SU‐E‐T‐440: Gain Calibration Stability Study of a MV Flat‐Panel‐Detector (FPD) On Siemens ARTISTE Linac

Abstract

Purpose: To study gain calibration variation over time for an MV flat‐panel‐detector (FPD). Methods: Gain calibration images (1024×1024 pixels) of a FPD (PerkinElmer AN‐9 on a Siemens ARTISTE) acquired in 40 consecutive months were studied. Using Python programming language, the images were processed to analyze the central 900×900 pixels with a threshold applied to exclude dead pixels. Month1 was set as the base image, then the difference images between it and the following months were calculated. The pixel intensity mean and standard deviation of the difference images were used to study the gain variation over time. Two other months’ images were also randomly selected as the base image and the above analyses were repeated. Finally, to investigate the equivalence of monthly calibration and quarterly calibration, the results from comparing neighboring months’ images, such as (month2‐month1), (month3 ‐month2) et al, were compared with those from comparing images taken in 3‐month intervals, such as (month4‐month1), (month7‐month4) et al, using Welch's t‐test. Results: For most months’ gain calibration images, the differences (mean and standard deviation) from base images were constant. In three months, the differences were relatively larger compared to other months but seemed instantaneous without any worsening trend. In those 3 months no clinical portal image quality degradation was observed. The difference between monthly and quarterly calibration tested by Welch's t‐test were insignificant for both means (p=0.9) and standard deviations (p=0.8). Conclusion: Long term stability of the FPD gain calibration images was observed. Therefore, the vendor‐recommended calibration frequency of 2–4 weeks is unnecessary. Less frequent, such as quarterly, calibration is sufficient.