Irreproducible motion of either the patient or the device during a cone-beam X-Ray scan remains a major issue limiting reconstruction quality in many practical applications. Computational approaches are starting to emerge, which allow to model general motion parameters during the reconstruction itself. Besides, intelligent image processing on the projection data may reveal clues about "what went wrong" during a scan. We present a novel algorithm which uses a combined analysis in projection and reconstruction space, to both detect and account for unknown motion. This allows not only for the detection of large-scale, non-periodic bulk motion, but also an automatic recovery of it, required for a reconstruction void of artifacts. Using the proposed method, we can restore the reconstruction of clinical head scans with severe unknown motion. Moreover, we evaluate our method on synthetic data with known motion trajectories in a radiotherapy scenario.     «

Irreproducible motion of either the patient or the device during a cone-beam X-Ray scan remains a major issue limiting reconstruction quality in many practical applications. Computational approaches are starting to emerge, which allow to model general motion parameters during the reconstruction itself. Besides, intelligent image processing on the projection data may reveal clues about "what went wrong" during a scan. We present a novel algorithm which uses a combined analysis in projection and r...     »