Evaluation of accuracy and precision in polymer gel dosimetry

PURPOSE: To assess the overall reproducibility and accuracy of an X-ray computed tomography (CT) polymer gel dosimetry (PGD) system and investigate what effects the use of generic, interbatch, and intrabatch gel calibration have on dosimetric and spatial accuracy. METHODS: A N-isopropylacrylamide (NIPAM)-based gel formulation optimized for X-ray CT gel dosimetry was used, and the results over four different batches of gels were analyzed. All gels were irradiated with three 6 MV beams in a calibration pattern at both the bottom and top of the dosimeter. Postirradiation CT images of the gels were processed using background subtraction, image averaging, adaptive mean filtering, and remnant artifact removal. The gel dose distributions were calibrated using a Monte Carlo (Vancouver Island Monte Carlo system) calculated dose distribution of the calibration pattern. Using the calibration results from all gels, an average or "generic" calibration curve was calculated and this generic calibration curve was used to calibrate each of the gels within the sample. For each of the gels, the irradiation pattern at the bottom of the dosimeter was also calibrated using the irradiation pattern at the top of the dosimeter to evaluate intragel calibration. RESULTS: Comparison of gel measurements with Monte Carlo dose calculations found excellent dosimetric accuracy when using an average (or generic) calibration with a mean dose discrepancy of 1.8% in the low-dose gradient region which compared to a "best-case scenario" self-calibration method with a mean dose discrepancy of 1.6%. The intragel calibration method investigated produced large dose discrepancies due to differences in dose response at the top and bottom of the dosimeter, but the use of a dose-dependent correction reduced these dose errors. Spatial accuracy was found to be excellent for the average calibration method with a mean distance-to-agreement (DTA) of 0.63 mm and 99.6% of points with a DTA < 2 mm in high-dose gradient regions. This compares favorably to the self-calibration method which produced a mean DTA of 0.61 mm and 99.8% of points with a DTA < 2 mm. Gamma analysis using a 3%/3 mm criterion also found good agreement between the gel measurement and Monte Carlo dose calculation when using either the average calibration or self-calibration methods (96.8% and 98.2%, respectively). CONCLUSIONS: An X-ray CT PGD system was evaluated and found to have excellent dosimeteric and spatial accuracy when compared to Monte Carlo dose calculations and the use of generic and interbatch calibration methods were found to be effective. The establishment of the accuracy and reproducibility of this system provides important information for clinical implementation.