Calculating photon beam characteristics with Monte Carlo techniques.

Abstract

This study describes the results from a simulation of a 10 MV photon beam from a medical linac using the BEAM code. To check the quality of the generated photon beam, the characteristics of this beam (depth dose curve, cross profiles, and output factors) have been calculated and compared to measured data. By splitting up the radiation head in two parts, the target section and the collimator section calculation times were long, but acceptable when aiming at phase space files containing some 5 million particles. Given the number of particles evaluated, the accuracy of all data was around 2%. Analysis of the phase space files for different field size supports results from previous studies about contaminant particles and sources for scattered radiation for photon beams from medical linacs. The total scatter output factor Scp as well as the collimator scatter output factor Sc have been calculated within 2% of measurements. Also, the ratio between dose at a reference point for the full scatter situation and the no-scatter situation has been calculated correctly. All depth dose curves and cross profiles have also been calculated correctly, although with only moderate statistics. Improvements are possible by increasing the number of particles in the simulations (up to 50 million for the largest field size) at least 4-8 times, although calculation times will increase with the same factor. Nevertheless, the method proved itself as reliable. Still, the accuracy should be improved to 1% or better. This is necessary as we plan to use Monte Carlo simulations to benchmark three-dimensional radiotherapy planning systems. By increasing the number of particles in the phase space files and subsequently increasing the number of particles in each simulation, this 1% accuracy will be achieved. The easy way to increase the number of particles in a simulation by increasing the number of times phase space files, which were already recycled ten times, are reused from ten times (this study) to 40 times or more will not work, as it introduces artifacts, especially in the cross profiles.