Multi-material spectral photon-counting micro-CT with minimum residual decomposition and self-supervised deep denoising
Abstract
Spectral micro-CT imaging with direct-detection energy discriminating photon counting detectors having small pixel size (< 100×100 µm2) is mainly hampered by: i) the limited energy resolution of the imaging device due to charge sharing effects and ii) the unavoidable noise amplification in the images resulting from basis material decomposition. In this work, we present a cone-beam micro-CT setup that includes a CdTe photon counting detector implementing a charge summing hardware solution to correct for the charge-sharing issue and an innovative image processing pipeline based on accurate modeling of the spectral response of the imaging system, an improved basis material decomposition (BMD) algorithm named minimum-residual BMD (MR-BMD), and self-supervised deep convolutional denoising. Experimental tomographic projections having a pixel size of 45×45 µm2 of a plastinated mouse sample including I, Ba, and Gd small cuvettes were acquired. Results demonstrate the capability of the combined hardware and software tools to sharply discriminate even between materials having their K-Edge separated by a few keV, such as e.g., I and Ba. By evaluating the quality of the reconstructed decomposed images (water, bone, I, Ba, and Gd), the quantitative performances of the spectral system are here assessed and discussed.
Read the paper
Related studies
- Developing high safety Li-metal anodes for future high-energy Li-metal batteries: strategies and perspectives
- Structure and Energy Transfer in Photosystems of Oxygenic Photosynthesis
- Ground- and Space-Based Gamma-Ray Astronomy
- Energy Functions in De Novo Protein Design: Current Challenges and Future Prospects
- Electronic structure of ions and molecules in solution: a view from modern soft X-ray spectroscopies
- Arising 2D Perovskites for Ionizing Radiation Detection
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.