Model-Based Iterative Reconstruction for Dual-Energy X-Ray CT Using a Joint Quadratic Likelihood Model.

Area of Science:

• Medical Imaging
• Computed Tomography
• Image Reconstruction

Background:

• Dual-energy CT (DECT) offers improved contrast and artifact reduction over traditional CT.
• Model-based iterative reconstruction (MBIR) can further reduce noise and enhance resolution in DECT.
• Direct dual-energy MBIR is computationally complex due to nonlinear forward models.

Purpose of the Study:

• To present a novel joint dual-energy MBIR (JDE-MBIR) method.
• To simplify the dual-energy MBIR forward model while maintaining statistical accuracy.
• To improve image quality in DECT, especially with fast kVp switching.

Main Methods:

• Developed JDE-MBIR using a quadratic approximation of the polychromatic log-likelihood.
• Incorporated an exact nonnegativity constraint in the image domain.
• Accounted for statistical dependencies in material-decomposed sinogram components.

Main Results:

• JDE-MBIR demonstrated favorable image quality compared to decomposition-based methods.
• The method effectively handles inaccuracies in interpolated sinograms from fast kVp switching.
• Both phantom and clinical studies confirmed the superior performance of JDE-MBIR.

Conclusions:

• JDE-MBIR provides a computationally efficient yet statistically robust approach to dual-energy CT reconstruction.
• The method achieves high-quality imaging comparable or superior to existing techniques.
• JDE-MBIR is particularly advantageous for DECT systems employing rapid kVp switching.