Purpose: The state-of-the-art photon-counting-detector (PCD) can distinguish various photon energes and correspondingly allocate each photon into its energy-related bin, which demonstrates a superior spectral resolution compared with the conventional energy-integrating detectors. However, when it is employed in micro-CT applications, the size limitation of the PCD cells, as a hardware bottleneck, directly decreases the spatial resolution in data acquisition, and further degrades the reconstructed image quality and material decomposition accuracy.
Methods: We innovatively design a multi-detector based spectral micro-CT system. A lens-coupled detector (LCD) performs optical magnification and contributes to a high spatial resolution scan. Simultaneously, a PCD works for energy-distinguishment and collects multi-channel data, i.e., a high spectral resolution scan. Then, we propose a novel hybrid data based multi-constraint optimization model for this system. It includes both intra- and inter-smoothness regularizations. The corresponding reconstruction and decomposition methods are also developed.
Results: We numerically simulate the scan process with an ideal point source, polychromatic photons, fan-beam geometry, and Poisson noise. We employ the Forbild head phantom and modify it by adding a resolution chart. The LCD contains a 10X optical magnification mechanism, and the PCD has two energy bins bisecting the flat-field photons. We visually validate the effectiveness and performance of the proposed approaches.
Conclusion: In this work, we develop a multi-detector based spectral micro-CT system with optical-magnification and energy-identification mechanisms. We incorporate both the scan datasets and the image-domain smoothness prior knowledge into a joint optimization objective for reconstructions and decompositions with high spatial and spectral resolution. Different from the conventional LCD-based micro-CT scanners, the inventive system design does not directly discard the unconverted X-ray photons but makes them collected by a PCD to achieve extra spectral information. The suggested reconstruction method well combines the hybrid datasets, exacts desired features (resolution merits), and applies supporting prior knowledge (smoothness characteristics).