Purpose: Presentation of a new method for IMRT QA comparisons that is more sensitive, more specific, and more clinically meaningful than gamma comparisons.
Methods: Gradient-dose segmented analysis (GDSA) works by segmenting the dose map from calculations on an IMRT QA phantom into regions of high vs. low dose and dose gradient. The mean local dose difference in the high-dose low-gradient regions of the phantom comparison directly predicts the changes in the patient PTV mean dose using only information from the calculation on the detector array. GDSA is based on analysis of >180,000 comparisons in 40 IMRT and VMAT cases using various induced errors to determine appropriate dose/gradient thresholds and dose differences for phantom dose segmentation for GDSA. Here, GDSA is compared to five gamma criteria, using 3 different detector geometries (ArcCHECK, MapCHECK, Delta4), to evaluate the ability to predict ±3% changes in PTV mean dose (as a binary metric) using ROC analysis. An independent set of 25 clinical IMRT/VMAT cases were the validation dataset for evaluation of the GDSA dose/gradient thresholds and overall behavior.
Results: ROC analyses comparing the sensitivity/specificity of GDSA to five different gamma criteria showed GDSA exhibited improved AUC compared to all five gamma criteria for all devices. AUC results from the validation dataset illustrate that sensitivity/specificity of GDSA is much improved compared to gamma with AUC values for GDSA vs gamma at 3%/3mm threshold 10%, global normalization were 0.893 vs. 0.79, 0.983 vs. 0.877, and 0.977 vs. 0.934 for ArcCHECK, MapCHECK, and Delta4, respectively.
Conclusion: GDSA analysis of IMRT QA is capable of directly predicting changes in patient PTV mean dose using only information from the QA phantom measurement geometry, while also providing more clinically meaningful results and better identification of errors without any increases in clinical workload or effort.