A helical cone beam filtered backprojection (CB-FBP) reconstruction algorithm using three-dimensional (3D) view weighting

摘要

Although they are approximate only from the theoretical point of view, the FDK cone beam filtered backprojection (CB-FBP) reconstruction algorithm for a circular source trajectory and the general CB-FBP algorithm for other source trajectories, including a helix, have been utilized for decades in extensive applications. A common feature of these two algorithms is that the filtering operation is carried out along each horizontal detector row. Although the horizontal filtering works well for small cone angles, it has been shown that, in applications where a helical source trajectory is employed, the filtering step needs to be carried out along a direction tangential to the source helix to improve reconstruction accuracy. Furthermore, view weighting techniques have to be applied in helical CB-FBP algorithm to suppress streak artifacts caused by data inconsistency. Usually, the view weighting techniques are directly borrowed from their fan beam counterparts, which work well at relatively low helical pitch, but fail at high helical pitches. In this paper, we propose a new helical CB-FBP reconstruction algorithm. The filtering in the proposed algorithm is naturally carried out along the tangential direction of the helical source trajectory after a row-wise fan-to-parallel rebinning. A 3D weighting function (which is dependent on view angle β, ray distance t from the ISO, and cone angle α), is utilized to: (a) suppress artifact caused by data inconsistency; (b) improve noise characteristics by using redundant projection data; (c) extend filed of view (FOV) at high pitches by maintaining data redundancy. Both computer simulations and real phantom scans are used to evaluate the proposed helical CB-FBP algorithm. Experimental evaluation verifies that the proposed algorithm produces excellent results from the perspective of reconstruction accuracy, noise characteristics, large FOV at high pitches, and temporal resolution.