Room: ePoster Forums
Purpose: Multiple algorithms exist to analyze Winston-Lutz (WL) images. With typical WL tolerances of 1mm for SBRT/SRS linacs, these algorithms must determine the location of a ball bearing (BB) with sub-millimeter accuracy. The ability of five algorithms: four commercial and one in-house to determine small BB displacements was studied.
Methods: A 12.70 mm diameter BB was mounted to a translational stage using a low density tube. The translation stage, capable of 0.01 mm motion increments, was mounted to the table of a Varian Edge linac and leveled. The BB was positioned at isocenter and known translations ranging from 0.05 mm to 1 mm were applied. After each translation, a 2 MU MV image was taken using the 1280x1280 pixel EPID (4x4 cmÂ² jaw, 150 cm source-to-imager distance). The reproducibility of the system was tested two ways: by taking multiple images without changing the setup, and by re-setting the position of the BB and re-imaging. Images were analyzed in four commercial software packages: Sun Nuclear Corporationâ€™s SunCHECK, Mobius Medical Systemâ€™s DOSELAB, Radiological Imaging Technologyâ€™s RITG142, and Varianâ€™s â€œDetect Ballâ€? tool along with an in-house MATLAB-based algorithm.
Results: With the exception of the â€œDetect Ballâ€? algorithm, the standard deviation of the position of the BB was less than 0.04 mm when the BB was held stationary. The standard deviation of the position of the BB across multiple physical setups was less than 0.11 mm for all algorithms. No algorithm was able to detect 0.05 mm translations. Translations of 0.1 mm and more were detected with an accuracy of 0.1 mm or better, depending on the algorithm. Pixel spacing at isocenter was 0.22 mm, indicating that all algorithms tested have sub-pixel resolution and accuracy.
Conclusion: All algorithms tested can detect submillimeter BB positional variations with an accuracy of 0.1 mm or better.