The commissioning and validation of Monaco treatment planning system on an Elekta VersaHD linear accelerator

Abstract Accurate beam modeling is essential to help ensure overall accuracy in the radiotherapy process. This study describes our experience with beam model validation of a Monaco treatment planning system on a Versa HD linear accelerator. Data were collected such that Monaco beam models could be generated using three algorithms: collapsed cone (CC) and photon Monte Carlo (MC) for photon beams, and electron Monte Carlo (eMC) for electron beams. Validations are performed on measured percent depth doses (PDDs) and profiles, for open‐field point‐doses in homogenous and heterogeneous media, and for obliquely incident electron beams. Gamma analysis is used to assess the agreement between calculation and measurement for intensity modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) plans, including volumetric modulated arc therapy for stereotactic body radiation therapy (VMAT SBRT). For all relevant conditions, gamma index values below 1 are obtained when comparing Monaco calculated PDDs and profiles with measured data. Point‐doses in a water medium are found to be within 2% agreement of commissioning data in 99.5% and 98.6% of the points computed by MC and CC, respectively. All point‐dose calculations for the eMC algorithm in water are within 4% agreement of measurement, and 92% of measurements are within 3%. In heterogeneous media of air and cortical bone, both CC and MC yielded better than 3% agreement with ion chamber measurements. eMC yielded 3% agreement to measurement downstream of air with oblique beams of up to 27°, 5% agreement distal to bone, and within 4% agreement at extended source to surface distance (SSD) for all electron energies except 6 MeV. The 6‐MeV point of measurement is on a steep dose gradient which may impact the magnitude of discrepancy measured. The average gamma passing rate for IMRT/VMAT plans is 96.9% (±2.1%) and 98.0% (±1.9%) for VMAT SBRT when evaluated using 3%/2 mm criteria. Monaco beam models for the Versa HD linac were successfully commissioned for clinical use.


| INTRODUCTION
Accurate beam modeling plays an important role in the overall accuracy of the radiation therapy treatment process. The International Commission on Radiation Units and Measurements specifies a total dose uncertainty tolerance of 5% in patients, 1 and decreasing dose calculation uncertainty is a means of achieving this goal. Furthermore, the choice of dose calculation algorithm has been shown to have a clinically significant impact on local tumor control rates. For example, in non-small cell lung cancer patients treated with stereotactic ablative radiation therapy a local control benefit was shown for patients whose treatment plan was generated using collapsed cone convolution vs a pencil beam algorithm. 2 This illustrates the importance in accurately commissioning and validating radiotherapy beam models used in the clinic.
Monte Carlo (MC) algorithms are the gold standard for dose computation in radiation therapy. MC dose engines simulate particles, track individual interactions and secondary generated particles, and tally dose deposition in a medium. Interactions at any given simulation step are determined through random number generation, the cross section of the respective stochastic process, and particles or photons are transported until their energy falls below a user-specified cutoff energy. 3 Due to the stochastic nature of these calculations, the calculated dose is subject to statistical uncertainty. In general, relative statistical uncertainty is proportional to the inverse square root of the number of histories generated. Large numbers of histories yield calculations with less statistical uncertainty but at the expense of increased calculation time. 4 The continual progression of computing power has enabled MC calculation times on the order of several minutes, which is clinically acceptable. Multiple manufacturers including Elekta, Varian, Ray-Search, and Accuray now offer treatment planning systems (TPSs) with MC dose algorithms. While many groups [5][6][7][8][9][10][11][12] have published their experience in commissioning MC-based treatment planning models and algorithms, few references on Elekta's Monaco TPS are available. 10,12 Narayanasamy et al. 10

| MATERIALS AND METHODS
The Monaco TPS (version 5.19.03d) was used for all calculations in this study. The Elekta Versa HD linear accelerator investigated in this work delivers photon energies of 6 MV (with flattening filter), 6 FFF (6 MV flattening filter free), 10 MV (with flattening filter), 10 FFF (10 MV flattening filter free), and 18 MV, and electron energies of 6, 9, 12, and 15 MeV. MC models were generated for each energy and modality. Collapsed cone convolution-superposition (CC) models were also created for 6, 10, and 18 MV photon beams, with and without wedges. The CC models simulate the effects of physical motorized wedge, whereas MC models cannot be used for wedged fields in the Monaco TPS.

