An image reconstruction strategy for truncated projections of planar object in cone-beam CT

Abstract

In the field of cone-beam computed tomography (CT), truncation artifacts caused by projection truncation will decrease the quality of reconstruction image. For planar object with large area to thickness ratio, projections are truncated at some views but not all views. Reconstruction of planar object is a special problem and rarely discussed ever before. In this paper, an image reconstruction strategy based on back-projection filtration (BPF) algorithm and projection onto convex set (POCS) method for truncated projections of planar object is developed. By considering the relationship between thickness of planar object and scan orbit radius, two cases are discussed in the strategy. In the first case, PI-lines parallel to the thickness direction which are non-truncated are selected and reconstructed using the BPF algorithm. In the second case, PI-lines parallel the thickness direction with no truncations are selected and reconstructed by use of BPF method firstly. And then, with the reconstructed result as a priori knowledge in the first reconstruction, PI-lines parallel to the width direction are selected and reconstructed by use of the POCS method. The proposed strategy is validated by simulations and real experiments. Results show that reconstruction image quality of our method is superior to that of the conventional FDK algorithm.