Comparison of CT on rails with electronic portal imaging for positioning of prostate cancer patients with implanted fiducial markers.

Abstract

The objective of this investigation was to measure the agreement between in-room computed tomography (CT) on rails and electronic portal image (EPI) radiography. Agreement between the location of the center of gravity (COG) of fiducial markers (FMs) on CT and EPI images was determined in phantom studies and a patient cohort. A secondary analysis between the center of volume (COV) of the prostate on CT and the COG of FMs on CT and EPI was performed. Agreement was defined as the 95% probability of a difference of <or=3.0 mm between images. Systematic and random errors from CT and EPI are reported. From 8 patients, 254 CT and EPI pairs were analyzed. FMs were localized to within 3 mm on CT and EPI images 96.9% of the time in the left-right (LR) plane, 85.8% superior-inferior (SI), and 89% anterior-posterior (AP). The differences between the COV on CT and the COG on EPI were not within 3 mm in any plane: 87.8% (LR), 64.2% (SI), and 70.9% (AP). The systematic error varied from 1.2 to 2.9 mm (SI) and 1.8-2.9 mm (AP) between the COG on EPI and COV on CT. Considerable differences between in-room CT and EPI exist. The phantom measurements showed slice thickness affected the accuracy of localization in the SI plane, and couch sag that occurs at the CT on rails gantry could not be totally corrected for in the AP plane. Other confounding factors are the action of rotating the couch and associated time lag between image acquisitions (prostate motion), EPI image quality, and outlining uncertainties.