Click here to


Are you sure ?

Yes, do it No, cancel

CT Based Simulation Framework for Motion Artifact and Ground Truth Generation of Cone-Beam CT

P Paysan1*, P Munro1, S Scheib1, (1) Varian Medical Systems Imaging Laboratory, Daettwil AG


(Sunday, 7/14/2019) 3:30 PM - 4:00 PM

Room: Exhibit Hall | Forum 9

Purpose: Patient motion is a major reason for artifacts in cone-beam CT (CBCT). One cause of motion is respiration that can be either suppressed via breath-hold or resolved by 4D reconstruction. However, residual motion such as breath-hold level drift, loss-of-breath, gas motion, peristalsis or muscle relaxation can still impair such reconstructions. Development of algorithms mitigating motion requires anatomic realism combined with ground-truth (GT) information. Such a requirement prevents the use of acquired CBCT patient data as well as the use of simple mathematical phantoms. We propose a (4D-)CT based method to simulate CBCT acquisitions with realistic respiratory motion and GT.

Methods: To simulate (4D-)CT based respiratory motion we apply the DEEDS deformation algorithm to non-rigidly register peak-inhale against peak-exhale phases. The registration yields a patient specific deformation vector field (DVF) that represents a linear approximation of the peak-to-peak deformation during regular breathing. By applying linear interpolation of the DVFs we simulate different inspiration levels. This enables us to use recorded breathing-curves to define breathing patterns. To simulate typical 17sec to 2min CBCT image acquisitions (400-900 projections), we interpolate the DVF according to the normalized breathing amplitude for each x-ray projection acquisition, deform the respective CT phase accordingly and forward project it. For reconstruction without motion streaks (but motion blurring) we calculate an average of all deformed volumes and a GT of choice using the DVF.

Results: The method synthesizes individual and realistic breathing motion expected during CBCT acquisition and provides GT. The flexibility in choosing various anatomical locations and breathing patterns enables training data generation for machine learning based motion artifact mitigation.

Conclusion: The method yields realistic looking motion artifacts and has been applied for regular as well as irregular breathing patterns. Future improvements will include the use of piecewise linear interpolation between several breathing phases to simulate nonlinear deformations.

Funding Support, Disclosures, and Conflict of Interest: All authors are full time employees of Varian Medical Systems


Cone-beam CT, Simulation


Not Applicable / None Entered.

Contact Email