Multi-Material Decomposition for Single Energy CT Using Material Sparsity Constraint.

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiology

Background:

• Multi-material decomposition (MMD) is crucial for identifying human body compositions in CT images.
• Spectral CT enables MMD but is not widely available clinically.
• Existing single energy CT (SECT) MMD methods suffer from limited accuracy and material basis due to data constraints.

Purpose of the Study:

• To develop an image-domain SECT MMD method utilizing material sparsity.
• To enhance decomposition accuracy for SECT-based MMD in clinical settings.
• To address the limitations of current SECT MMD techniques.

Main Methods:

• An image-domain SECT MMD method incorporating a material sparsity constraint (MSC) was proposed.
• The L0 norm was used to represent the MSC within the decomposition objective function.
• An accelerated primal-dual (APD) algorithm with a line-search scheme was employed for optimization.
• Initial estimates were derived using a pixelwise direct inversion method with the two-material assumption (TMA).

Main Results:

• The proposed MSC method achieved a volume fraction accuracy (VFA) of 98.5% in phantom studies, a significant improvement over the TMA method's 92.0%.
• In patient studies, the method clearly visualized calcification areas in virtual non-contrast images, matching ground-truth CT images.
• High decomposition image quality was demonstrated, facilitating clinical applications of SECT-based MMD.

Conclusions:

• The proposed SECT MMD method with material sparsity significantly improves decomposition accuracy.
• This technique enhances the clinical utility of SECT for material composition analysis.
• The method shows promise for broader adoption in radiological departments lacking spectral CT.