Abstract
Methods for computing forward projection such ray-driven, pixel-driven, distance-driven or their interpolated versions, result in varying degrees of angular-dependent errors. New optimized reprojection methods have been developed for use in Single Photon Emission Computed Tomography (SPECT) with parallel-beam geometry, which provide superior corrections of these errors. These methods are based on a comparative angular study between various conventional reprojection methods. This optimization is based on a selection of the operator which provides the minimum of errors for a given projection angle. This method was compared to the conventional methods using the digital Shepp-Logan phantom. The normalized mean square error, the normalized mean error and the max percent error between reprojected sinograms and a standard reference were calculated. Results show that the proposed operators have at least comparable performance. This work also investigates a comparative study on iterative reconstruction task with ML-EM. The preliminary result of this study shows that these optimized reprojection operators can promise more accuracy of reconstructed images.