Poster - Thur Eve - 21: ROC analysis in patient specific quality assurance.

Abstract

Many institutions rely on a patient specific measurement for IMRT/VMAT patient QA. In diagnostic imaging, radiologists use Receiver Operator Curves (ROC) to help quantify the value of a diagnostic imaging test. The purpose of this work is to investigate the value or ROC methodology for patient specific IMRT QA. Beam fluences for 34 prostate IMRT patients were analyzed using gamma analysis. For half of these, measurements were done using the planned beam fluences. For the rest, perturbations to the MLC leaf positions were introduced. Gamma analysis was then used to measure fluence differences. Assuming that the unperturbed fluencies were positive measurements, distributions of true positive and false negatives were calculated. For poorly performing beam delivery systems the choice of γ-DTA criterion has little effect on test sensitivity and specificity. The AUC is increased by about 10% for high performance beam delivery systems. For a 3%/3mm γ-DTA condition, ideal cut off values are reasonably independent of MLC performance. At a tighter γ-DTA condition of 2%/2mm, then the optimal sensitivity and specificity of the test is more dependent on MLC performance. For a pass-fail test such as the γ-DTA map is, it is important to choose an optimal cut off value to maximize the sensitivity and specificity of the test. ROC methodology allows users to follow a prescriptive method to obtain ideal cut-off values for gamma analysis, and to assess improvements in sensitivity and specificity for higher performing beam delivery system.