2.A | Beam data collection and open-field dosimetric verification
All beam scanning was conducted following the guidelines of TG-106. 13  scanning, PDD scans, output factor measurements, collimator scatter factors, and absolute dose measurement all performed at 90 cm source to surface distance (SSD). Photon CC beam models required additional scanning of wedged fields, diagonal scans, and wedge transmission factor measurements. Output factors for field sizes smaller than 5 × 5 cm 2 were measured with the Edge diode detector and daisy-chained to an ion chamber measurement, while ion chamber measurements alone were used for larger field sizes. Electron MC data included profiles in air at 90 and 70 cm SSD, profiles in water at 100 cm SSD, PDD measurements with and without applicators, output factors measured in air without the applicators, and absolute dose measurements. Collimator scatter factors were measured using the formalism provide in AAPM TG-74 14 using acrylic and brass mini-phantoms.
In addition to open field data, the manufacturer provides users a set of eight "Express QA" plans. These fields have been described in detail by Narayanasamy et al. 10

2.A.1 | Profile validation
A CT scan of air was acquired and imported into the Monaco TPS. A 50 × 50 × 50 cm 3 cube was contoured in air with an assigned electron density (ED) of 1.0 for the MC computations, and was set to be treated as water for the CC models (for CC models 1.0 ED is not pure water). Open fields were computed using a 2 × 2 × 2 mm 3 dose grid and a statistical uncertainty of 1.0% per calculation for MC. Open-field dose planes calculated by the Monaco TPS were exported for comparison with scanning data collected during commissioning. ScanDoseMatch (http://www.qxrayconsulting.com/sdm/), an open source scanning data analysis tool, was used to perform gamma analyses between modeled and measured data. 5,15 Gamma analysis was performed using a 2% and 2 mm passing criteria (relative mode) for both photons and electrons.  Additionally, obliquity and extended SSD calculations were tested and compared to dose measurements obtained from a 0.125 cm 3 ion chamber which was cross calibrated against a PTW 30013 ion chamber with a valid ADCL calibration.

2.B. | Nondosimetric testing
A series of nondosimetric tests were performed in accordance with the recommendations AAPM TG-53. 16 Main components of this testing include: generation of CT number to ED curves, contour generation and 3D expansion accuracy, and creation of tolerance tables for use in Mosaiq oncology record and verify system (Elekta Inc. Atlanta, Georgia). Additionally, data export from Monaco to Mosaiq for multiple patient and phantom orientations was tested. CT to ED curves were created in Monaco from results of scanning an Electron Density CT Phantom (Gammex Inc., Middleton, WI) with known ED inserts using various CT kVp values. Additionally, the accuracy of the automatic rigid image registration was tested using datasets provided by AAPM TG-132. 17 These multimodality images include, CT, cone beam CT (CBCT), MRI, and PET. The images were imported with the known offsets provided by TG-132 and the transformation matrix given in the TPS was compared to these known offsets.

2.C. | Plan validation
Plan validations were performed using a Sun Nuclear ArcCheck. The ED of the ArcCheck was set following the recommendations specified by Sun Nuclear for use with Monaco. A virtual ArcCheck phantom was imported into the Monaco TPS with a manual ED override.
The dose from a 10 × 10 cm 2 field size at 100 cm source to axis distance (SAD) was calculated using the TPS and the entrance to exit diode ratio recorded. This setup is then delivered to the ArcCheck using the linear accelerator, and the measured entrance to exit diode ratio must match the calculated ratio to within 1%. Furthermore, the local gamma using 2% and 2 mm distance to agreement (DTA) passing criteria must be greater than 90%. If these criteria are not met, the ED of the virtual ArcCheck phantom must be iteratively adjusted.

