Helical cone beam reconstruction in volumetric CT: maintaining a large filed-of-view (FOV) at very high pitches

摘要

In volumetric CT, the suppression of cone beam (CB) and helical artifacts is very challenging. At very high pitches larger than 1.5, not only artifact suppression but also maintenance of a large field of view (FOV) become even more challenging. At large helical pitches, the data redundancy corresponding to each pixel within a reconstruction FOV varies dramatically. An inappropriate dealing with the fast-varying data redundancy over pixels to be reconstructed can result in severe shading/glaring artifacts in reconstructed images and lead a clinical diagnosis impossible. One existing approach to combat the non-uniform data redundancy and extend the FOV at very high pitches is to reconstruct images on tilted planes. However, tilting a reconstruction plane can significantly impacts data flow of filtered backprojection -the most attractive one practiced by all major CT manufacturers - and slows down image generation speed considerably. Along with an experimental evaluation, a helical CB filtered backprojection reconstruction algorithm using three-dimensional view weighting (namely 3D view weighted CB-FBP algorithm) is proposed here. The novelty of the proposed algorithm is the 3D-nature of the weighting function, which enables the proposed algorithm to reach the optimal image qualities and provides the freedom in controlling image quality over accuracy, noise characteristics, spatial resolution and temporal resolution. Particularly, it is the 3D view weighting that enables the proposed algorithm to maintain an FOV at helical pitch larger than 1.5, which is as large as that at helical pitches below 1.0. It is believed that the proposed 3D view weighted CB-FBP algorithm is applicable to and robust over all CT imaging applications, including diagnostic imaging and dynamic imaging for functional evaluation. It is also believed that the proposed algorithm will work well for larger cone angles when more sophisticated 3D view weighting strategies are employed.