Radon Space and Frequency Space Interpolation Methods for Circular-Orbit Cone-Beam Imaging

Circular-orbit cone beam imaging is widely used to obtain 3D medical images, due to its simple mechanical setup. However, the circular-orbit cone-beam projection data do not satisfy Tuy\´s data-sufficiency condition, and are not sufficient for artifact-free image reconstruction. Many researchers, such as H. Yang, H. Bruder, B. Nett and so on, have investigated some possible approaches to reduce the artifacts due to data insufficiency. Data extrapolation methods may help in "completing" the unmeasured projection. In the paper, we consider two data extrapolation methods, one in the Radon space (i.e., the projection space), and the other one in the frequency space (i.e., the Fourier transform of the Radon space). These two methods are implemented using computer simulations. Neither of these methods is as good as the image reconstructed with an iterative ML-EM algorithm. We believe that in order to further reduce the data insufficiency artifacts, an iterative POCS (projection onto convex sets) method should be used.