Evaluation of candidate template beam models for a matched TrueBeam treatment delivery system

Abstract Purpose To explore candidate RayStation beam models to serve as a class‐specific template for a TrueBeam treatment delivery system. Methods Established validation techniques were used to evaluate three photon beam models: a clinically optimized model from the authors’ institution, the built‐in RayStation template, and a hybrid consisting of the RayStation template except substituting average MLC parameter values from a recent IROC survey. Comparisons were made for output factors, dose profiles from open fields, as well as representative VMAT test plans. Results For jaw‐defined output factors, each beam model was within 1.6% of expected published values. Similarly, the majority (57–66%) of jaw‐defined dose curves from each model had a gamma pass rate >95% (2% / 3 mm, 20% threshold) when compared to TrueBeam representative beam data. For dose curves from MPPG 5.a MLC‐defined fields, average gamma pass rates (1% / 1 mm, 20% threshold) were 92.9%, 85.1%, and 86.0% for the clinical, template, and hybrid models, respectively. For VMAT test plans measured with a diode array detector, median dose differences were 0.6%, 1.3%, and 1.1% for the clinical, template, and hybrid models, respectively. For in‐phantom ionization chamber measurements with the same VMAT test plans, the average percent difference was −0.3%, −1.4%, and −1.0% for the clinical, template, and hybrid models, respectively. Conclusion Beam model templates taken from the vendor and aggregate results within the community were both reasonable starting points, but neither approach was as optimal as a clinically tuned model, the latter producing better agreement with all validation measurements. Given these results, the clinically optimized model represents a better candidate as a consensus template. This can benefit the community by reducing commissioning time and improving dose calculation accuracy for matched TrueBeam treatment delivery systems.


| INTRODUCTION
Various resources are available to help guide the development of a new beam model in a treatment planning system. The TG-106 report 1

from the American Association of Physicists in Medicine
(AAPM) provides information on the use of phantoms and detectors to acquire the measurement data typically used to generate a beam model. For TrueBeam treatment delivery systems (Varian Medical Systems, USA), multi-institutional commissioning data have been reported and serve as a reference. [2][3][4] Additionally, the AAPM has provided recommendations on end-user beam model validation through reports for MPPG 5.a 5 and TG-119. 6 Despite this commissioning guidance, articles reporting on dosimetry credentialing results from the Imaging and Radiation Oncology Core (IROC) have demonstrated difficulties in creating an accurate beam model, particularly for highly modulated plans. [7][8][9] Furthermore, significant dose calculation differences have been noted when utilizing automated model generation based on closely corresponding beam commissioning data. 10 In addition, with the prevalence of IMRT planning, multi-leaf collimator (MLC) parameter values take on additional importance. 11,12 Historically, radiation oncology clinics have been required to create and validate unique machine-specific beam models in their treatment planning system (TPS) due to variations in treatment delivery system (TDS) performance. In general, this process entails significant effort often under a compressed timeline, and this situation can be exacerbated in multi-vendor environments. Starting with a template beam model may facilitate parameter value optimization, but significant effort is still required for validation and commissioning measurements. Nevertheless, with many newly arriving treatment machines meeting consistent performance specifications, the opportunity exists for use of an effective beam model template to potentially eliminate parameter value optimization and offer a reduced validation workload.
Historically, a TPS user often begins with published data as a starting place when constructing their own clinical beam model.

