Motion-compensated 4D cone-beam computed tomography

In image-guided radiation therapy (IGRT) beside the linear particle accelerator an additional kV system provides information for an accurate patient positioning. The precise execution of the treatment plan is assured that way. An extra goal is the refinement of further planned treatment fractions based on the on-board imaging system information. Additional inter-fractional planning CTs become thus obsolete. However, the acquisition time of the system is much longer than the patient´s breathing cycle due to technical limitations on the gantry rotation speed. Severe artifacts like blurring or streaks are the consequence. They affect image quality and thus also the refinement of the treatment plan. A method for the motion-compensated reconstruction of high quality respiratory-correlated 4D volumes from flat panel detector cone-beam CT (CBCT) scans with slowly rotating gantry is proposed. The motion vector field estimation is done by a deformable registration algorithm. This estimation does not incorporate a planning CT to reduce the influences of intra- and inter-fractional variations in patient´s motion and tissue. We propose to use the phase-correlated reconstruction method of McKinnon and Bates. Thus determined motion vector fields are applied to conventional phase-correlated reconstructions to deform the volumes of all respiratory phases into one single phase. The proposed method is verified by processing simulated rawdata and patient data. The simulated rawdata are obtained by deforming a clinical patient dataset using realistic deformation vector fields. The patient data were acquired with a flat panel cone-beam CT scanner. The results show noise levels comparable to 3D standard reconstructions due to the increased dose usage of the motion compensation approach. The method furthermore outperforms the phase-correlated Feldkamp reconstruction regarding undersampling artifacts and the McKinnon-Bates algorithm regarding noise. In contrast to well-known publication- on motion compensation with low sampled datasets the proposed algorithm has increased robustness to undersampling artifacts and and does not require a priori knowledge like planning CT to reduce extra-fractional factors.