IROC phantoms accurately detect MLC delivery errors

We evaluated the impact of random and whole-bank multileaf collimator (MLC) delivery errors on dosimetric delivery accuracy in the Imaging and Radiation Oncology Core (IROC) phantom audits, as well as differences in delivery accuracy between the IROC phantom prescription and typical clinical fraction sizes.

Plans were created for the IROC IMRT head and neck (H&N) and SBRT spine phantoms. MLC leaf errors were introduced into the plans: random shifts between -2 and 2 mm, and whole bank shifts of 0.5, 1, and 2 mm. Plans were recalculated and delivered on a Varian Truebeam, and the log files were analyzed using Mobius Fx software. A second study examined the impact of fraction size on MLC position accuracy and corresponding dose delivery accuracy. The standard IROC phantom prescriptions (∼6 Gy) were scaled to the extremes of 2 Gy for H&N and 27 Gy for spine. All plans (original and scaled) were delivered on a Varian Truebeam and 21EX machine.

Random MLC positioning errors produced small average dose deviations in the PTV of up to -2.8% for H&N and 0.7% for spine. Whole-bank MLC shifts resulted in larger average PTV dose deviations up to 8% for H&N and 7.1% for spine. The Varian 21EX irradiations had greater MLC root mean square (RMS) error than Truebeam plans. Plans with smaller prescriptions (and faster leaf motion) had greater MLC RMS errors, but plan accuracy was not affected dosimetrically - all results remained within 1% regardless of fraction size.

Both random and whole bank MLC shifts caused dose deviations in the IROC phantoms that were comparable to clinical results previously found in the literature. Deviations measured with ion chambers were well matched with delivery log file analysis. Smaller dose-per-fraction prescriptions caused larger MLC RMS errors that were detected with log files, but were clinically insignificant compared to the dosimetric accuracy of the plan.