Development of an Automated Method for Analysis of Winston-Lutz Test Results Using Digital Radiography and Photostimulable Storage Phosphor

Abstract

Stereotactic radiosurgery (SRS) and radiotherapy (SRT) are intricate techniques that deliver a highly precise radiation dose to a localized target, usually a tumor. At our hospital, we perform SRS and SRT on brain tumors using a linear accelerator (linac) mounted with an external micro multi-leaf system. The Task Group TG-142 Report by the American Association of Physicists in Medicine recommends the coincidence of the radiation and mechanical isocenter to be within ±1 mm. The Winston-Lutz test is commonly used to verify the linac isocenter position: it has the advantages of being a simple method that uses a film or electronic portal imaging device (EPID). However, the film method requires a higher radiation dose, which makes it more time-consuming than the EPID method, and the results are highly dependent on the skills of the observer. The EPID method has certain advantages over the film method, but it has low resolution and can only be used for a few combinations of gantry and couch angles. This prompted us to develop an in-house-designed radiation receptor system based on digital radiography, using a photostimulable storage phosphor and automated analysis algorithm for Winston-Lutz test images using a template-matching technique based on cross-correlation coefficients. Our proposed method shows a maximum average absolute error of 0.222 mm (less than 2 pixels) for 0.5 mm and 1.0 mm displacement from the isocenter toward the inline and crossline directions. Our proposed method is thus potentially useful for verifying the Linac isocenter position with a small error and good reproducibility, as demonstrated by improved accuracy of evaluation.

Teste de Winston-Lutz: Uma análise quantitativa

“…Nagafuchi, et al 13 , apresenta método de análise automatizado dos resultados do teste de WL, através do uso de placas de fósforo fotoestimuladas e radiografia digital. No seu método proposto, é realizado análise automatizada e um modelo baseado em coeficiente de correlação cruzada, obtendo média de erro absoluto de até 0.222mm.…”