In Image-Guided RadioTherapy (IGRT) of lung tumors, patients undergo a 4D CT, on the basis of which their treatment is planned. It is implicitely assumed that their breathing motion will not change much throughout the treatment. At the beginning of a treatment fraction, a Cone Beam CT (CBCT) acquisition is performed, and used to re-position the patient. Obtaining a 4D reconstruction from this cone beam data would allow the therapists to check whether the breathing motion of the day still matches that of the planning CT, and if not, take appropriate corrective actions. But 4D tomog-raphy from a single cone beam CT acquisition implies a severe lack of projection data, and efficient methods have only started to appear during the last few years. Some perform regularization along time to explicitely enforce similarity between consecutive frames, which considerably improves image quality. In IGRT, breathing motion can be estimated on the 4D planning CT, and used to refine the 4D CBCT reconstruction results. We describe a 4D CBCT reconstruction method that combines regular-ization techniques with the use of motion information.