Performance demonstration of a novel photon-counting CT for preclinical application

Photon-counting computed tomography (PC-CT) has attracted attention over the last few years as the next-generation CT technique that solves the problems encountered in clinical CT. In PC-CT, dark current and electronic noise can be reduced by setting the energy threshold to exceed the noise level, which leads to a low-dose scan. Furthermore, multiple energy thresholds realize multicolor CT imaging, which is not possible with clinical CT. Recently, we proposed a novel PC-CT system consisting of a multipixel photon counter (MPPC) coupled with a high-speed scintillator, performing simultaneous imaging of multiple contrast agents and estimate concentration. However, the PC-CT images obtained by our PC-CT system faces some limitations, such as degradation of image quality due to the lack of photon statistics and/or image resolution loss due to the pixel size of the detectors. In this study, the signal-to-noise ratio (SNR) of the PC-CT images was improved by applying machine-learning models, that is, U-Net and Noise2Noise, to the PC-CT images. In addition, a new imaging method was developed to acquire the high-resolution CT images required for clinical use. As a result, the resolution of the CT images improved from 1.04 mm to 0.77 mm. Finally, the visualization of contrast agents in plants was set as a challenge for the next step towards the clinical application of MPPC-based PC-CT. The results demonstrate that our PC-CT system can provide color imaging not only in phantom-based experiments, but also in plants close to an organism.