Statistical image-domain multimaterial decomposition for dual-energy CT.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Dual-energy CT (DECT) offers advanced tissue characterization through basis material decomposition.
• Clinical applications often require multi-material decomposition (MMD), which is an ill-posed problem with conventional DECT methods.
• Existing flexible image-domain MMD methods suffer from noise magnification due to direct inversion schemes.

Purpose of the Study:

• To develop a statistical image-domain MMD method for DECT to suppress noise.
• To improve the accuracy and preserve anatomical structures in decomposed material images.
• To address the limitations of existing MMD techniques in dual-energy CT.

Main Methods:

• Implemented a penalized weighted least-square (PWLS) reconstruction with negative log-likelihood and edge-preserving regularization for each material.
• Determined statistical weights using a data-based method accounting for noise variance in high- and low-energy CT images.
• Utilized optimization transfer principles to create pixel-wise separable quadratic surrogates (PWSQS) for simultaneous pixel updates.

Main Results:

• The proposed method reduced noise standard deviation in soft tissue by up to 95.35% compared to direct inversion.
• Achieved improved volume fraction accuracy by approximately 6.85% and reduced RMSE of electron densities by 20.89% compared to low-pass filtration.
• Increased spatial resolution by factors of 1.64 (digital phantom) and 2.16 (Catphan©600 phantom) at 50% MTF magnitude.

Conclusions:

• A novel statistical image-domain MMD method effectively suppresses noise in DECT.
• The method maintains quantification accuracy and anatomical structure fidelity in decomposed material images.
• This technique shows promise for practical and advanced clinical applications of DECT imaging.