Spectral diffusion: an algorithm for robust material decomposition of spectral CT data.

Area of Science:

• Medical Imaging
• Computational Imaging
• Materials Science

Background:

• Spectral CT is emerging as a powerful functional imaging tool, driven by advancements in detector technology and nanoparticle contrast agents.
• Robust material decomposition algorithms are crucial for spectral CT, especially in noisy conditions, to accurately quantify material concentrations.

Purpose of the Study:

• To develop a noise-robust material decomposition algorithm for spectral CT.
• To improve quantitative accuracy and reduce bias in material decomposition for spectral CT applications.

Main Methods:

• Developed 'spectral diffusion,' an algorithm combining spectrally joint kernel regression and split Bregman iteration.
• The algorithm enforces gradient sparsity for accurate and minimally biased material decomposition.
• Utilized a 3D digital bar phantom and a calibrated sensitivity matrix for algorithm evaluation.

Main Results:

• Quantified the limits of detectability for iodine, gold, and gadolinium using spectral diffusion.
• Demonstrated accurate triple-energy material decomposition with concentrations as low as 0.9 mg/mL for gold.
• Successfully applied spectral diffusion for in vivo material separation in mouse organs (kidneys, liver, spleen).

Conclusions:

• Spectral diffusion offers a robust solution for material decomposition in spectral CT, enhancing its functional imaging capabilities.
• The algorithm's accuracy and noise resilience are critical for clinical translation and research involving nanoparticle contrast agents.