commissioned. To evaluate the beam models, percentage depth dose (PDD) values, MC calculation profiles, and measured beam data were compared at various depths (Dmax, 5 cm, 10 cm, and 20 cm), field sizes, and energies. To evaluate the feasibility of the QA system for clinical use, doses measured for clinical VMAT plans using films were compared to dose maps calculated using our MC‐based QA system. Results A LINAC QA system was analyzed by PDD and profile according to the secondary collimator and multileaf collimator (MLC). Values for MC calculations and TPS beam data obtained using CC13 ion chamber (IBA Dosimetry, Germany) were consistent within 1.0%. Clinical validation using a gamma index was performed for VMAT treatment plans using a solid water phantom and arbitrary patient data. The gamma evaluation results (with criteria of 3%/3 mm) were 98.1%, 99.1%, 99.2%, and 97.1% for energies of 6 MV, 10 MV, 6 MV FFF, and 10 MV FFF, respectively. Conclusions We constructed an MC‐based QA system for evaluating patient treatment plans and evaluated its feasibility in clinical practice. We observed robust agreement between dose calculations from our QA system and measurements for VMAT plans. Our QA system could be useful in other clinical settings, such as small‐field SRS procedures or analyses of secondary cancer risk, for which dose calculations using TPS are difficult to verify.

field SRS procedures or analyses of secondary cancer risk, for which dose calculations using TPS are difficult to verify.

K E Y W O R D S
GATE, HD120, Monte Carlo, TPS QA, TrueBeam 1 | INTRODUCTION Treatment plans of intensity modulated radiotherapy (IMRT) and volumetric modulated arc radiotherapy (VMAT) have dynamic motion of multileaf collimators (MLC), gantry, or dose rate, during dose delivery, and could deliver highly conformal prescribed dose to the target volume while sparing normal volume by modulating intensities. 1,2 Model-based dose calculation algorithms in the commercial treatment planning systems such as anisotropic analytical algorithm (AAA) or the collapsed cone convolution class (CCC) have fast dose calculation time and accuracy as level as clinically acceptable. 3,4 Although accurate beam data measurement and beam modeling can reduce uncertainty of dose calculation in IMRT and VMAT plans, many sources of errors in IMRT planning, including uncertainty of beam modeling, output for small fields, unmeasured out-of-field area, heterogeneity, and so on, still remains. 5 Especially, as VMAT is widely used for stereotactic ablative body radiotherapy (SABR) because of its fast treatment time and high conformality, [6][7][8][9] importance of the patient-specific pretreatment quality assurance (QA) has been increased. 5,[10][11][12] Monte Carlo (MC) simulation is a popular method used in comparative studies to verify the accuracy of dose calculation using commercial treatment planning systems (TPSs). [13][14][15][16][17][18][19] Recently, many methods have been applied to improve the calculation time for Monte Carlo simulation, which is used to verify the accuracy of radiotherapy. 20 GATE v8. 1 which is an open-source toolkit compatible with the Gean-t4 medical application system 21 was released. GATE, which has been mainly used for single photon emission computed tomography (SPECT) and positron emission tomography (PET), is a Geant4-based MC platform with three-dimensional simulation and parallel computation.
In the current study, we developed a QA program which can automatically convert treatment plan files in digital imaging and communications in medicine (DICOM) format into our QA system and export three-dimensional dose in DICOM format for analysis. We used GATE v8.1 as dose calculation engine in the QA system. A TrueBeam with HD120 (Varian Medical Systems, Palo Alto, CA, USA) was modeled and validated as the source of the QA system.
To verify the clinical usefulness of the QA program, dose calculations for VMAT treatment plans using the QA program were performed and compared to dose measured using radiochromic film.

2.A | Design of the QA system
The QA system analyzes plans and converts data into macro files that can be used in GATE. The VMAT plan generated using a TPS exports data to the QA system in DICOM format. The QA system imports plan files and automatically analyzes the treatment parameters, including number of fields, positions of the x and y jaws, positions of the gantry and collimator, and position of the MLC over time for each field and segment. Based on the analyzed information, the QA system creates a folder corresponding to the patient ID and name. It then generates a macro capable of performing dose calculations within the QA system according to number of fields, position of the gantry, field weight, and MLC in the data folder. The MLC is stored with the gantry location information file, with outboard leaf, half-leaf on target, half-leaf on isocenter, quarter-leaf on isocenter, and quarter-leaf on target divided into parts A and B. The macros (consisting of the time sequences) are split into 200 job players and combined into one file after calculation.
To reduce time consumption in MC-based dose calculation, we designed a compact cluster exclusively. The compact cluster is configured with 88 nodes using the Rocks cluster Linux for dose calculation. The 88 nodes are configured for high-performance computing (8-16 GB of RAM and a CPU for Intel i7-3770 and Xeon E3-1220).
The cluster system contains one master node in addition to the calculation nodes. The master node controls the job split, output merge, F I G . 1. Processing diagram for the Monte Carlo-based QA system. QA, quality assurance. job submission, and queue management. 21,22 The master node communicates with the external network to import a treatment plan in DICOM format and perform dose calculation in parallel via a highspeed ethernet switching hub. The constructed cluster submits and manages the macro converted by the QA system using Condor software. 23 This QA system was designed to use time weighting applied within the QA program according to the treatment plan. The Fig. 1 shows the configuration and sequence of the system.

