Stereotactic Body Radiation Therapy for Non-Spine Bone Metastases: A Single Institution's Experience

Purpose/Aim: Liver cancer patients with minimal amplitude of breathing or unable to tolerate breath-hold for radiotherapy are treated free-breathing with image guided SBRT. The purpose of this study is to assess the accuracy and reproducibility of automated registration of 3D versus 4D CBCT in target localization for free-breathing liver stereotactic body radiotherapy (SBRT). Method/Process: Daily localization CBCT images are acquired and any positional discrepancies corrected prior to treatment delivery for free-breathing liver SBRT. For image-guided radiotherapy (IGRT) at our institution, amplitudes of breathing 5 mm undergo 3D-CBCT acquisition while larger breathers (u003e 5 mm) utilize a 4D-CBCT workflow for target localization. Retrospective analysis of translational (T) and rotational (R) displacements of treated fractions was performed through 3 methods: 3D-CBCT dual registration using bony anatomy (T+R) and average liver position (T+R); 3D-CBCT registration using average liver position (T+R) followed by 4D-CBCT registration to the liver exhale position (T); and 4D-CBCT registration to the liver exhale position (T). Each retrospective image registration was performed automatically by the imaging software, and evaluated by three independent observers who performed manual adjustments to ensure clinical target coverage. Krippendorff’s alpha was used to assess agreement between the methods. Results/Benefits/Challenges: Ten liver cancer patients treated for 30-50 Gy over 5 or 6 fractions, with amplitudes of breathing motion ranging from 3.6 13 mm, were included. A total of 54 CBCT images were analyzed by each observer. Mean (+/SD) observer disagreement for 3D-CBCT dual registration was 1.1 0.8 mm, 2.1 1.3 mm, and 1.4 1.3 mm in the right-left, cranio-caudal and anterior posterior direction respectively. Agreement between observers was very high with Krippendorff’s alpha greater than 0.81 in all directions. Dual registration on 3D-CBCT and 4D-CBCT of the liver produced similar shifts; the root-mean-square (RMS) difference between mean manual and automatic techniques was 0.75mm in a 3D vector. Registration using 4D-CBCT to the liver exhale position alone produced an RMS observer error of 0.72 mm. Automatic matches of 3D and 4D liver registrations converged at small amplitudes, indicating a single 4DCBCT workflow is sufficient for any amplitude of breathing motion. Inter-observer variability between the 3 observers and automatic imaging software matches was smallest with 4D-CBCT liver exhale registration alone. Conclusion/Impact/Outcomes: Automatic registration using 4D-CBCT provides accurate target localization and is unaffected by larger breathing amplitudes as seen with 3D-CBCT image registration. Fourdimensional CBCT reduces the inter-observer variability and potential for systematic error, making it a desirable approach for free breathing liver IG-SBRT.