Linac photon beam fine-tuning in PRIMO using the gamma-index analysis toolkit

Abstract

Abstract Background In Monte Carlo simulations, the fine-tuning of linac beam parameters to produce a good match between simulated and measured dose profiles is a lengthy, time-consuming and resource-intensive process. The objective of this study is to utilize the results of the gamma-index analysis toolkit embedded inside the windows-based PRIMO software package to yield a truncated linac photon beam fine-tuning process. Methods Using PRIMO version 0.1.5.1307, a Varian Clinac 2100 is simulated at two nominal energy configurations of 6 MV and 10 MV for varying number of histories from 106 to more than 108. The dose is tallied on a homogeneous water phantom with dimensions 16.2 × 16.2 × 31.0 cm3 at a source-to-surface-distance of 100.0 cm. For each nominal energy setting, two initial electron beam energies are configured to reproduce the measured percent depth dose (PDD) distribution. Once the initial beam energy is fixed, several beam configurations are sequentially simulated to determine the parameters yielding good agreement with the measured lateral dose profiles. The simulated dose profiles are compared with the Varian Golden Beam Data Set (GBDS) using the gamma-index analysis method incorporating the dose-difference and distance-to-agreement criteria. The simulations are run on Pentium-type computers while the tuned 10 MV beam configuration is simulated at more than 108 histories using a virtual server in the Amazon.com Elastic Compute Cloud. Results The initial electron beam energy configuration that will likely reproduce the measured PDD is determined by comparing directly the gamma-index analysis results of two different beam configurations. The configuration is indicated to yield good agreement with data if the gamma-index passing rates using the 1%/1 mm criteria generally increase as the number of histories is increased. Additionally at the highest number of histories, the matching configuration gives a much higher passing rate at the 1%/1 mm acceptance criteria over the other competing configuration. With the matching initial electron beam energy known, this input to the subsequent simulations allows the fine-tuning of the lateral beam profiles to proceed at a fixed yet lower number of histories. In a three-stage serial optimization procedure, the first remaining beam parameter is varied and the highest passing rate at the 1%/1 mm criteria is determined. This optimum value is input to the second stage and the procedure is repeated until all the remaining beam parameters are optimized. The final tuned beam configuration is then simulated at much higher number of histories and the good agreement with the measured dose distributions is verified. Conclusions As physical nature is not stingy, it reveals at low statistics what is hidden at high statistics. In the matter of fine-tuning a linac to conform with measurements, this characteristic is exploited directly by the PRIMO software package. PRIMO is an automated, self-contained and full Monte Carlo linac simulator and dose calculator. It embeds the gamma-index analysis toolkit which can be used to determine all the parameters of the initial electron beam configuration at relatively lower number of histories before the full simulation is run at very high statistics. In running the full simulation, the Amazon.com compute cloud proves to be a very cost-effective and reliable platform. These results are significant because of the time required to run full-blown simulations especially for resource-deficient communities where there could just be one computer as their sole workhorse.