SU-E-T-94: Multileaf Collimator Performance and Validation of Quality Control Tolerances

The automated quality assurance system (AQUA) is a centralized quality control (QC) software designed to automate QC tests. Statistical analysis of AQUA results was performed to assess the geometric accuracy and long-term reproducibility of a commercially available multileaf collimator (MLC) and examine the applicability of the American Association of Physicists in Medicine (AAPM) tolerances for MLC QC.

The MLC was first calibrated with AQUA by minimizing leaf-positioning errors on megavoltage images for 5 different leaf-bank positions (-60 to 100 mm from radiation isocenter). Leaf-positioning accuracy and reproducibility was assessed by repeating the AQUA test 5 times/week. The range of leaf-positioning error over leaf-bank positions and time was reported. Measured leaf-positioning errors were then separated into systematic and random error components. The systematic error corresponds to the variation (standard deviation) in mean positioning errors between leaves over leaf-bank positions and time. The random error quantifies the leaf position variations around its mean and is calculated as the root-mean-square of the individual leaf position standard deviations.

To date, 2 different MLCs have been calibrated using AQUA and 9-18 datasets have been acquired to assess performance. For the unit with the longest follow up, the range of leaf-positioning errors was -0.62 to 0.85 mm and 98% of the measured leaf positions (n=7200) were within ±0.5 mm of the nominal position. The systematic error was the main error component (±0.15 to ±0.2 mm) and was attributed to the residual errors after calibration. The random error was ±0.07 mm for both units and demonstrated good leaf-positioning reproducibility and limited uncertainty of the AQUA measurements.

Preliminary results show that after MLC calibration with AQUA, leaf-positioning errors on two different units are well within the AAPM-recommended ±1 mm tolerances. Additional MLC performance improvement is possible if residual errors after calibration can be reduced further as the MLC demonstrated high reproducibility. Funded in part by Elekta Inc.