Multi-material spectral photon-counting micro-CT with minimum residual decomposition and self-supervised deep...

Area of Science:

• Medical Imaging
• Materials Science
• Physics

Background:

• Spectral micro-CT imaging faces challenges with energy resolution and noise amplification.
• Small pixel size detectors (< 100x100 µm²) exacerbate charge sharing and noise issues.

Purpose of the Study:

• To develop a cone-beam micro-CT setup addressing spectral imaging limitations.
• To enhance material discrimination and image quality in spectral micro-CT.

Main Methods:

• Utilized a CdTe photon counting detector with hardware charge summing.
• Implemented an image processing pipeline with spectral response modeling and minimum-residual basis material decomposition (MR-BMD).
• Applied self-supervised deep convolutional denoising to acquired projections (45x45 µm² pixel size).

Main Results:

• Successfully discriminated between materials with K-edges separated by only a few keV (e.g., Iodine and Barium).
• Demonstrated sharp discrimination capabilities using combined hardware and software solutions.
• Evaluated quantitative performance of reconstructed decomposed images (water, bone, I, Ba, Gd).

Conclusions:

• The developed spectral micro-CT system effectively overcomes hardware and software limitations.
• The innovative approach significantly improves material differentiation and image quality in spectral micro-CT.
• This work provides a robust platform for quantitative spectral micro-CT analysis.