Mlc QA Based on Epid Picket Fence Test: An Integrated Method to Partially Eliminate Radiation Scatter from Strip Beams
Abstract
Purpose
Current multi-leaf collimator (MLC) quality assurance (QA) methods based on the picket fence (PF) test are affected by radiation scatter from strip beams, which may reduce analytical accuracy. This study aims to quantify the impact of inter‑strip scatter on algorithm parameters and partially mitigate it, then propose a robust integrated MLC QA method incorporating log files.
Methods
A series of electronic portal imaging device (EPID) PF patterns were designed by varying strip width (7-13 mm) and spacing (10-20 mm) to quantify scatter effects on three parameters: full width at half maximum (FWHM), peak height, and peak area. A scatter library was built for partial compensation. Three QA algorithms (FWHM‑based, peak height‑based, peak area‑based) were developed using log‑file data. Two integrated approaches—a voting‑based method and a linear fitting‑based method—were proposed to enhance robustness. The effects of data processing strategies and strip spacing on positional accuracy were analyzed to determine the optimal algorithm design.
Results
With increasing strip spacing, FWHM and peak area increased while peak height decreased. At fixed spacing, FWHM and peak height declined from the central strip outward; peak area remained nearly constant except for the outermost strips. Developing algorithms independently per strip and leaf pair gave the highest positional accuracy. When strip spacing was ≥16 mm, the overall absolute positional error remained consistent (0.155 ± 0.168 mm). Individual absolute errors were: peak height method 0.294 ± 0.213 mm, peak area method 0.100 ± 0.086 mm, FWHM method 0.071 ± 0.054 mm, voting integration 0.057 ± 0.062 mm, and fitting integration 0.036 ± 0.528 mm.
Conclusion
This study proposed two effective and robust MLC QA integration methods that can reduce partial radiation scattering. When the strip spacing exceeds a certain threshold, the algorithm's positional accuracy is almost unaffected by the strip position or spacing.