Penalized weighted least-squares approach for multienergy computed tomography image reconstruction via structure...

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computed Tomography

Background:

• Multienergy computed tomography (MECT) enables material differentiation using energy-selective data.
• Conventional MECT reconstruction methods suffer from low signal-to-noise ratios and streak artifacts due to insufficient photon counts in specific energy windows.

Purpose of the Study:

• To develop an advanced image reconstruction algorithm for MECT that overcomes the limitations of existing methods.
• To improve the quantitative and visual quality of MECT images.

Main Methods:

• A penalized weighted least-squares (PWLS) scheme was developed, incorporating structure tensor total variation (STV) regularization (PWLS-STV).
• STV regularization penalizes higher-order derivatives, offering robust image variation measures and mitigating patchy artifacts common in total variation (TV) regularization.
• An alternating optimization algorithm was employed to minimize the objective function.

Main Results:

• The PWLS-STV algorithm demonstrated superior performance compared to conventional filtered backprojection (FBP) and existing TV-based algorithms.
• Quantitative and visual evaluations confirmed significant gains in image quality, including reduced noise and artifacts.
• Experiments were conducted using a digital XCAT phantom and a meat specimen.

Conclusions:

• The proposed PWLS-STV algorithm effectively enhances MECT image quality.
• This method offers a significant improvement over traditional and current regularization techniques for MECT.
• PWLS-STV holds promise for more accurate material identification and differentiation in MECT applications.