Misalignment Detection Accuracy with Dual Orthogonal Kv Imaging and Surface-Guidance for SBRT Spine
Abstract
Purpose
Target alignment is critical for successful SBRT treatment for spinal oligometastatic disease. While CBCT imaging is standard, in-treatment triggered imaging with dual orthogonal kV imaging for internal target monitoring and surface-guidance for external patient motion can further ensure alignment. This phantom study aims to investigate the effects of DRR Volume of Interest (VOI) and contrast on detection accuracy of simulated longitudinal misalignments.
Methods
Three DRR VOI’s (Large_VOI, Medium_VOI, Small_VOI) with variable visibility of ribs and vertebra and six DRR contrast settings (Default Contrast and 25% increments ranging from Bony to Tissue) were used to test auto-registration-based detection of twelve simulated longitudinal misalignments (1mm, 5mm, 10mm, 20mm, 25mm, 30mm superiorly and inferiorly), totaling 216 measurements. An anthropomorphic phantom with spine, ribs, and mandible was positioned with CBCT and intentionally misaligned. The simulated misalignment was verified with dual orthogonal kV imaging (120kVp, 20mAs). A detection accuracy threshold of 0.3mm was chosen to consider couch motion, image registration, and isocenter coincidence uncertainties.
Results
For 25mm and 30mm misalignments, 72% (n=26) and 100% (n=36) of measurements exceeded the threshold, respectively. Within 20mm, superior ISO shifts had 0% (n=0), 16% (n=3) and 5% (n=1) of measurements compared to inferior shifts had 39% (n=7), 16% (n=3) and 61% (n=11) measurements exceeding the threshold for Large_VOI, Medium_VOI and Small_VOI, respectively. Within 20mm, the 75%bony/25%tissue and 50%bony/tissue contrast had 12.5% (n=3) and 17% (n=4) measurements, respectively, exceeding the threshold while all other contrast levels surpassed 25%.
Conclusion
Dual orthogonal kV imaging accurately displayed the superior and inferior shifts within 20mm which matches the vendor registration search range. Potentially due to confounding similar anatomy, the system was more accurate with superior ISO shifts compared to inferior. While 75%bony/25%tissue and 50%bony/tissue contrast were most accurate in this study, conclusions cannot be generalized due to variable patient depth and density.