An investigation of gantry angle data accuracy for cine-mode EPID images acquired during arc IMRT

EPID images acquired in cine mode during arc therapy have inaccurate gantry angles recorded in their image headers. In this work, methods were developed to assess the accuracy of the gantry potentiometer for linear accelerators. As well, assessments of the accuracy of other, more accessible, sources of gantry angle information ( i.e., treatment log files, analysis of EPID image headers) were investigated. The methods used in this study are generally applicable to any linear accelerator unit, and have been demonstrated here with Clinac/Trilogy systems. Gantry angle data were simultaneously acquired using three methods: i) a direct gantry potentiometer measurement, ii) an incremental rotary encoder, and iii) a custom-made radiographic gantry-angle phantom which produced unique wire intersections as a function of gantry angle. All methods were compared to gantry angle data from the EPID image header and the linac MLC DynaLog file. The encoder and gantry-angle phantom were used to validate the accuracy of the linac's potentiometer. The EPID image header gantry angles and the DynaLog file gantry angles were compared to the potentiometer. The encoder and gantry-angle phantom mean angle differences with the potentiometer were 0.13 degrees +/- 0.14 degrees and 0.10 degrees +/- 0.30 degrees, respectively. The EPID image header angles analyzed in this study were within +/- 1 degrees of the potentiometer angles only 35% of the time. In some cases, EPID image header gantry angles disagreed by as much as 3 degrees with the potentiometer. A time delay in frame acquisition was determined using the continuous acquisition mode of the EPID. After correcting for this time delay, 75% of the header angles, on average, were within +/- 1 degrees of the true gantry angle, compared to an average of only 35% without the correction. Applying a boxcar smoothing filter to the corrected gantry angles further improved the accuracy of the header-derived gantry angles to within +/- 1 degrees for almost all images ( 99.4%). An angle accuracy of 0.11 degrees +/- 0.04 degrees was determined using a point-by-point comparison of the gantry angle data in the MLC DynaLog file and the potentiometer data. These simple correction methods can be easily applied to individual treatment EPID images in order to more accurately define the gantry angle.