2.C.1 | 3D-CRT plan validation
The collapsed cone model was validated by importing a CT dataset of a previously treated prostate cancer patient and generating 3D-CRT plans. One, three-field technique plan incorporating parallel opposed wedged fields and an AP beam, and one four-field box technique plan was created for each CC modeled energy (6, 10, and 18 MV). All plans were delivered to the ArcCheck and evaluated using gamma analysis with absolute dose, global normalization, a low dose threshold of 10%, and passing criteria of 3% and 2 mm. A single point-dose measurement was performed using a 0.125 cm 3 ion chamber inserted into the ArcCheck. Point-dose measurements were scaled by the ratio of mass energy-absorption coefficient of water to that of the medium. 18 This is required because the dose calculated in the virtual phantom is dose to poly(methylmethacrylate) (PMMA) while the dose measured by the ion chamber is related to a dose to water measurement. Failure to do this will lead to a systematic discrepancy in point-dose measurements.

2.C.2 | IMRT/VMAT plan validation
IMRT and VMAT treatments were commissioned using select datasets provided by TG-119 19-21 as well as previously treated patient datasets from our institution. Test plans were selected to cover a range of treatment sites with corresponding energies that would likely be used clinically. The "C shape" CT dataset and structure set were imported into Monaco, and 1 IMRT and 1 VMAT plan were created for each energy (eight total plans on this dataset); target and OAR doses were designed to meet the "harder C shape" objectives. Head and neck and prostate datasets from TG-119 and previously treated patients were also imported and planned with VMAT treatments, where 6 MV and 6 FFF energies were used for head and neck, and 10 MV and 10 FFF were used for prostate. All VMAT plans generated on TG-119 data sets used two full arcs for VMAT plans or nine equally spaced beams for IMRT. The previously treated patient plans used two full 360°arcs.
All of the TG-119 datasets and the clinical head and neck and prostate plans were generated using 2 Gy dose per fraction schemes.

3.A.2 | Point-dose and output factor validation
Open-field dosimetric point-dose validations were performed for both MC and CC models. Seventy-nine (79) dose points were evaluated for each of the five photon energy MC models (395 points total) and 93 dose points were evaluated for each of the three photon energy CC models (279 points total). Across all data, 99.5% and 98.6% of TPS calculations are within the 2% tolerance recommended by Medical Physics Practice Guidelines 5a 24 for the MC and CC algorithms, respectively. All calculated output factors in square fields at the depth of 10 cm for photon MC and CC models are within 2% of hand calculations and the majority is within 1%. Sample data points are shown in Table I

3.C | Plan verification
An ED of 1.144 was determined to best match the material composition of the ArcCheck used at our institution. At this electron density, the local gamma passing rate for a 10 × 10 cm 2 field using 2% and 2 mm criteria is 96.3% and the calculated ratio of entrance to exit diode doses differed from measured by only 0.59%. Both values are within the stated guidelines provided by Sun Nuclear.    the gamma analysis at 3% and 2 mm criteria which exceed the tolerance limit stated in TG-218. In total 94.7% (18/19) of VMAT/IMRT plans measured are within the tolerance limit of TG-218 and no plan measurements exceed the action limit. These results compare favorably to the work of Narayanasamy et al. 10 who, using a Monaco TPS for dose calculation, reported an average gamma passing rate of 95.0% with 3% DD and 3 mm DTA, which are less stringent criteria than reported in this study. This is in good agreement with the 95% confidence limit expectation. All plans were successfully exported to Mosaiq and delivered without any significant deviations to our institutions current workflows.

CONF LICT OF I NTEREST
The authors declare no conflicts of interest pertaining to this work.