The consequences of fixed-pattern noise and image movement on electronic portal images.

Abstract

Fixed-pattern noise in electronic portal images (EPIs) is normally eliminated by dividing raw image data by an open-field calibration image (OFCI). However, for successful elimination, there must be exact registration between the two image sets. Any movement within the imaging system, such as that which occurs with gantry angle, will result in misregistration and a subsequent increase in noise. This paper describes, both qualitatively, by way of example, and quantitatively, by way of variance analysis, the consequences of misregistration and its effects on image quality. Our results show that image quality is found to degrade significantly with change in gantry angle, when a single OFCI is used, with a loss of low-contrast, fine detail. Variance is observed to increase over 2.5-fold. A simple solution of using multiple OFCIs is described, along with a technique for optimizing the number of images required, and the gantry angles at which they are acquired. When five OFCIs are used, the variance changes with gantry angle are limited to less than 20%. These changes are observed in both long and very short exposures (5 MU or 1 s).