Multi-basis-image reconstruction from conventional data acquired in standard CT.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Standard computed tomography (CT) relies on a non-linear data model.
• Reconstructing multi-basis images typically requires specialized CT systems (e.g., dual-energy, multi-spectra, photon-counting).
• Conventional CT data, collected with a single spectrum, is usually insufficient for direct multi-basis image reconstruction.

Purpose of the Study:

• To develop a novel algorithm for accurate and stable reconstruction of multi-basis images from conventional CT data.
• To investigate the inversion of the non-linear data model in standard CT.
• To demonstrate the feasibility of obtaining advanced material decomposition images using a single-spectrum CT system.

Main Methods:

• Formulated the reconstruction as a non-convex optimization program using the basis-region technique and volume-conservation constraint.
• Developed the dynamic non-convex primal-dual (dNCPD) algorithm to solve the inverse problem.
• Validated the algorithm's numerical accuracy with simulated data and evaluated its stability with real-world, noisy data.

Main Results:

• The dNCPD algorithm achieved numerically accurate and stable reconstruction of multi-basis images from conventional CT data.
• Successfully generated virtual monochromatic images alongside the basis images.
• Demonstrated the algorithm's robustness against noise and other physical factors present in real data.

Conclusions:

• The developed dNCPD algorithm enables the reconstruction of multi-basis images directly from conventional, single-spectrum CT data.
• This represents a significant advancement, potentially reducing the need for specialized CT hardware.
• The findings hold theoretical interest and practical implications for broader applications of CT imaging.