8. Analysis of dose variations and sources of error in VMAT treatments with an EPID-based in vivo dosimetry system

Purpose To evaluate the daily dose delivery discrepancies catching sources of errors in VMAT treatments with a 3D EPID-based in vivo dosimetry (IVD) system. Methods and materials The IVD software PerFRACTION (Sun Nuclear Corporation) was used. PerFRACTION is based on the information concerning MLC and collimator positions obtained by EPID images and on the information of log files, such as MUs and the gantry angles. From these informations, a collapsed-cone/superposition algorithm (SDC) is used to calculate the dose distribution on daily CBCT images [1] . It also uses the SDC to produce an entire dose volume to check against that from the TPS. First of all, we calculated the dose discrepancies ( Δ D%) at the isocenter (ID) and the 3D gamma passing rate (%GP) between our reference (AAA) and SDC algorithm, for 81 plans. About IVD, 3D dose distributions were reconstructed for 405 fractions of 53 patients, including prostates (21%), lungs (15%), Head & Neck (17%), PBI (10%), and palliatives (38%). We evaluated Δ D% between reference and daily dose distribution in terms of dose to 95% of the volume ( D 95 % ) , dose to 2% of the volume ( D 2 % ) and average dose of the target. Results We obtained for ID an average Δ D% equal to (1.8 ± 0.8)% and an average %GP equal to (99 ± 2)%. For IVD, the mean Δ D% was equal to (−0.3 ± 2.2)% for D 95 % , (0.9 ± 1.4)% for D 2 % , and (0.5 ± 1.2)% for average dose. 75% of fractions showed Δ D% inferior to 2%. For 74 (18%) of the analyzed dose points, Δ D% was greater than the 3%. For the fractions that exceed 3%, the errors are due to setup (28.4%), patient anatomy (40.5%), setup + anatomy (27%) and bolus positioning (4%). No machine-based errors were observed for these patients. Conclusion 3D EPID-based IVD is a powerful method to cacht and quantify delivery discrepancies during radiotherapy process.