Modeling Elekta VersaHD using the Varian Eclipse treatment planning system for photon beams: A single‐institution experience

Abstract The aim of this study was to report a single‐institution experience and commissioning data for Elekta VersaHD linear accelerators (LINACs) for photon beams in the Eclipse treatment planning system (TPS). Two VersaHD LINACs equipped with 160‐leaf collimators were commissioned. For each energy, the percent‐depth‐dose (PDD) curves, beam profiles, output factors, leaf transmission factors and dosimetric leaf gaps (DLGs) were acquired in accordance with the AAPM task group reports No. 45 and No. 106 and the vendor‐supplied documents. The measured data were imported into Eclipse TPS to build a VersaHD beam model. The model was validated by creating treatment plans spanning over the full‐spectrum of treatment sites and techniques used in our clinic. The quality assurance measurements were performed using MatriXX, ionization chamber, and radiochromic film. The DLG values were iteratively adjusted to optimize the agreement between planned and measured doses. Mobius, an independent LINAC logfile‐based quality assurance tool, was also commissioned both for routine intensity‐modulated radiation therapy (IMRT) QA and as a secondary check for the Eclipse VersaHD model. The Eclipse‐generated VersaHD model was in excellent agreement with the measured PDD curves and beam profiles. The measured leaf transmission factors were less than 0.5% for all energies. The model validation study yielded absolute point dose agreement between ionization chamber measurements and Eclipse within ±4% for all cases. The comparison between Mobius and Eclipse, and between Mobius and ionization chamber measurements lead to absolute point dose agreement within ±5%. The corresponding 3D dose distributions evaluated with 3%global/2mm gamma criteria resulted in larger than 90% passing rates for all plans. The Eclipse TPS can model VersaHD LINACs with clinically acceptable accuracy. The model validation study and comparisons with Mobius demonstrated that the modeling of VersaHD in Eclipse necessitates further improvement to provide dosimetric accuracy on par with Varian LINACs.


2.A | Commissioning materials and equipment
At our institution, two Elekta VersaHD LINACs (Elekta Instruments AB, Stockholm, Sweden) were "matched" with each other and commissioned utilizing one mutual beam dataset for treatment planning and delivery.
The VersaHD LINACs were accepted from Elekta with three flattened (6x, 10x and 15x) and two flattening-filter-free (6xFFF and 10xFFF) megavoltage photon energies. The available 6, 9, 12 and 15 megavoltage electron energies were accepted but not commissioned for clinical use.
Each VersaHD LINAC is equipped with a 160-leaf MLC set and a single pair of jaws in the other direction. The MLC consists of two opposed leaf banks with 80 leaves per bank. Each leaf has a projected width of 5 mm at the isocenter level. VersaHD is also equipped with a universal wedge 11 which can be used with the 6x, 10x and 15x open fields. The Varian Eclipse (Varian Medical Systems, Palo Alto, CA) TPS version 13.6 was used to model the LINACs, and the Anisotropic Analytical Algorithm (AAA) was used for photon beam dose calculation. 12 The Blue Phantom 2 water tank (IBA Dosimetry GmbH, Neu-Isenburg, Germany) was used for beam scanning. 13 The tank servo of 48 cm × 48 cm × 41 cm scanning volume was controlled by the OmniPro-Accept 7 software (OmniPro-Accept version 7.4, IBA Dosimetry GmbH, Neu-Isenburg, Germany). 14 The software provides automatic radiation detector navigation and data collection through a motion control unit. The software was also used for beam data postprocessing. All beam scanning and data collection were performed in agreement with manufacturer manuals and AAPM professional guidelines, including AAPM Task Group (TG) Reports No. 45 15 and No. 106 16 . These materials provide detailed recommendations for acceptance testing and beam commissioning measurements, for both regular and small field sizes. The radiation detectors used with the corresponding detector properties and the associated tasks are listed in Table 1.

