Document Type

Conference Proceeding

Publication Date


Publication Title

Medical Physics


Purpose: On Elekta linacs, cone-beam CTs (CBCT) can be acquired during VMAT delivery to monitor intrafraction motion during treatment. Stereotactic spine treatments can be lengthy, increasing patient motion likelihood during treatment. We evaluated during-treatment CBCT (DT-CBCT) feasibility to assess patient motion for stereotactic spine treatments in place of a mid-treatment CBCT. Methods: CBCTs from seven spine treatments were analyzed retrospectively. Various fractionation schemes were used (16-25Gy delivered in 1-5 fractions). All plans contained two VMAT arcs of 2209.5-13,332.4 MUs with treatment times of 7-58 minutes. Projections from pre-correction CBCTs were combined with post-correction CBCT to simulate patient motion (1cm, 2 deg) occurring after 50% to 90% of the first beam being delivered. Images were reconstructed and registered (bony alignment) in Elekta XVI 5.0 by an imaging specialist. For reconstructions, image quality and the difference from true CBCT shifts were assessed. Ground truth shifts ranged from translations of 0.60-1.12cm and rotations of 0.0-1.9 degrees. Results: Simulated motion impacted DT-CBCT image quality, displaying double anatomy and decreased delineation of vertebral bodies when patient movement occurred after 50%-80% of the first beam delivery. For all treatments, when motion occurred in this range it was obvious to the imaging specialist that the CBCT should not be used. However, image quality was not impacted for 6/7 treatments when simulated motion occurred after 90% of first beam delivery. Registration of these images resulted in erroneous patient translations and rotations from the ground truth up to 1.02cm and 1.9 degrees. Conclusion: Even large patient motion that occurs near the end (at 90%) of the beam 1 delivery is not detectable with DT-CBCT resulting in erroneous patient setup for the 2nd beam delivery. Use of a mid-treatment CBCT for stereotactic spine treatments is recommended.





First Page


Last Page