Presenter(s)
H. Yang1, J. Park2, D. Han1, and H. Cho2; 1Yale school of medicine, New Haven, CT, 2Yonsei University, wonju, Korea, Republic of (South)
Purpose/Objective(s): To develop a respiratory target motion analysis framework using routine on-board three-dimensional cone-beam CT (3D CBCT) without external optical surrogate systems for stereotactic body radiotherapy (SBRT). Upper abdomen organs, including the liver and pancreas, exhibit significant respiratory motion, which complicates accurate target localization and image registration. When registration confidence is low, internal target motion is extracted directly from pretreatment CBCT projections without additional imaging. Four-dimensional cone-beam CT (4D-CBCT) reconstruction can be selectively triggered as needed.
Materials/Methods: A target motion model within the framework was developed to quantify fiducial marker trajectories directly from CBCT projections. Continuous respiratory phase was derived from a detrended Superior-Inferior (SI) surrogate using analytic signal processing. Anterior–posterior (AP) and Left-Right (LR) motion were parameterized using phase-dependent first-order harmonic functions. Three-dimensional marker positions were iteratively refined via nonlinear least-squares optimization by minimizing reprojection error between measured and model-predicted 2D marker positions. Robust Huber weighting and mild regularization improved numerical stability.
Results: The proposed framework was validated using a respiratory motion phantom with implanted 2 mm diameter radiopaque marker. The estimated respiratory displacements in the phantom were 1.15 mm in LR, 16.03 mm in AP, and 5.99 mm in SI. Trajectory optimization achieved mean reprojection errors of 0.863 mm and 0.130 mm along the detector’s lateral and longitudinal directions, respectively. Internal fiducial tracking quantified simultaneous SI and AP displacement, with a mean AP–SI peak shift of approximately 60 ms, demonstrating near-synchronous respiratory motion. As the AP displacement was largest in the phantom experiment, retrospective phase sorting was performed based on the AP surrogate signal, resulting in reconstruction of phase-resolved 4D-CBCT volumes.
Conclusion: The proposed framework enables accurate multi-directional respiratory target motion estimation directly from routine on-board 3D CBCT projections and supports selective phase-resolved 4D-CBCT reconstruction. These results demonstrate the feasibility of projection-based daily internal motion monitoring in SBRT without additional imaging.