Statistical reconstruction of material decomposed data in spectral CT

Area of Science:

• Medical Imaging Physics
• Radiological Sciences
• Computational Imaging

Background:

• Photon-counting detector technology enables energy-resolved computed tomography (CT) and K-edge imaging.
• Limitations in detector count rates lead to high noise in spectral CT images.
• Iterative reconstruction (IR) offers superior signal-to-noise ratio (SNR) over filtered backprojection (FBP) in low-dose CT.

Purpose of the Study:

• To investigate statistically-principled iterative image reconstruction for material-decomposed sinograms in spectral CT.
• To develop a reconstruction algorithm that minimizes a penalized likelihood-based cost functional.
• To assess the performance of the proposed method for K-edge imaging applications.

Main Methods:

• Developed a novel iterative reconstruction algorithm minimizing a penalized likelihood cost functional.
• Estimated likelihood function parameters using the Fisher information matrix from material decomposition.
• Evaluated the method using computer-simulated, experimental phantom data, and an animal experiment.

Main Results:

• The proposed statistically-principled iterative reconstruction method effectively reduces noise in spectral CT images.
• The algorithm allows for a trade-off between noise and spatial resolution via a roughness penalty.
• Demonstrated improved K-edge imaging performance in phantom and animal studies.

Conclusions:

• Statistically-principled iterative reconstruction is a viable approach to address noise limitations in photon-counting spectral CT.
• The method shows significant potential for enhancing K-edge imaging accuracy and diagnostic utility.
• This work advances spectral CT reconstruction, paving the way for improved low-contrast detectability and material differentiation.