Sinogram restoration in computed tomography with an edge-preserving penalty

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Iterative penalized-likelihood reconstruction is computationally intensive.
• Quadratic penalties in sinogram restoration do not effectively preserve edges.
• Nonquadratic penalties offer potential for improved edge preservation.

Purpose of the Study:

• Derive a restoration update equation for nonquadratic penalties.
• Extend sinogram restoration to helical cone-beam geometry and clinical data.
• Evaluate edge-preserving penalties for improved image quality.

Main Methods:

• Derived restoration update equation using separable parabolic surrogates (SPS).
• Proposed method for calculating sinogram degradation coefficients for helical cone-beam geometry.
• Performed sinogram restoration with quadratic and Huber penalties, followed by analytical reconstruction.

Main Results:

• Huber penalty sinogram restoration yielded better resolution-noise trade-offs than quadratic penalty.
• Edge preservation in sinogram domain is influenced by object size and contrast.
• Sinogram restoration is feasible for 3D helical cone-beam CT, with pitch having minimal impact.

Conclusions:

• Sinogram restoration with Huber penalty offers superior resolution-noise performance compared to quadratic penalty.
• Huber-penalty sinogram restoration is feasible for helical cone-beam CT and applicable to clinical data.