Clinical implementation of a secondary dose calculation system for patient-specific quality assurance of complex VMAT and SBRT treatments

Purpose: Advanced radiotherapy techniques require robust patient-specific quality assurance (PSQA). This study validated a secondary calculation software for plan dose verification, evaluating accuracy across different treatment sites, beam qualities, and plan complexities. Methods: Data from two beam-matched VersaHD linacs were used to commission RadCalc-V7.3.2.0. 33 end-to-end tests in homogeneous and anthropomorphic phantoms compared RadCalc Monte Carlo (RC-MC) and Collapsed Cone Convolution Superposition (RC-CCCS) algorithms with Pinnacle3 TPS, using gamma analysis (1.5 %/2mm) and ionization chamber measurements. 140 clinical VMAT plans of varying complexities, including 35 head and neck (H&N) and 65 SBRT treatments, were evaluated using gamma analysis (3 %/2mm) and relevant DVH metrics for PTV (D98%, D2%). RadCalc calculations were compared with pre-treatment Octavius 4D measurements. Results: Phantom studies showed excellent RadCalc-TPS agreement for homogeneous plans and lung SBRT with flattened beams (mean passing rates > 98 %, mean measured dose differences < 1 %). Larger discrepancies were observed in the anthropomorphic thorax phantom for FFF SBRT. For clinical plans, mean passing rates exceeded 98.5 %. Site-specific differences emerged: RC-MC performed better for H&N, RC-CCCS for other sites. RadCalc calculated slightly less homogeneous dose distributions than Pinnacle3, but averaging RC-MC and RC-CCCS results in reduced DVH discrepancies (mean ΔD98% −1.1 ± 1.1 %, mean ΔD2% +1.1 ± 1.5 %). Octavius measurements may underestimate calculation discrepancies due to tissue inhomogeneities. Conclusions: RadCalc produced very consistent results with Pinnacle3 and can be integrated into our PSQA program for efficient 3D dose verification, reducing measurement workload while maintaining high standards of dosimetric accuracy. Using both RadCalc algorithms effectively reduced calculation uncertainties.