Iterative spectral CT reconstruction based on low rank and average-image-incorporated BM3D

Area of Science:

• Medical Imaging
• Computational Imaging
• Photon Counting Detectors

Background:

• Spectral CT scanners use photon counting detectors to generate projections across multiple energy channels.
• Reconstructed images from individual energy channels suffer from significant noise due to photon statistics.
• Strong correlations exist between inter-channel images and similarities within spatial patches of CT images.

Purpose of the Study:

• To develop an advanced image reconstruction method for spectral CT.
• To address the challenge of noise in spectral CT image reconstruction.
• To leverage inter-channel and spatial redundancies for improved image quality.

Main Methods:

• Proposed a novel average-image-incorporated block-matching and 3D (aiiBM3D) filtering method.
• Integrated aiiBM3D with low-rank regularization for iterative spectral CT reconstruction.
• Employed the alternating direction method of multipliers (ADMM) to exploit inter-channel correlations and applied BM3D to spatial redundancies.

Main Results:

• The proposed iterative reconstruction method effectively reduces noise in spectral CT images.
• Demonstrated significant improvements in signal-to-noise ratio (SNR).
• Showcased excellent preservation of structural details in reconstructed images.

Conclusions:

• The aiiBM3D filtering combined with low-rank regularization offers a powerful approach for spectral CT image reconstruction.
• Validated effectiveness on both simulated and preclinical datasets.
• Represents a significant advancement in improving the diagnostic quality of spectral CT imaging.