Multi-energy computed tomography reconstruction using a nonlocal spectral similarity model

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiology

Background:

• Multi-energy computed tomography (MECT) offers material differentiation but suffers from image quality degradation due to limited photon counts per energy bin.
• Traditional CT reconstruction methods may not fully exploit the spectral information available in MECT data.

Purpose of the Study:

• To develop a novel statistical iterative reconstruction algorithm for high-quality MECT imaging.
• To improve the accuracy of material-specific image reconstruction from MECT data.

Main Methods:

• Developed the MECT-NSS algorithm, incorporating nonlocal self-similarities within energy bins and spectral similarities across energy bins.
• Implemented an efficient optimization strategy to solve the MECT-NSS objective function.
• Conducted comprehensive parameter selection evaluations.

Main Results:

• The MECT-NSS algorithm demonstrated superior MECT image quality compared to existing methods.
• MECT-NSS achieved more accurate material estimation.
• Validation performed on phantom and patient datasets confirmed the algorithm's effectiveness.

Conclusions:

• The MECT-NSS algorithm effectively addresses MECT image quality limitations.
• This approach provides enhanced material differentiation and quantification capabilities for MECT.
• MECT-NSS represents a significant advancement in MECT image reconstruction technology.