FRoG: An independent dose and LET_d prediction tool for proton therapy at ProBeam® facilities

Particle therapy is becoming increasingly available world-wide for precise tumor targeting, its favorable depth dose deposition, and increased biological damage to tumor tissue compared to conventional photon therapy. As demand increases for improved robustness and conformality, next-generation secondary dose calculation engines are needed to verify treatment plans independently and provide estimates for clinical decision-making factors, such as dose-averaged linear energy transfer (LET_d ) and relative biological effectiveness (RBE).

FRoG (Fast dose Recalculation on GPU) has been installed and commissioned at the Danish Centre for Particle Therapy (DCPT). FRoG was developed for synchrotron-based facilities and has previously demonstrated good agreement with gold-standard Monte Carlo simulations and measurements. In this work, additions and modifications to FRoG's pencil beam algorithm to support the ion beam delivery with cyclotron-based technology as used at the DCPT, range shifter (RS) implementation, and robustness analysis methods are presented. FRoG dose predictions are compared to measurements and predictions of the clinical treatment planning system (TPS) Eclipse (Varian Medical Systems, Palo Alto, United States of America, CA, v.13.7.16) in both homogenous and heterogeneous scenarios using a solid-water/water and a half-head anthropomorphic phantom, respectively. Additional capabilities of FRoG are explored by performing a plan robustness analysis, analyzing dose and LET_d for ten patients.

Mid-target measurements in spread-out Bragg Peaks (SOBP) were on average within -0.19% ± 0.30% and ≤0.5% of FRoG predictions for irradiations without and with RS, respectively. Average 3%/2mm 3D γ-analysis passing rates were 99.1% for ~200 patient plan QA comparisons. Measurement with an anthropomorphic head-phantom yielded a γ-passing rate >98%. Overall, maximum target differences in D_02% of <2% between the TPS and FRoG were observed for patient plans. The robustness analysis study accounting for range, delivery, and positioning uncertainties revealed small differences in target dose and a maximum LET_d VH_02% (LET_d received by 2% of the volume having dose larger than 1% of maximum dose) values below 10.1 keV/µm to the brain stem.

We demonstrate that auxiliary dose calculation systems like FRoG can yield excellent agreement to measurements comparable to clinical beam models. Through this work, application of FRoG as a secondary engine at third party cyclotron-based particle treatment facilities is now established for dose verification as well as providing further insight on LET_d and variable RBE distributions for protons, currently absent from the standard clinical TPS.