Visualization of the iterative process.
The two orthogonal circles trajectory used for simulations of the 3D Forbild head phantom. The two solid lines mark the two circular source trajectories. The first source uses 80 kV tube voltage and rotates around the axis. The second source uses 140 kV tube voltage and rotates around the axis.
Definition of the micro-CT mouse phantom. Materials and densities are listed in Table III.
Reconstruction results of the Forbild head phantom. The left part of the figure shows the transversal center slice and the right side shows the sagittal center slice. The image-based reconstructions and the MDIR reconstructions use the two orthogonal circles as source trajectories (see Fig. 2). The first row shows a raw data-based reconstruction from geometrically consistent rays, where both sources rotate around the axis. The image-based method yields banding artifacts between bones and a general capping of the water-equivalent brain matter. Furthermore, the difference images in the fourth row reveal that the CT values of bone are incorrect in the monochromatic composite image. The proposed MDIR method corrects those errors and the remaining entries in the difference images in the fifth row are caused by different spatial resolution and different aliasing artifacts that are caused by the different source geometries.
Reconstruction results of the mouse phantom simulation. The top row shows the standard images and the second row shows the material decomposition results when using the image-based standard method. The results after two iterations of the proposed MDIR method do not show any banding artifacts like the image-based method does. The difference images in the last row reveal that besides artifacts there is a significant correction of the mean density values, and further the absence of noise and edges in those images proves that the spatial resolution is maintained using MDIR.
Measurement results of the PE-HA400 phantom scanned with a clinical CT scanner. In the image-based reconstruction (top row), artifacts are very prominent between the HA400 inserts. Those artifacts are completely corrected after only one iteration of MDIR (second row). Thereby, the difference images show that the spatial resolution and the image noise are constant.
Profile through the PE density image using image-based reconstruction and MDIR. The dashed line marks the true value.
Profile through the HA400 density image using image-based reconstruction and MDIR. The dashed line marks the true value.
Image quality and applicability to inconsistent rays of selected DECT reconstruction methods. All methods are applicable to consistent rays.
Overview of the vector notation used throughout this paper.
Materials and densities of the micro-CT mouse phantom definition according to Fig. 3.
Quantitative evaluation of one ROI of homogeneous bone (left) and another ROI of homogeneous water in the Forbild head phantom using different material decomposition methods. The mean value and the standard deviation are evaluated on noise-free images.
Article metrics loading...
Full text loading...