Evaluation of differences and dosimetric influences of beam models using golden and multi-institutional measured beam datasets in radiation treatment planning systems

Purpose: The beam model in radiation treatment planning systems (RTPSs) plays a crucial role in determining the accuracy of calculated dose distributions. The purpose of this study was to ascertain differences in beam models and their dosimetric influences when a golden beam dataset (GBD) and multi-institution measured beam datasets (MBDs) are used for beam modeling in RTPSs. Methods: The MBDs collected from 15 institutions, and the MBDs' beam models, were compared with a GBD, and the GBD’s beam model, for Varian TrueBeam linear accelerator. The calculated dose distributions of the MBDs' beam models were compared with those of the GBD’s beam model for simple geometries in a water phantom. Calculated dose distributions were similarly evaluated in volumetric modulated arc therapy (VMAT) plans for TG-119 C-shape and TG-244 head and neck, at several dose constraints of the planning target volumes (PTVs), and organs at risk. Results: The agreements of the MBDs with the GBD were almost all within ±1%. The calculated dose distributions for simple geometries in a water phantom also closely corresponded between the beam models of GBD and MBDs. Nevertheless, there were considerable differences between the beam models. The maximum differences between the mean energy of the energy spectra of GBD and MBDs were −0.12 MeV (−10.5%) in AcurosXB (AXB, Eclipse) and 0.11 MeV (7.7%) in collapsed cone convolution (CCC, RayStation). The differences in the VMAT calculated dose distributions varied for each dose region, plan, X-ray energy, and dose calculation algorithm. The ranges of the differences in the dose constraints were −5.6% to 3.0% for AXB and −24.1% to 2.8% for CCC. In several VMAT plans, the calculated dose distributions of GBD's beam model tended to be lower in high-dose regions and higher in low-dose regions than those of the MBDs' beam models. Conclusions: We found that small differences in beam data have large impacts on the beam models, and on calculated dose distributions in clinical VMAT plan, even if beam data correspond within ±1%. GBD’s beam model was not a representative beam model. The beam models of GBD and MBDs and their calculated dose distributions under clinical conditions were significantly different. These differences are most likely due to the extensive variation in the beam models, reflecting the characteristics of beam data. The energy spectrum and radial energy in the beam model varied in a wide range, even if the differences in the beam data were <±1%. To minimize the uncertainty of the calculated dose distributions in clinical plans, it was best to use the institutional MBD for beam modeling, or the beam model that ensures the accuracy of calculated dose distributions.