Evaluating the Robustness of CBCT-Based Dose Calculation across Anatomical Sites and Imaging Platforms for Adaptive Radiotherapy
Abstract
Purpose
To assess dose calculation accuracy on cone-beam computed tomography (CBCT) images relative to the RayStation CBCT correction framework (CBCTcorr) across multiple CBCT platforms, anatomical sites, and clinical scenarios.
Methods
This retrospective study used CBCT images from Elekta XVI, Varian TrueBeam, and Varian Halcyon HyperSight systems. The cohort included thoracic, abdominal, pelvic, head and neck, and sarcoma cases. Two scenarios were evaluated: minimal anatomical change, representing baseline dose calculation accuracy, and substantial anatomical change requiring replanning during the treatment course. For each fraction, dose was calculated on uncorrected CBCT images using an institutional CT-to-density calibration, with missing tissue outside the CBCT field of view propagated from the reference CT, and on CBCT images processed using CBCTcorr. Warped CT (wCT), generated by deformable registration of the reference or replanning CT to the CBCT geometry, served as the dosimetric reference. Dose calculation accuracy was quantified using global dose-difference pass rates from 1% to 5%.
Results
For minimal anatomical change, uncorrected CBCT calculation showed platform- and site-dependent variability that increased with stricter criteria. At the 3% dose-difference criterion, thoracic (lung) median pass rates were 98.2% (XVI), 99.6% (TrueBeam), and 99.8% (HyperSight). At the 1% criterion, median pass rates were 66.0%, 95.5% and 89.7%, respectively. CBCTcorr systematically improved agreement across platforms and sites, increasing lung median 1% pass rates to approximately 99.9%. With CBCTcorr, median 2% pass rates exceeded 99.8% for all sites. In cases with substantial anatomical change requiring replanning, CBCTcorr maintained high agreement, with median 3% pass rates exceeding 95%.
Conclusion
Dose calculation on uncorrected CBCT depends on imaging platform and anatomical site. In this multi-platform evaluation, CBCTcorr consistently improved dose calculation accuracy and demonstrated robustness under clinically realistic anatomical change, supporting its use for fraction-wise dose reconstruction and adaptive decision-making without repeat CT imaging.