Open Access Thesis
Date of Award
Michael G. Snyder
Monte Carlo (MC) based dose calculation methods trade-off accuracy at the expense of computational time, which is, correlated to the user input values of statistical uncertainty and pixel spacing (1). It was first hinted by low et. al. that noise generated within either the calculated or measured plan distributions can affect the result of the plan verification by method of ‘Gamma Index Analysis’(GI) (2). The purpose of this research experiment is to investigate a possible correlation between added noise from increasing MC statistical uncertainty and increasing the odds of a plan passing the GI verification criteria. For this research experiment, we calculated 10 head and neck radiation therapy treatment plans using the MC dose calculation method within Monaco TPS. We varied the statistical uncertainty values from 5%, 3%, 1% and 0.25% and varied the voxel size values from 3mm, 2mm and 1mm. The treatment plans were then administered on an Elekta Versa linear accelerator and measured using Mapcheck dose measurement device. Each plan was evaluated for clinical pass/fail using the GI Analysis with criteria 3%/3mm and 2%/2mm. For 1 mm voxel size, 3%/3mm GI, there was an increase in average gamma pass rates from 98.91% calculated at 0.5% statistical uncertainty to 99.61% calculated at 5% statistical uncertainty. For 1 mm voxel size, 2%/2mm GI, there was an increase in average gamma pass rates from 97.02% calculated at 0.5% statistical uncertainty to 98.80% calculated at 5% statistical uncertainty. At 2 mm and 3 mm voxel sizes, there was not a clear demonstrable increase in average gamma pass rates. The experimental results conclude that the user must be careful when selecting a statistical uncertainty prior to performing a MC dose calculation. The input of a high statistical uncertainty does not lead to more points failing the GI, but paradoxically, can increase the chances that the evaluated radiation therapy plan will pass the acceptance evaluation.
Van Delinder, Kurt William, "Higher Statistical Uncertainty With Small Pixel Sizes Gives Higher Pass Rates." (2016). Wayne State University Theses. 538.