Abstract
This study aims to determine the image registration interoperator variability using the image fusion interface on the TomoTherapy system. The study focused on the registration results among five Ottawa Hospital Cancer Centre personnel comparing 15 retrospective prostate, bladder, and head and neck patients from day 2 megavoltage computed tomography images with planning computed tomography images. Personnel were instructed to manually adjust the fused images after automatic bony registration matching disease-specific criteria. The Pearson correlation matrix was applied to the summed vector lengths of the personnel's overall translations in all three disease sites. The results were compared with a chosen benchmark personnel and were analyzed for significance in correlation. Finally, the overall translations compared with the automatic bony registration were analyzed for any trends in the data. In each disease site, the average correlation from all personnel compared to a benchmark registering prostate, bladder, and head and neck patients were 0.94, 0.94, and 0.64, respectively. Then, analysis of the interuser overall translations illustrated that each disease site had its own trends with regard to interfraction image fusion corrections. Finally, the mean translational variations were assessed over all personnel for each disease site in X (lateral), Y (longitudinal), and Z (vertical) directional planes. The results demonstrate that consistent, accurate image registration is dependent on factors involving overall user experience with TomoTherapy software and user knowledge of human sectional anatomy, among others.