Unencrypted login | home

Program Information

The Performance Test of EPID for In-Vivo Dosimetry

no image available
N Saotome

N Saotome1,2*, S Kida1, T Imae1, K Sasaki1, A Sakumi1, Y Masutani1, A Haga1, K Nakagawa1, (1) The University of Tokyo Hospital, Tokyo, ,(2) Toho Omori Hospital, Tokyo,

SU-E-J-219 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose:
An electronic portal imaging device (EPID) imaging is one of the promising tools to observe the dose in real time with a dynamic treatment. From our previous measurement, it was found likely that irradiation field was displaced more than 0.5 mm on the gantry-rotating axis, and such an error has never detected by MLC log file. An EPID, on the other hand, may overcome this problem. In this study, we show the feasibility for MLC detection for in-vivo dosimetry.

Methods:
An EPID (iViewGT; Elekta) and a gantry angle sensor (DosimetryChack; Math Resolutions) were employed. An in-house program written in c++ code was developed to detect the MLC position and the detected position was compared with MLC log file. In this study, we employed two algorithms for the detection; threshold algorithm and Gradient algorithm. MLC position was determined as most nearest position from intensity of each threshold.
To estimate the uncertainty of the MLC detection using EPID, a reproducibility of EPID panel position in initial setting and position difference depending of the gantry angle at the open were measured.
Static irregular shape fields and rotating rectangular field were used for the performance evaluation test of algorithm. Convex and concave shapes was acquired and processed by two algorithms.

Results:
Reproducibility of EPID panel position results less than 0.5 mm. Position difference depending of the gantry angle at the open was less than 1.0 mm. For convex-shape field, algorithm dependency was enhanced. Uncertainty of gradient method was found to be least in this study. Whereas threshold method was sensitive to scatted radiation, gradient method detecting the inflection point was less sensitive.

Conclusion:
It was determined that uncertainty of this system was totally 1.1 mm when gradient method was employed.

Contact Email: