Abstract ID: 34 Comparison of two Monte Carlo calculation engines for proton pencil beam scanning

Abstract

Two independently configured Monte Carlo simulation engines for pencil beam scanned proton therapy have been compared in both geometric setups and real patient data. Both are based on Geant4.10.02.p01 and have been tuned with the same commissioning data. One however uses Gate7.2, the other TOPAS3.0.p1. In addition, different physics lists were chosen; the QGSP_BIC reference physics list (Gate) versus the Topas default physics list. The source planes also differed, being placed at the MU chamber in Gate, and at the nozzle exit in Topas. Finally, the preabsorber was included as a physical component in Topas, while in Gate, it was modeled by modifying the beam input parameters. To compare the two systems, single spots (70–230 MeV with and without preabsorber), simulated in homogeneous volumes of air, water, bone and brain (defined using elemental compositions), were first calculated in both systems. Then 7 clinical fields were compared to verification measurements taken in water using a PTW-2D-ARRAY-XDR (T10031). Finally, the same 7 fields were recalculated in the relevant CT’s. Initially, differing default ionization potentials in Gate and Topas led to range differences for single spots of 1.5–2.1% and absolute differences in CT simulations of 1.8–2.4%. Matching ionization potentials, and retuning the models, resolved these range differences however. Using gamma-analysis to compare clinical fields simulated in water, 99.9% of points agreed between the two simulations at 2%/2 mm, and 100% and 79% agreed relatively with QA measurements for 3%/3 mm and 2%/2 mm respectively. Simulated absolute doses were however 1.0–2.7% lower than measurements. For simulations in CT geometries, 95% of points agreed between both calculations @2%/2 mm relatively, and to within 1% for absolute dose. With the different preabsorber implementations, relative dose distributions in water agreed to better than >98% (2%/2 mm) between the simulations, but since proton loss in the preabsorber is not taken into account in the Gate simulations, absolute doses differed by 7–8%. With this work, the importance of correctly defining ionization potentials and modeling beam modifying devices as physical devices has been highlighted. Excellent agreements between the dose distributions resulting from the two setups and between Monte Carlo and measurements have been shown.