FONT-SIR: Fourth-Order Nonlocal Tensor Decomposition Model for Spectral CT Image Reconstruction

Area of Science:

• Medical Imaging
• Image Reconstruction
• Photon Counting Detectors

Background:

• Spectral computed tomography (CT) uses photon counting detectors (PCDs) for image reconstruction.
• Reconstructed spectral CT images often suffer from severe noise due to limited photons and detector counting rates.

Purpose of the Study:

• To propose a novel fourth-order nonlocal tensor decomposition model for spectral CT image reconstruction (FONT-SIR).
• To enhance image quality by reducing noise and artifacts while preserving structural details.

Main Methods:

• Developed a fourth-order nonlocal tensor decomposition model (FONT-SIR).
• Grouped similar patches in spectral and spatial domains to form fourth-order tensor units.
• Utilized principal component analysis (PCA) for feature extraction and cosine similarity for patch comparison.
• Applied weighted nuclear norm and total variation (TV) norms for decomposition into low-rank and sparse components.
• Employed the alternating direction method of multipliers (ADMM) for model optimization.

Main Results:

• The FONT-SIR model effectively removes noise and artifacts from spectral CT images.
• Structural details and spatial relationships within the images are well-preserved.
• Demonstrated superior qualitative and quantitative performance compared to existing state-of-the-art methods on simulated and real data.

Conclusions:

• The proposed FONT-SIR model offers a significant advancement in spectral CT image reconstruction.
• This method provides high-quality spectral CT images with reduced noise and preserved details.
• FONT-SIR shows great potential for clinical applications requiring accurate spectral CT imaging.