Room: AAPM ePoster Library
Purpose: To use the Receiver Operator Characteristic (ROC) analysis to determine the threshold that can be used in gamma function during patient specific quality assurance for VMAT plans
Methods: 13 VMAT plans for head & neck cases were delivered as planned. Two errors were introduced, isocenter shift in order of 1.0, 2.0, 3.0, 4.0 and 5.0mm. The second error was a modification of beam MU either by increase or decrease by 1%, 2%, 3%, 4% and 5%. Dose calculations were performed using Eclipse V13 and plans were delivered by Varian 2100C/D linear accelerator with 120 leaf MLC. Dose measurements were performed using PTW Octavius II with Detector 729 array. Five gamma function criteria were used 1%/1mm, 2%/2mm, 3%/3mm, 4%/4mm and 5%/5mm
ROC curves were produced for each Gamma function criteria. For each curve the area under the curved AUC was calculated. Ideal thresholds were selected by finding point where sensitivity and specificity are close to 100% on the ROC curve
Results: For 3.0mm shift the system detectability is perfect, and it will continue to be perfect for 2.0mm if the gamma function criteria reduced to 2%,2mm and 1%,1mm. while it shows a poor behavior to detect 1.0mm.
For the dose difference the system detectability is perfect for 3%, while it drops down by 15% for 2% and by 35% for 1%.
The ideal threshold to detect 2mm shift is 96% and 100% for 2%/2mm and 3%/3mm respectively and to detect 3mm is 85.5 and 98.5 respectively.
For the dose difference, the ideal threshold to detect 3% error is 94.5 and 100 for 2%/2mm and 3%/3mm
Conclusion: ROC method is a useful tool that help to understand our QA equipment and their ability, and to decide what Gamma criteria to be used depending on type of error and its variation
ROC Analysis, Quality Assurance