2.B | Linac modeling
To construct a QA system with Monte Carlo-based dose calculation engine, geometries of TrueBeam including target, collimator, and HD120 MLC, provided by the manufacturer under non-disclosure agreement, were modeled 24 using GATE v 8.1. Modeling of HD120 MLC is the most important for calculating absorbed dose with IMRT/ VMAT treatment plan. Geometries from the HD120 MLC were simulated using tessellated volumes from stereolithography (STL) files containing triangular facet data. These data were used to define the surfaces of three-dimensional objects. It is composed of 32 pieces with a thickness of 2.5 mm at the center and 28 pieces with a thickness of 5.0 mm at the periphery.
Although the detailed geometries of target assembly were modeled, we used phase space files provided by the manufacturer. In the phase space files, photons with nominal energies of 6 MV, 6 MV flattening filter free (FFF), 10 MV, and 10 MV FFF were recorded above the jaws. The photon sources were generated from target and incident electrons with mean energies for each source were 6.18, 5.9, 10.7, and 10.2 MeV, respectively. 22 The phase space files were formatted according to recommendation of the International Atomic Energy Agency (IAEA) and have been verified in published studies. 21,25 The average of photon energy is 1.92 MeV and 1.87 MeV for 10 MV and 10 MV FFF, respectively. Because the ratios of photon spectrum with energy over 10 MeV were 0.08% and 0.014%, respectively, hadronic processes was not considered in this study.
In order to validate the modeling of TrueBeam, the three-dimensional dose in water phantom with size of 60 × 60 × 60 cm 3

2.C | Clinical feasibility of the QA system
In order to validate the clinical feasibility, four VMAT plans with photon energies of 6 MV, 10 MV, 6 MV FFF, and 10 MV FFF in stereotactic ablative radiotherapy (SABR) for lung, adrenal gland, liver, and lung cancer, respectively, were enrolled and the characteristics of the plans summarized in Table 1  The MLC shapes for all the constraints were converted automatically using the QA system, as shown in Fig. 4, and absorbed doses for all the VMAT plans were calculated successfully. Fig. 5   to accuracy of dose calculation. 5 And also, tissue and material composition could affect accuracy. 32,33 In order to reduce factor causing dose calculation error, three-dimensional dose using homogenous solid water phantom was calculated.

| RESULTS
In the processes of positioning and delivery, various sources of uncertainties, for example, film positioning, accuracy of leaf motion, and gantry rotation, could affect accuracy of dose delivery. Also, uncertainties of registration and dose conversion 34 in the step of analysis could affect accuracy of the results in this study. In pretreatment QA for VMAT treatment plan, irradiation with fixed gantry angle and composite dose for the all fields were efficient; however, dose error resulted from uncertainties of MLC motion, gantry rotation, dose rate variation, and daily output variation might be obscured. 5 Because the purpose of this study was to evaluate accuracy of dose calculation and feasibility of the MC-based QA system for VMAT treatment plan, we fixed gantry angle to 0 degree in order to reduce and minimize uncertainties of MLC motion and gantry rotation in delivery for VMAT treatment plans. and leakage dose. According to the ICRP 86, the incidence of TPS in radiation therapy is 28%. Therefore, the availability of an additional QA protocol system for TPS is becoming increasingly necessary. 41 As monitor unit (MU) in the use of IMRT or VMAT is usually higher than that in conventional RT, effect of neutron dose in IMRT and VMAT has been studied and reported. [42][43][44][45] Peripheral dose of thermal neutron in VMAT with 10 MV was less than 100 micro-Gray of kerma equivalent. 45 All the mean energies of the photon simulated in this study were less than 10 MeV and only 0.08% for 10 MV was over 10 MeV, we did not consider effect of the neutron dose in film measurement. However, neutron dose generated by the use of photon with higher energy than 10 MV may affect in pretreatment QA and dosimetry in VMAT or IMRT plans.

| CONCLUSION
We established a pretreatment QA system for clinical use of Monte Carlo calculations for TrueBeam and HD120 MLC and confirmed the possibility of using this system for VMAT plans. Our QA system can be used clinically as an additional verification or replacement QA method for VMAT plans, which are increasing. Importantly, our system confirms the accuracy of commercial TPS dose algorithms currently in clinical use. Our system can also be applied in other settings, including calculation of dose accuracy in small fields or secondary cancer risk for out of field which are difficult to assess with TPS.

CONFLI CT OF INTEREST
The authors declare no conflict of interest.