Affiliations: Institute of Image Processing & Pattern
Recognition, Xi'an Jiaotong University, Xi'an, Shaanxi, 710049, China | CT/Micro-CT Lab, Department of Radiology, University
of Iowa, Iowa City, IA, 52242, USA
Abstract: In the X-ray CT research area, the linear integral projections of
mathematical phantoms were extensively used to evaluate the proposed algorithms
with a hypothesis of monochromatic model. However, X-ray tube emits a spectrum
of energies and hence it becomes difficult to study some important aspects,
such as the reconstructed density accuracy of special reconstruction algorithm
and beam hardening effects of polychromatic imaging. If the energy spectrum of
X-ray tube, geometry of phantom, and material compositions of all sub-regions
are known, the actual projections can be simulated with more accuracy.
Motivated by this fact, we propose an improved three-step discrete scheme to
simulate the X-ray projections based on physical model. First, CT-numbers of
all sub-regions of a phantom are determined in advance. Next, CT-numbers are
decomposed into different material compositions under the constraint that
sub-regions correspond to different human tissues. Finally, the projections are
measured according to the energy spectrum distribution of X-ray tube, using the
knowledge of X-ray imaging physics and our strategy to simulate the projecting
process. Our scheme requires significantly less storage and computation to
achieve a higher accuracy and precision. To demonstrate the feasibility of our
scheme, we use the well-known FORBILD head phantom to generate projections and
the Feldkamp algorithm for image reconstruction. The anticipated beam hardening
effects are clearly seen in images reconstructed from projections simulated
using our scheme. This scheme can be utilized in many medical X-ray imaging
research aspects, such as algorithm design, performance analysis, distortion
calibrations, artifact reduction, etc.
Keywords: X-ray projection, physical imaging model, energy spectrum, material composition, FORBILD phantom, Feldkamp algorithm
