Fifty-seventh annual meeting of the American association of physicists in medicine
TH-CD-204-04: Automated Beam Modeling for a Hybrid MRI Radiotherapy System
The first clinical MR-linac will soon become operational. We developed a one-day commissioning framework, including a novel 3D data collection method using film (to overcome the limited watertank scan range), automated beam modeling (Monaco TPS) and verification procedure.
For three field sizes (3×3, 10×10 and 20×20 cm2) beam data is obtained by exposing multiple EBT3 films sandwiched in a RW3 phantom at different depths, as well as one axial film through the isocenter. Subsequently, an output measurement is done for each field size using an ionization chamber.All films were digitized and converted to (relative) dose. For each field size the films were registered to the film in reference position (depth = 10 cm) and stacked to form a 3D dose distribution. PDDs from the axial film are used for resampling in the depth direction. The resulting high resolution dose distributions are used as input for the automated beam modeling. The automated beam modeling tool minimizes the differences between the 3D measured and calculated dose for a certain parameter set. Some parameters can be determined by directly; others need to be fitted using an optimization procedure that changes one or more parameters every iteration.Verification of the beam model is done by gamma comparison (2%/2mm) of the orthogonal profiles extracted from the measurement and calculation.
The 3D dose reconstruction using film has been validated by complementary watertank measurements. Acquisition and processing of the film data takes four hours. Beam model parameter fitting typically takes about 15 hours and verification 2 hours. Finally, a 7-beam treatment plan was created and delivered on the research MR-linac using the beam model. Local differences were below 4%.
A full beam commissioning framework in a single day is feasible and might lead to better beam models. The tool can also be used for conventional linacs.