Quantitative assessment of digitized portal images: effect of sampling frequency on observer performance.

Abstract

To investigate spatial resolution requirements for digitized portal images in radiation therapy, observer performance tests were performed. One hundred twenty portal images were digitized with sampling frequencies of 0.700, 0.350, and 0.175 mm for observation. Receiver operating characteristic analysis was used to determine the acceptable sampling frequency for clinical portal images. The detectability of setup errors was significantly better on the original images than on the digitized images with sampling frequencies of 0.700 mm (P = .005) and 0.350 (P = .046). Some clinical disadvantages might accrue with the use of a sampling frequency of 0.350 mm or larger.