Improving axial image reconstruction by off-centering ROI with data truncation and 3D weighted cone beam DBPF algorithm

We have proposed to improve the accuracy of axial image reconstruction by off-centering the region of interest (ROI) to be imaged and reconstructing the images using the 3D weighted cone beam (CB) filtered backprojection (CB-FBP) algorithm. By applying the 3D weighting scheme at each voxel within the off-centered object, the minimum cone angle of the ray out of the conjugate ray pair corresponding to the vast majority of the voxel can be effectively reduced, and thus reduce CB artifact in reconstructed image. In practice, a small to medium off-centered object, e.g., head or extremities, may readily be within the scan field of view (FOV) determined by the gantry geometry and latitudinal detector span, and accordingly the voxel-wise 3D weighted CB-FBP algorithm can be applied straightforwardly. However, it may be inevitable for a large off-centered object, e.g., the thorax embodying the heart, to be partially outside the scan FOV, which may result in severe latitudinal truncation in projection data acquisition and accordingly artifacts in reconstructed image if the FBP algorithm is employed. Recognizing the strength of derivative backprojection filtering (DBPF) algorithm in dealing with the latitudinal truncation, we propose to apply the voxel-wise 3D weighting scheme in the DBPF reconstruction. Preliminary results from the numerical simulation study demonstrate that the DBPF reconstruction algorithm can substantially reduce the artifacts caused by the latitudinal data truncation, while the voxel-wise 3D weighting can effectively suppress the CB artifacts within the off-centered ROI. (orig.)