Sparse-View Spectral CT Reconstruction and Material Decomposition Based on Multi-Channel SGM.

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Spectral CT quantifies K-edge contrast agents, crucial for assessing physiological function.
• High radiation doses from repetitive spectral CT scans are a significant concern.
• Sparse-view scanning reduces dose but introduces artifacts, hindering accurate quantification.

Purpose of the Study:

• To develop an unsupervised algorithm for sparse-view spectral CT reconstruction and material decomposition.
• To enable accurate quantification of contrast agent distribution and content with reduced radiation exposure.
• To address the challenge of streaking artifacts in sparse-view spectral CT.

Main Methods:

• Utilized a multi-channel score-based generative model (SGM) trained on multi-energy and tissue images.
• Employed sparse-view projections to drive SGM for generating multi-energy and tissue images.
• Integrated SGM-generated tissue images as priors into a material decomposition algorithm for contrast agent imaging.

Main Results:

• The proposed SGM-based method successfully reconstructed images from sparse-view spectral CT data.
• Accurate quantification of contrast agent distribution and content was achieved, overcoming artifacts.
• Experimental validation in mouse models demonstrated the method's effectiveness.

Conclusions:

• The unsupervised SGM approach offers a promising solution for dose reduction in spectral CT.
• Accurate contrast agent quantification is feasible even with sparse-view acquisition.
• This technique has significant potential for advancing medical imaging and diagnostics.