2.B | Beam data collection: PDDs and profiles
The PDDs and profiles were measured in accordance with the Varian Eclipse TPS reference manual for beam modelling. All measurements were performed at a 100 cm fixed source-to-surface distance (SSD). Field sizes ranged from 1 × 1 cm 2 to 40 × 40 cm 2 and were determined by jaws moving along radial axis direction and by MLCs moving along transverse direction. All fields used for scanning had two pairs of "guard leaves" abutting each other along the field edges under the diaphragm jaws. The other MLC leaves beyond the guard leaves, 1.0 cm from the field edge in the jaw direction, were closed by the LINAC per Elekta's format. These field settings were created, stored as "Quick Beam" or "Stored Beam", and used for beam scanning in the Elekta service mode. The cylindrical chamber position was automatically corrected for the effective point of measurement in the OmniPro-Accept software. Beam scanning using the diode detector did not need the correction. All mandatory and recommended beam data measurements, such as PDD, crossline and inline profiles, were performed for desired field sizes and scanning depths (dmax, 5, 10, 20 and 30 cm), with and without the universal wedge. The diagonal profiles were measured for the maximum field size (40 × 40 cm 2 ) at required scanning depths (dmax, 5, 10, 20 and 30 cm) with open beams. The central axis correction of the water phantom was performed for each beam energy setting prior to measurements. For field size (FS) ≤ 3 × 3 cm 2 , a Sun Nuclear Edge diode detector was used for its high spatial resolution and minimal susceptibility to the partial volume effect. The diode measurements were performed with no reference detector in the step-by-step scan mode as to increase the acquisition sampling time and improve the signal-to-noise ratio. In comparison, the beam data for FSs ≥ 3 × 3 cm 2 were scanned continuously with a traditional dual ionization chamber setup, with field and reference Scanditronix CC13 cylindrical ionization chambers. As measurement conditions change, the scan speeds were adjusted to account for variations in dose rate, field size, and presence of beam modifiers (wedge) ensuring accurate and consistent quality measurements of the beam data. For 3 × 3 cm 2 FS, both diode and cylindrical chamber measurements were performed for cross-validation. For profile scans larger than the dimension of the water tank like 40 × 40 cm 2 inline, crossline and diagonal profiles, half of the profile was acquired and the other half was mirrored. All PDDs and profiles were smoothed using a median filter with a sliding window size of 0.5 cm. The acquired PDDs and profiles of two VersaHD machines were matched within 1% for PDD and 1%/1 mm for beam profiles. The machines were adjusted accordingly until the deviations were smaller than these stated thresholds.  No. 50. 17 Using the same setup, the DLG factor which accounts for transmission through leaf ends, was preliminarily measured for all photon energies using the sweeping-gap technique 18     The plans encompassed a full range of 2D/3D cases and intensitymodulated radiation therapy (IMRT)/volumetric-modulated arc therapy (VMAT) cases to cover our clinical practices. 2D/3D plans included three simple 4-field box plans (field sizes ranging from 3 cm × 3 cm, 10 cm × 10 cm to 20 cm × 20 cm), a breast tangential plan using the field-in-field technique, a breast tangential plan using two wedged fields, a whole brain plan, a 3D lung SBRT plan using noncoplanar static fields, a 3D conformal-arc spine plan, and a 3D pelvis plan using two wedged fields. All wedged plans used the

3.A | Beam data measurements
All essential beam modeling dataset entries were measured and scrutinized before being imported into Eclipse. As an example, Fig. 1(a) shows measured PDD curves for 6x photons for a 3 cm × 3 cm field size. Red curve for ionization chamber (Scanditronix CC13, IC) and blue curve for diode (Sun Nuclear Edge, Edge) detector measurements reveal an excellent mutual agreement. The corresponding inline and crossline beam profiles at 1.5 cm depth are plotted in Fig. 1(b). The IC and Edge profiles

3.B | Beam modeling in Eclipse
The measured data were used as input in Eclipse for autofitting procedure which resulted in a VersaHD beam model. Fig. 3(a) shows measured and modeled 6x photons PDD curves for a 10 × 10 cm 2 field size. Fig. 3

3.C | DLG measurements
The measured MLC leaf transmission factors were less than 0.5% for all energies, ranging from 0.3% to 0.5%. The values were in line with previously reported value 25    the measured doses were consistently higher than the TPS calculated doses. For the IMRT/VMAT plans, there was no defined trend, though the measured doses were slightly lower than the TPS ones.

3.D | Ionization chamber and film QA results
In general, the 2D/3D plans are intrinsically without much modulation, which implies sensitivity to the daily machine output variations.
On the other hand, the point doses of IMRT/VMAT plans are determined by complex mixture of factors including modulation level, DLG, output, and dose gradient across the chamber. As a result, the point dose errors had no distinct trends for IMRT/VMAT plans.

3.E | Mobius commissioning results
Mobius as an IMRT QA tool was initially commissioned only for the flattened 6x, 10x and 15x beams. The flattening-filter-free energies (6xFFF and 10xFFF) were deferred till later as enough clinical plans are accumulated for proper analysis. In interim, IC and film QA system is used for routine IMRT QA. For comparison, in Table 3 Table 3.
A glance at Of note is that the Mobius vs. ionization chamber measurements were suboptimal for 15x. The current model represents a compromise after multiple trials with a realization that it is unachievable to simultaneously optimize both head and neck VMAT plans and liver VMAT plans. Optimizing one site will in turn make the other site worse, which is essentially echoing the scenario within Eclipse and indirectly revealing a modeling deficiency.

| CONCLUSIONS
It is feasible to use Eclipse to model Elekta VersaHD LINACs for dose calculation and treatment planning. Nevertheless, future developments and major model improvements are still warranted for Elekta LINAC models within Eclipse platform to be on par with the corresponding Varian LINAC models.

CONF LICT OF I NTEREST
No conflict of interest.