Feasibility study of photon-counting CT for material identification based on YSO/SiPM detector: A proof of concept
Du Zhang,Bin Wu,Daoming Xi,Rui Chen,Peng Xiao,Qingguo Xie
DOI: https://doi.org/10.1002/mp.17341
2024-08-12
Abstract:Background: Current photon-counting computed tomography (CT) systems utilize semiconductor detectors, such as cadmium telluride (CdTe), cadmium zinc telluride (CZT), and silicon (Si), which convert x-ray photons directly into charge pulses. An alternative approach is indirect detection, which involves Yttrium Orthosilicate (YSO) scintillators coupled with silicon photomultipliers (SiPMs). This presents an attractive and cost-effective option due to its low cost, high detection efficiency, low dark count rate, and high sensor gain. Objective: This study aims to establish a comprehensive quantitative imaging framework for three-energy-bin proof-of-concept photon-counting CT based on YSO/SiPM detectors developed in our group using multi-voltage threshold (MVT) digitizers and assess the feasibility of this spectral CT for material identification. Methods: We developed a proof-of-concept YSO/SiPM-based benchtop spectral CT system and established a pipeline for three-energy-bin photon-counting CT projection-domain processing. The empirical A-table method was employed for basis material decomposition, and the quantitative imaging performance of the spectral CT system was assessed. This evaluation included the synthesis errors of virtual monoenergetic images, electron density images, effective atomic number images, and linear attenuation coefficient curves. The validity of employing A-table methods for material identification in three-energy-bin spectral CT was confirmed through both simulations and experimental studies. Results: In both noise-free and noisy simulations, the thickness estimation experiments and quantitative imaging results demonstrated high accuracy. In the thickness estimation experiment using the practical spectral CT system, the mean absolute error for the estimated thickness of the decomposed Al basis material was 0.014 ± 0.010 mm, with a mean relative error of 0.66% ± 0.42%. Similarly, for the decomposed polymethyl methacrylate (PMMA) basis material, the mean absolute error in thickness estimation was 0.064 ± 0.058 mm, with a mean relative error of 0.70% ± 0.38%. Additionally, employing the equivalent thickness of the basis material allowed for accurate synthesis of 70 keV virtual monoenergetic images (relative error 1.85% ± 1.26%), electron density (relative error 1.81% ± 0.97%), and effective atomic number (relative error 2.64% ± 1.26%) of the tested materials. In addition, the average synthesis error of the linear attenuation coefficient curves in the energy range from 40 to 150 keV was 1.89% ± 1.07%. Conclusions: Both simulation and experimental results demonstrate the accurate generation of 70 keV virtual monoenergetic images, electron density, and effective atomic number images using the A-table method. Quantitative imaging results indicate that the YSO/SiPM-based photon-counting detector is capable of accurately reconstructing virtual monoenergetic images, electron density images, effective atomic number images, and linear attenuation coefficient curves, thereby achieving precise material identification.