A comparison of two image fusion techniques in ct-on-rails localization of radiation delivery

A computed tomography (ct) scanner on Rails has been installed in a linear accelerator room at Morristown Memorial Hospital since 2000. The ct-on-Rails has been used for the localization of patient position during radiation delivery for prostate, lung and liver cancer patients. The image management system, the Siemens Syngo system, is the primary software employed in the registration of the planning ct and the treatment ct images. This study compares the two image fusion methods available in the system: Landmark Registration and Visual Alignment. Shifts in 6 ct scans with Rando phantom were deduced from Landmark Registration (automatic algorithm) and from Visual Alignment (manual registration), and compared with the shifts directly measured on the phantom. For Visual Alignment, the isocenter shifts deduced from the fused images generally agreed well with the directly measured shifts on the Rando phantom, with average absolute error of 0.9 mm in anterior-posterior (ap) direction, 1.0 mm in right-left (rl) direction, and 2.0 mm in superior-inferior (si) direction. The image fusion algorithm was confirmed to be accurate. Some scans with Landmark Registration gave erroneous ap shifts when the anterior radio-opaque marker (bb) registration was of in the ap direction. Visual Alignment was more robust than Landmark Registration in these clinical situations.