Refinement of MLC modeling improves commercial QA dosimetry system for SRS and SBRT patient-specific QA

PurposeMobius 3D (M3D) provides a volumetric dose verification of the treatment planning system's calculated dose using an independent beam model and a collapsed cone convolution superposition algorithm. However, there is a lack of investigation into M3D's accuracy and effectiveness for stereotactic radiosurgery (SRS) and stereotactic body radiotherapy (SBRT) quality assurance (QA). Here, we collaborated with the vendor to develop a revised M3D beam model for SRS/SBRT cases treated with a 6X flattening filter-free (FFF) beam and high-definition multiple leaf collimator (HDMLC) on an Edge linear accelerator. MethodsEighty SRS/SBRT cases, planned with AAA dose algorithm and validated with Gafchromic film, were compared to M3D dose calculations using 3D gamma analysis with 2%/2 mm gamma criteria and a 10% threshold. A revised beam model was developed by refining the HD-MLC model in M3D to improve small field dose calculation accuracy and beam profile agreement. All cases were reanalyzed using the revised beam model. The impact of heterogeneity corrections for lung cases was investigated by applying lung density overrides to five cases. ResultsFor the standard and revised beam models, respectively, the mean gamma passing rates were 94.6% [standard deviation (SD): 6.1%] and 98.0% [SD: 1.7%] (for the overall patient), 88.2% [SD: 17.3%] and 93.8% [SD: 6.8%] (for the brain PTV), 71.4% [SD: 18.4%] and 81.5% [SD: 14.3%] (for the lung PTV), 83.3% [SD: 16.7%] and 67.9% [SD: 23.0%] (for the spine PTV), and 78.6% [SD: 14.0%] and 86.8% [SD: 12.5%] (for the PTV of all other sites). The lung PTV mean gamma passing rates improved from 74.1% [SD: 7.5%] to 89.3% [SD: 7.2%] with the lung density overridden. The revised beam model achieved an output factor within 3% of plastic scintillator measurements for 2 x 2 cm(2) MLC field size, but larger discrepancies are still seen for smaller field sizes which necessitate further improvement of the beam model. ConclusionSpecial attention needs to be paid to small field dosimetry, MLC modeling, and inhomogeneity corrections in the beam model for SRS/SBRT QA. The improvements noted in this study, and further collaborations between clinical physicists and the vendor to refine the M3D beam model could enable M3D to become a premier SRS/SBRT QA tool.