With the advent of matched TrueBeam systems that satisfy Varian
Enhanced Beam Conformance (EBC) specifications 13,14 for representative beam data, the use of template models may be considered even more tenable. Nevertheless, with a template-based candidate beam model in place, it remains the responsibility of users to validate against a spectrum of test plans that match clinical practice to the extent that any present weaknesses are identified.
RayStation (RaySearch Laboratories, Sweden) MLC model parameter values have been shown to strongly affect its dynamic MLC delivery dose calculation. Multiple studies have analyzed these parameters for Elekta (Sweden) machines. [15][16][17] Clinically used MLC parameter values have also been reported for the Varian Trilogy system. 18,19 For a TrueBeam treatment delivery system, Chen et al. presented a systematic approach for optimizing MLC parameter values based on IMRT QA dose measurements. 20 Additionally, Saez et al. described a procedure using ionization chamber measurements of sweeping gap beams to determine RayStation parameters for both Millennium 120 and HD120 MLC systems. 21 A recent publication by Glenn et al. reported a reference data set featuring the above RayStation MLC parameters as provided by clinical end-users through an IROC survey for different treatment planning systems. 22 The objective of this work is to use established validation techniques to identify the most optimal beam model from three candidates: a clinically optimized model from the authors' institution ("clinical"), the built-in RayStation template ("template"), and a hybrid based on the RayStation template except for substituting average parameter values from a recently published IROC survey ("hybrid"). 22 We performed comparisons of measured versus calculated output factors, dose curves from jaw-defined and MLC-defined static treatment fields, as well as dose from representative VMAT test plans. In the aforementioned IROC survey 22 from 2020, user-submitted values were provided for primary source size as well as a number of RayStation MLC parameters. The leaf tip width (LTW) is used to account for x-ray transmission through the rounded end of an MLC.
The tongue-and-groove width (TGW) accounts for transmission along exposed leaf sides defining an aperture edge. MLC positioning is accounted for as a function of field size using the terms offset, HANSEN and FRIGO | 93 gain, and curvature which are polynomial coefficients in the expression: Output factors in RayStation were calculated under the same setup conditions with a virtual water phantom.
In the comparison with measured percent depth dose (PDD) curves and lateral profiles, TrueBeam representative beam data 14 provided by the vendor were used. These data were acquired at 100 cm SSD using a CC13 ionization chamber (IBA, Belgium). 3 A custom Python script was used to perform local 1D gamma analysis with 2% / 3 mm criteria above a 20% dose threshold. 24 Criteria were based on MPPG 5.a recommendations 5 for basic dose profile comparisons given as AE2% locally in the high dose region with 3 mm distance-to-agreement in the penumbra region. Field sizes ranging from 3 × 3 cm 2 to 30 × 30 cm 2 were again evaluated with a calculation grid size of 1 mm. Dose profiles were assessed at water depths between 1.5 cm and 30 cm.

2.C | MPPG 5.a static beam analysis
In addition to beam model comparisons for jaw-defined fields, an analysis was also performed for various MLC apertures following MPPG 5.a guidelines. 5

3.A | Jaw-delineated beam analysis
VMAT plan with values 2.0-3.8%. All remaining median dose differences reported in Table 6 were ≤1.0%, except for the C-shape target plan with the template and hybrid beam models at 2.9% and 1.9%, respectively.

4.B | Test sensitivity
Jaw-defined output factors (Table 4)  Additionally, for FFF beams, the agreement was generally worse since secondary source contributions were not available for the refinement of in-field and out-of-field profile shape. For PDD curves, similar behavior was observed where the closest agreement with measured data occurred for a 10 × 10 cm 2 field size (95.8-100.0% across beam models). The lowest gamma pass rates were associated with either a 3 × 3 cm 2 field or a 30 × 30 cm 2 field for each beam energy. This behavior comes about in part because the photon energy spectrum within RayStation is specified along the central axis and users can optimize bin weights to achieve the best agreement for an intermediate field (e.g. 10 × 10 cm 2 ). Under this prioritization, low-energy photon contributions from gantry scatter would be expected to contribute more to large field sizes while being attenuated by collimation with small fields. In addition, larger field profile agreement is affected by the kernel no-tilt approximation in the RayStation dose algorithm implementation. The clinical model also prioritized flat output factor correction variation with jaw size over low-shoulder profile agreement for larger field sizes, and this is borne out in the profile gamma behavior seen with field size.
The gamma analysis results for MPPG 5.a dose curves ( Fig. 4 and

ACKNOWLEDG MENTS
The authors gratefully acknowledge the support of Dustin Jacqmin and David Adam for the development of software tools used in this work. The authors would also like to acknowledge Andrew Shepard and Max Belanger for conducting beam validation measurements.

CONFLI CT OF INTEREST
The authors have no relevant conflict of interest to disclose.

D A T A A V A I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.