Session Item

Dose given by the treatment planning system (TPS) reflects an ideal situation as we classically assume the patient as a rigid object and delivery uncertainties from the treatment machine are not present. For the uncertainties related to the patient, different strategies (image guidance, motion/setup uncertainty simulation) have found the way to the treatment process. From there we asked if and how typical machine delivery errors like gantry and  MLC  positions, isocenter and gantry sag  uncertainties  will have an effect on the dose distribution or if they can be neglected within the radiotherapy process as we classically do.

Main delivery errors like gantry sag (.5, 1, 1.5, 2mm), isocenter uncertainty (.5, 1, 1.5, 2mm), MLC calibration errors (+/-.5, 1, 1.5, 2mm) and gantry angle errors (+/-.5, 1, 1.5, 2°) are tested for 3 patient cases (Prostate, BC, H+N). Original treatment plans were used to generate modified DICOM RT plan files using an editor tool designed in IDL (Harris Geospatial Solutions). Manipulated plans have been reimported into Raystation 8.0 (Raysearch) and the resulting dose was calculated. For evaluation a volume dose factor was calculated based on the DVHs. This was done by multiplying each DVH dose point and the corresponding volume of the organ or target for all modified plans. To get an idea of this factor, redosed original plans (+/- 2Gy) were calculated. From that, a linear fit for each volume (dose to dose volume) was established. This made it possible to calculate a rescale (offset) dose the plan would have to be rescaled with to show the same effect as the error related dose volume change.

Gantry sag: Prostate and BC had higher doses for higher sag for all volumes while H+N had lower doses at targets only. Isocenter: H+N and BC had lower doses for all volumes while prostate only had some higher doses. MLC positioning: clear correlation: larger field is more dose (Fig 1). Gantry angle: H+N and BC mostly had lower and prostate had some higher doses. In general the errors resulted in higher dose offsets when volumes were small as for the H+N case. Therefor delivery errors might be of higher interest for such cases.  

The limited number of cases in this study showed complex dependencies. Commonly delivery errors do vary the dose more than expected but the dose change is difficult to predict as it depends on the case and beam arrangement. Therefore new dose calculation models should include a kind of dose uncertainty display based on general or better individual machine QA results. This might even help with the decision on the QA procedures used for plan acceptance.