CBCT-MOTUS: Model-Based Non-Rigid Motion Estimation and Correction for CBCT Imaging

Cone beam computed tomography (CBCT) imaging has many applications including position verification in radiotherapy. However, patient motion introduces artifacts in the reconstructions, degrading their quality and resolution. To this end, we have developed a framework for joint non-rigid motion estimation and correction from reference images and projection data (CBCTMOTUS). Deformation vector fields are estimated through a spline-based motion model and spatial regularization, reducing the number of parameters to optimize. It is validated in silico and applied on clinical acquisitions. The framework can estimate and correct periodic and random motion with high temporal resolution (182 ms), where the motion-corrected reconstructed images show improved image quality for both experiments, such as deblurring and removal of contour artifacts. In conclusion, we estimate and correct non-rigid motion through CBCT-MOTUS by parametrization of spatially smooth motion.