How low can you go? A CBCT dose reduction study

Abstract Purpose Cone beam computed tomography (CBCT) is often used for patient setup based solely on bony anatomy. The goal of this work was to evaluate whether CBCT dose can be lowered to the level of kV image pair doses when used for bony anatomy‐based IGRT without compromising positioning accuracy. Methods An anthropomorphic phantom was CT scanned in the head, head and neck, chest, and pelvis regions and setup on the linear accelerator couch with the isocenter near the planned location. Cone beam computed tomographies were performed with the standard “full dose” protocol supplied by the linac vendor. With sequentially lowering the dose, three‐dimensional (3D) matching was performed for each without shifting the couch. The standard kV image pair protocol for each site was also used to image the phantoms. For all studies, six degrees of freedom was included in the 2D or 3D matching to the extent they could be employed. Imaging doses were determined in air at isocenter following the TG‐61 formalism. Results Cone beam computed tomography dose was reduced by 81–98% of the standard CBCT protocol to nearly that of the standard kV image pair dose for each site. Relative to the standard CBCT shift values, translational shifts were within 0.3 and 1.6 mm for all sites, for the reduced dose CBCT and kV image pair, respectively. Rotational shifts were within 0.2 degree and 0.7 degrees for all sites, for the reduced dose CBCTs and kV image pair, respectively. Conclusion For bony anatomy‐based image guidance, CBCT dose can be reduced to a value similar to that of a kV image pair with similar or better patient positioning accuracy than kV image pair alignment. Where rotations are important to correct, CBCT will be superior to orthogonal kV imaging without significantly increased imaging dose. This is especially important for image guidance for pediatric patient treatments.

select patients. This daily CBCT imaging has in many cases replaced daily or weekly MV imaging performed in the era before onboard kV imaging existed. Full three-dimensional (3D) imaging information coupled with a six degrees of freedom (6DoF) treatment couch provides state-of-the art patient positioning, enabling accurate treatment delivery.
Recent interest in limiting radiation exposure from diagnostic CT exams, especially for children, by tailoring the technique to the size of the patient, has carried over to considerations for IGRT dose.
Although radiotherapy patients will get orders of magnitude greater dose from their treatment compared to the IGRT dose inside their treated volume, the additional daily CBCT dose outside the treated volume is comparable to the scatter dose. 2 In addition, as the magnitude of imaging dose is inversely related to the body mass index (BMI), 3 and children have in general lower BMI values, imaging dose magnitude is higher in children for the same imaging protocol used in adults. Therefore, there has been an interest in quantifying the dose from CBCT for IGRT and potentially reducing that dose if feasible, if only to adhere to ALARA principles. This has been stressed in the AAPM Task Group Report 180 as well, trying to reduce the dose burden from imaging while considering the risks and benefits of imaging to the patient. 4 When imaging pediatric cases, some clinicians have elected not to use CBCT daily, but instead to perform lower dose orthogonal kV imaging daily or even weekly imaging. 5 This lack of full 3D information may reduce the patient positioning accuracy and understanding of changes in anatomy during treatment. In this work, we have investigated the potential for orders of magnitude dose reduction for CBCT to be comparable to kV image pair dose, which is low enough to not generally cause concern. The consequence of reducing the dose for any imaging procedure is the potential for loss of information and reducing or eliminating the usefulness of the image.
In this study, we focus on the scenario where only bony anatomy is being used for image registration and matching during IGRT. Where soft tissue delineation is required, CBCT dose reduction may not be possible, at least not to the extent we explored. There have been a few studies that have explored low-dose IGRT, especially for pediatrics 6,7 but we are not aware of any prior study that aimed to answer the question, "how low can you go" in the context of CBCT dose for radiotherapy IGRT.

| MATERIALS AND METHODS
An anthropomorphic phantom (Alderson Rando, Phantom Laboratory, RSD, Inc., Long Beach, CA) was CT scanned in the head, head and neck, chest, and pelvis regions and a plan for each was created in the treatment planning system so that an isocenter at each site could be created. The phantom was setup on Varian TrueBeam (Varian, Palo Alto, CA) linear accelerators with 6DoF couch with the isocenter near the planned location, but with about 0.3 to 1.5 cm translational shifts and 0.2-2.9 degree rotational shifts imposed to make the process more clinically realistic. CBCTs were performed with the standard "full dose" protocol supplied by the linac vendor.
The lower dose protocols were achieved by lowering the mAs, kVp, and frame rate from standard protocols (Table 1). We started with the default kVp and mAs provided on the TrueBeam. The kVp was then reduced from 100 to 80 for head and head and neck, 125 to 100 for chest, but kept at 125 kVp for pelvis. We then progressively reduced the mAs, imaged the phantom using the new protocol, and registered the reduced dose CBCT to the planning CT. As long as the shifts agreed with the full dose CBCT within 0.3 mm and 0.3 degrees, we continued to reduce the mAs. We eventually arrived at the lowest mAs the system allowed, so to continue to reduce the dose, we lowered the frame rate. The frame rate was by default 15 frames per sec (fps) but was reduced to either 7 or 3 fps as needed to achieve lower doses. Other combinations of reducing the kVp, mAs, or frame rate are possible with potentially equivalent results but were not tested. While sequentially lowering the dose, image matching between CBCT and planning CT was performed, without applying the shifts.
Although the auto match feature is commonly used in many clinics, manual image matching can also be performed and can produce slightly different shift results. For consistency of image matching in this study, the auto match feature was used for matching for all sites except for the head, where three radiation therapists independently performed the matching in addition to the auto shift being performed. This was done to compare the auto shift algorithm with manual matching and to demonstrate that the auto shift algorithm is reasonable to rely on for this study.
The standard kV image pair protocol for each site was also used to image the phantoms. Two-dimensional (2D)-3D auto matching was performed for head and neck, chest and pelvis phantoms while 2D-2D auto matching and manual matching were performed for the head phantom. For all studies, 6DoF was included in the 3D matching. In the case of the 2D-2D matching using kV image pairs for the T A B L E 1 Image protocol parameters. NA = not applicable.  Table 3).
The shifts for the head site were performed both by three expe-   as the site location changed from head to L spine. They also found roll to be responsible for the largest deviations from the CBCT result. 13 These image accuracy considerations are especially important for image guidance where highly precise patient positioning is required to optimize normal tissue sparing.

| CONCLUSIONS
Where bony anatomy matching is appropriate, substantially lower AUTHOR CONTRIBU TI ON S TATEMENT Arthur J. Olch and Parham Alaei both contributed measurements, data analysis, and manuscript preparation and approve the final submitted version of the manuscript.