Verifying Plan QA for Online Adaptive Treatment in MR Linac.

Purpose/Objective(s) In addition to real-time tracking, a major advantage of the MR Linac is the ability to perform online plan adaptation in accordance with the patient anatomy of the day. This allows delivery of adequate tumor doses while ensuring surrounding organ at risk sparing. Treatment sites include pancreas prostate, lung and liver, among others. Typically, daily plan adaption leads to a change in monitor units (MU) from the original plan. This change can be over 100% at times. The vendor provides an online QA tool which is a secondary MU check. The purpose of this work is to evaluate the quality of plan QA in online adaptive IMRT treatments compared to physical, post-treatment measurements. Materials/Methods We evaluated 32 online adapted plans (26 patients) which passed vendor-supplied secondary MU plan checks with a gamma criterion of 2mm/2%, and were treated clinically. We used both a diode array and an ion chamber for absolute dose measurement at isocenter. MU changes were evaluated in four groups: 0-25%, 26-50%, 51-75% and above 75% difference between adapted and original plans. Overall, MU changes ranged from 0%-102% between original and adapted fractions. Results The pass rates of the adapted plans were statistically significantly different from those of the original plans by up to 5% (2-tailed t-test). In 6 cases (19%) the relative dose passed and the absolute dose did not. In 4 cases (12%), the QA did not pass even using a 3mm/3% criteria. In 22 of 27 plans (82%) the ion chamber measurements were within 3% of the predicted dose. The other 5 plans had a larger than 3% difference between planned and measured dose at isocenter. In 5 plans the isocenter was too lateral to measure with an ion chamber. Conclusion Our results show the vendor-supplied QA tool is in good correspondence with post treatment QA measurement. We did not find any correlation between the magnitude of MU changes between the original and adapted plans and the QA pass rates, which is reassuring given the large differences in some of the plans. Treatment site was not correlated to either the magnitude of MU changes or the QA passing rates. The vendor-supplied QA tool cannot be used when treating a patient with the original plan, and thus could not be assessed on these plans. As the optimization algorithm is identical in the treatment planning system and in the adaptive planning process in the treatment delivery computer, the underlying cause for the differences observed in passing rates is unclear. Overall, our results prove the reliability of the vendor-supplied MU secondary check as adequate real-time QA for confidently treating adapted plans.