Reconstruction algorithm for tilted helical volumetric CT

With the development of volumetric computed tomography (VCT), technical challenges in image reconstruction increases significantly. Majorities of the proposed algorithms are for helical acquisition in which the gantry is perpendicular to the rotation axis. In many clinical applications, however, the gantry needs to be tilted to avoid direct X-ray exposure to the sensitive organs. If appropriate correction is not rendered, degraded image quality will result. There are two effects when the CT gantry is tilted with respect to the z-axis. The first is that the detector rotation axis and the reconstruction axis is no longer the same. The second effect is that in helical acquisition, the patient table does not travel perpendicular to the detector center plane. These two effects have been studied in the past. Complication arises, however, when the cone beam geometry is first rebinned to a set of parallel samples prior to the filtered backprojection. Under such conditions, the source locations for each projection map to a segment of the source-helix instead of a single point. In this paper, we derive a close-form reconstruction algorithm. We separate the entire process into two components. The first component assumes the detector center plane coincides with the reconstruction plane. The second component deals with the two separate planes. The key step in the reconstruction is that for each pixel, the original coordinate is first mapped to a new coordinate before the backprojection. For validation, we conducted extensive phantom and clinical experiments. Results have shown that the proposed algorithm can successfully remove image artifacts. Since the correction is implemented as part of the backprojection process, little impact on spatial resolution is observed.