Statistical Variability and Confidence Intervals for Planar Dose QA Pass Rates
D Bailey1*, B Nelms2, L Kumaraswamy1, K Attwood3, M Podgorsak1, (1) Dept of Radiation Medicine, Roswell Park Cancer Institute, Buffalo NY (2) Canis Lupus LLC, Merrimac WI (3) Department of Biostatistics, Roswell Park Cancer Institute, Buffalo NYSU-E-T-16 Sunday 3:00:00 PM - 6:00:00 PM Room: Exhibit Hall
Purpose: The most common metric for comparing measured to calculated dose planes is a pass rate generated using percent difference, distance-to-agreement (DTA), or some combination of the two (e.g. gamma evaluation). The grid of analyzed points often corresponds to a dosimeter array with low areal-density of point detectors. This work examines the statistical uncertainty of planar dose comparison pass rates and proposes methods for establishing confidence intervals for pass rates obtained with low detector-density arrays.
Methods: Absolute dose planes were acquired via EPID for twenty intensity-modulated fields of varying complexity. Matching calculated dose planes were created via treatment planning system. Pass rates for each dose plane pair (centered to CAX) were calculated with various %/DTA composite analysis techniques. Software was designed to selectively sample the high-density EPID matrix to simulate many low-density measured grids, each representing a different alignment with respect to CAX. Simulations were repeated (100 positional iterations per field) using grids of varying detector-densities and both random and orthogonal point-detector orientation. For each simulation, pass rates were calculated with various composite analysis techniques.
Results: Repositioning simulated low-density grids leads to a distribution of possible pass rates for each measured/calculated dose plane pair, independent of whether the detector grid is random or uniform. Distributions can be predicted using a binomial distribution by which a confidence interval (function of sampling density and observed pass rate) is approximated for each pass rate. For example, 95% confidence intervals for IMRT pass rates (2%,2mm) average +/-5.3% and +/-3.8% with 1-detector/cm² and 2-detector/cm² grids, respectively.
Conclusions: Pass rates for low-density array measurements are not absolute and should be reported with both a full description of calculation method and confidence intervals quantifying their uncertainty. Results extend to 3D detector arrays. The concept of fixed 'action levels' for pass rates must be reexamined for low-density array measurements.