Initial experience with megavoltage (MV) CT guidance for daily prostate alignments.

PURPOSE: The on-board megavoltage (MV) computed tomography (CT) capabilities of a TomoTherapy Hi*ART unit were used to obtain daily MVCT images of prostate cancer patients. For patient alignment the daily MVCT image needs to be registered with the planning CT image to calculate couch shifts. Three manual techniques of registering the MVCT images with the planning kilovoltage (kV) CT images were evaluated. The techniques are based on visual alignment of (1) fiducial prostate markers, (2) CT anatomy, and (3) kVCT contours. METHODS AND MATERIALS: One hundred and twelve alignments from 3 patients were available for analysis. The radiation therapists visually registered the MVCT images with the planning kVCT images based on fiducial markers for actual patient alignment. Retrospectively, the therapists registered each image set using anatomy and contour-based techniques. In addition to the therapists, a physician retrospectively registered each image set based on each of the three techniques. For each MVCT to kVCT image pair a reference alignment was computed from the center-of-mass (COM) of the three fiducial markers. All registration results were compared with these reference alignments. The physician's image registrations were compared with the radiation therapists' registrations to assess the user variability of the different techniques.
RESULTS: The marker-based registration results agree best with the reference alignments, while the contour-based registrations show the least degree of agreement. Using anatomy and contour-based registrations, the radiation therapist's alignments differed by > or = 3 mm from the reference alignments in 24%, 33%, and 3% and 55%, 48%, and 21% of all registrations in the anterior-posterior, superior-inferior, and lateral directions, respectively. The respective values for the marker-based alignments were 3%, 6%, and 3%. The physician's registrations showed the same general trend. The marker-based registrations showed the least amount of inter-user variability while the contour-based ones showed the most.
CONCLUSION: The use of fiducial markers for MVCT image guidance is advantageous to reduce the inter-user variability of the image registration. If fiducial markers are not used, anatomy-based registrations outperform contour-based registrations in terms of (1) agreement with a reference alignment and (2) inter-user variability.