Click here to


Are you sure ?

Yes, do it No, cancel

Reconstruction of a Respiratory Motion Model and a High-Quality Reference Image From CBCT Projections Acquired in Treatment Position. A Retrospective Study Using Patient Images

M Guo1,2*, G Chee1 , D O'Connell1 , J Fu1 , K Singhrao1 , D Ruan1 , P Lee1 , D Low1 , J Zhao2 , J Lewis1 , (1) UCLA School Of Medicine, Los Angeles, California, (2) Shanghai Jiao Tong University, Shanghai


(Thursday, 8/2/2018) 1:00 PM - 3:00 PM

Room: Room 202

Purpose: To reconstruct a motion model and dynamic image sequence from free-breathing cone-beam projections and external surrogate measurements acquired with a patient in treatment position. These results could be used to generate 3D images representing deforming patient anatomy during treatment delivery.

Methods: The proposed method was grouped into two parts: A) reference image reconstruction, and; B) motion model construction. We first divided the projections into several bins (6–8) according to their breathing amplitudes. In part A, projections of each bin were used to reconstruct a reference image using a motion compensated simultaneous algebraic reconstruction technique (McSART). We incorporated motion compensation into both the projector and back-projector using the estimated motion model (initialized as 0, and updated in part B). In part B, the reference images were used to update the respiratory motion model. The 5D respiratory model was used. By alternating between parts A and B, we obtained reference images and motion parameters.

Results: Eight patient CT datasets of the thorax and upper abdomen regions were used to evaluate the method. Average HU differences compared to “ground truth� fast helical CT images were 42.4±70.2 HU. The average voxel displacement between the reconstructed image and the ground truth image at each phase, used to evaluate the motion model accuracy, was 0.36±0.07 mm.

Conclusion: We have developed a new motion-compensated CBCT reconstructing method that provides a high-quality reference image and a motion model representing the patient’s deforming anatomy at the time of treatment. The reference image and corresponding motion model can be used to generate images representing arbitrary breathing amplitudes and phases. This could be used to generate 3D images during treatment delivery for targeting verification, delivered dose calculations, and adaptive radiotherapy. In future work, we will evaluate the presented method using prospectively acquired CBCT projections.

Funding Support, Disclosures, and Conflict of Interest: Partially funded by Varian master research agreement


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email