Uncertainties in IMRT dosimetry

Abstract

The purpose of this study is to investigate some characteristics of the beam delivery system and their effects on the dose distribution of intensity-modulated radiation therapy (IMRT) and the results of patient-specific IMRT quality assurance (QA). These characteristics include the accelerator source distribution and multileaf collimator (MLC) geometry. Monte Carlo dose calculations based on intensity maps that were built from the actual deliverable IMRT leaf sequences with and without considering the characteristics of the beam delivery system were performed in this study using in-house Monte Carlo software. The effect of the resolution of the intensity maps on the dose distribution was investigated first. The mean dose of the treatment target and the voxel doses in the high dose region of seven IMRT plans generated by different treatment planning systems for Varian 21EX and Siemens Primus linear accelerators were used for comparison and evaluation. The results show that a 0.2 x 0.2 mm2 or smaller pixel size is needed for the intensity maps for accurate IMRT dose calculation. The extrafocal source, MLC leaf thickness, leakage, tongue-and-groove structure, and the effective leaf offset can affect the mean dose by up to 1.5%, 4.5%, 5.6%, 5.3%, and 7.8%, respectively, when these factors are considered separately. They also cause significant uncertainties to the voxel dose with standard deviations up to 2.5%, 0.7%, 2.1%, 1.3%, and 5%, respectively. The overall effect on the mean dose is up to 8% and the standard deviation of the voxel dose uncertainty is up to 6.4% when all the effects are included. The maximum standard deviation is reduced to 4.6% if the voxel size of the dose calculation matrix is increased from 0.04 to 0.3 cm3 to make it comparable with the sensitive volume of the ionization chamber used for IMRT QA measurements. It can be concluded that the characteristics of the beam delivery system are the major contributors to the uncertainty of measurement-based IMRT QA because most of them are not fully considered in the currently available treatment planning systems.