Commissioning of a model based 3d second check and PSQA system.
Bismack B, Gopal A, Wen N, Kim J, Laugeman E, Dumas M, Miller B, and Chetty I. Commissioning of a model based 3d second check and PSQA system. Med Phys 2017; 44(6):2891-2892.
Purpose: Commissioning of new software for independent verification of treatment plan calculation (Dose Check) using a systematic approach to that draws on recommendations from TG119 and TG53 to isolate situations for potential modeling issues. Methods: Beam models for the calculation engine (Sun Nuclear Dose Calculator, SDC, based on collapsed cone convolution) were tuned by the vendor using a subset of primary TPS beam data and tweaking MLC model properties to match SDC to previously delivered dose distributions from our primary TPS (AAA). The beam model was first evaluated for jaw and MLC defined open fields in regions of the radiation field as delineated by AAPM TG-53. Accuracy for dynamic deliveries was evaluated by comparing simple sweeping gap calculations between AAA and SDC, and by doing a 3 way comparison of ion chamber and film measurements of TG-119 prescribed IMRT (and similar VMAT) plans between AAA, SDC, and measurement. Results: For open-field measurements, SDC generally agreed within 1.5% (2.5% for < 3x3) in the in-field region. For the inner penumbra, SDC agreed within 0.2 mm DTA, outer penumbra agreed within 0.4 mm. In the out of field region all deviations were within 0.3 cGy. Agreement between AAA and SDC for sweeping gap patterns was within 3% point dose and had a 97% pass rate for 3% 3 mm. Ion chamber plan measurements showed a mean within 1.72% of measurement with a std. dev. of 3.1%; similar stats were -1.23% and 2.57% between SDC and AAA respectively. Net planar dose statistics for 3% 3 mm passing was 93.47% mean, 5.05% std dev. with measurement and 99.88%, 0.19% with AAA respectively. Conclusion: SDC is a robust model capable of agreement within TG53 and TG119 suggested tolerances compared to both TPS AAA calculation and measurement. Further optimization based on results of evaluation may further improve model performance.