Studies on circular isocentric cone-beam trajectories for 3D image reconstructions using FDK algorithm

Image reconstruction from X-ray cone-beam projections collected along a single-circular source trajectory is commonly done using the Feldkamp (FDK) algorithm which performs well only with a small cone-angle. Although this method does not provide an "exact" reconstruction, the approximation is considered adequate for many purposes. In FDK reconstruction the degree of inaccuracy is highly object-dependent, and the largest errors are to be expected for planes parallel to and remote from the midplane. In this study we investigated the possibility to accurately reconstruct these regions by applying FDK algorithm along three-orthogonal to each other circular scanning trajectories. After appropriate weighting, based on the expected errors for each individual reconstruction, the final 3D volume contains the most precisely recovered values. By comparing the quality of 3D reconstructed images using FDK algorithm on projections acquired along classical single-circular and two- and three-orthogonal circular trajectories, we show that using three-orthogonal circular isocentric orbits with an error-based weighted averaging, image quality of reconstructed slices significantly improves, reconstruction error due to circular scanning is reduced and becomes almost independent of the slice position even for relatively large cone-angles.