Anti-aliasing Weighting Functions for Single-Slice Helical CT
PJ La Riviere,XC Pan
DOI: https://doi.org/10.1109/tmi.2002.803104
IF: 10.6
2002-01-01
IEEE Transactions on Medical Imaging
Abstract:Spatially variant longitudinal aliasing plagues most volumes reconstructed from single-slice helical computed tomography data, and its presence can degrade resolution and distort image structures. We have recently developed a Fourier-based approach to longitudinal interpolation in helical computed tomography that can, for scans performed at pitch 1 or lower, essentially eliminate this longitudinal aliasing by exploiting a generalization of the Whittaker-Shannon sampling theorem whose conditions are satisfied by the interlaced pairs of direct and complementary longitudinal samples. However, the algorithm is computationally intensive and cannot be pipelined. In this paper, we address this shortcoming by deriving two spatial-domain, projection-data weighting functions that approximate the application of the Fourier-based approach, and preserve its aliasing suppression properties to some degree, while allowing for a pipelined implementation. The first approach, which we call simply 180AA, for anti-aliasing, is a direct spatial-domain approximation of the 180FT approach. The second approach, which we call 180BSP, is based on an approximate generalized interpolation approach making use of B-splines. Studies of aliasing and resolution properties in reconstructions from simulated data indicate that while the 180AA and 180BSP approaches do not perfectly replicate the favorable aliasing suppression and resolution properties of the 180FT approach, they do represent an improvement over the clinically standard 180LI approach on these fronts.