Spectral CT reconstruction via low-rank representation and structure preserving regularization

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Spectral computed tomography (CT) acquires multi-energy data, but photon collection limitations in narrow energy bins degrade image quality.
• Conventional CT methods are insufficient for the unique challenges of spectral CT data.

Purpose of the Study:

• To develop an advanced spectral CT reconstruction algorithm.
• To improve image quality and material decomposition accuracy in spectral CT.

Main Methods:

• A novel algorithm combining low-rank representation and structure-preserving regularization for spectral CT reconstruction.
• Utilizing inter-channel correlation and gradient domain sparsity as prior regularization terms.
• A split-Bregman iterative algorithm and a multi-channel adaptive parameter strategy were developed.

Main Results:

• The proposed algorithm demonstrated superior reconstruction accuracy and material decomposition compared to SART, TVM, LRTV, and SSCMF.
• Achieved an average 40.4% improvement in feature similarity (FSIM) over SART in numerical simulations.
• Significant improvements were observed in both numerical simulations and real mouse data.

Conclusions:

• The developed multi-channel reconstruction algorithm is tailored for spectral CT imaging.
• The algorithm offers substantial improvements in image quality, showing significant potential for spectral CT applications.