UCLA

Purpose

To quantify and elucidate factors affecting error sensitivity in current IMRT QA comparisons performed with the gamma comparison, to investigate causes for gamma comparison insensitivity, and to utilize these results to develop and validate a new method for analyzing IMRT QA dose distributions in the clinic.

Methods

Over 20,000 gamma comparisons were performed for three detector geometries – ArcCHECK, MapCHECK, and Delta 4 – for a variety of IMRT and VMAT cases in the presence of induced errors. Differences in error sensitivity for each device geometry and delivery technique were studied with the use of 1mm vs. 1mm calculation-only comparisons and in-house MATLAB gamma comparison software developed specifically for this project. Additionally, the effects of spatial sampling for each device were evaluated using gamma comparisons performed at the true spatial sampling of each detector compared to those at 1mm. Patterns of gamma failures were also investigated in the presence of induced errors of increasing magnitude.

Results from these gamma comparisons were considered in developing a new comparison technique for IMRT QA analysis. A new analysis method that segments the IMRT QA comparisons by different dose and gradient thresholds was developed with the use of known induced errors in a calculation-only scenario for the ArcCHECK, MapCHECK, and Delta 4 measurement geometries. Results from the new method were validated for a separate cohort of patient plans, as well as with the use of real plan measurements with and without intentional errors on the MapCHECK device.

Results

Differences in gamma comparison error sensitivity were observed for the ArcCHECK, MapCHECK, and Delta 4 geometries when removing the effects of different spatial sampling for each device. While sensitivity was error type-specific for most studied gamma criteria, a gamma criterion with a 10% low dose threshold and local dose difference normalization appeared to offer similar sensitivity across the three devices. For more commonly used gamma criteria, the Delta 4 appeared more sensitive for the majority of induced error types. Additionally, error sensitivity was lower for VMAT cases compared to IMRT cases across all detector geometries. Reducing the spatial sampling of each device from 1mm to the true spatial sampling of the device did not noticeably affect gamma comparison error sensitivity. In evaluating patterns of gamma failures and gamma value maps in the presence of induced errors of increasing magnitude, it was observed that high dose gradients likely limit the sensitivity of the gamma comparison, regardless of dose difference normalization or detector geometry. Additionally, for some cases the number of diodes in real measurements not falling along these gradients may be alarmingly low, which may help explain why the gamma comparison can fail to flag large errors for certain cases.

A new method, gradient-dose segmented analysis (GDSA) was developed to allow more clinically meaningful and sensitive IMRT QA comparisons. This method segments the comparison points into regions of high-gradient, high-dose low-gradient, and low-dose low-gradient points. The mean local dose difference in high-dose low-gradient regions of the comparison was found to predict true changes in PTV mean in the patient DVH. The development of GDSA made use of over 180,000 comparisons to select appropriate dose and gradient thresholds for IMRT and VMAT cases on the MapCHECK, Delta 4, and ArcCHECK devices. Predictions for change in PTV mean dose performed best for the MapCHECK and Delta 4 geometries, with a nearly 1:1 correlation between predicted and true change in PTV mean dose. Additionally, as a binary pass/fail metric, GDSA exhibited higher sensitivity and specificity than five studied gamma criteria. GDSA results were validated with a separate cohort of patients as well as real MapCHECK measurements. GDSA is feasible for clinical implementation as it would not require an increase in time spent analyzing the results.

Conclusions

A variety of measurement scenarios were considered in controlled calculation-only comparisons that suggest device-specific and delivery technique-specific gamma criteria may be appropriate in order to achieve similar sensitivity in IMRT QA comparisons across the field. Additionally, the complexity of gradient maps in current IMRT QA appears to be a driving factor in error sensitivity for the gamma comparison. Finally, the gradient-dose segmented analysis (GDSA) method has been developed and validated for the purpose of IMRT QA analysis for three different detector devices. GDSA was shown to predict changes in PTV mean in the patient DVH using only information from the calculations and measurements in the phantom geometry. As a binary pass/fail metric, GDSA was also shown to be more sensitive and specific than the gamma comparison. Results from this new analysis technique can help predict the clinical relevance of dose differences in IMRT QA measurements, thus offering more meaningful IMRT QA results.