Liver fat quantification using deep silicon photon-counting CT: an in silico imaging study.
BACKGROUND: Accurate liver fat quantification is essential for early diagnosis and effective management of fatty liver disease. PURPOSE: To investigate the potential clinical utility of a deep silicon-based photon-counting CT (dSi-PCCT), currently in development, for liver fat quantification using human models in an in silico imaging study. MATERIALS AND METHODS: dSi-PCCT is a cutting-edge photon-counting CT (GE HealthCare), with several investigational systems installed globally, used under IRB approval for imaging animals and human volunteers to support FDA clearance. We developed a dSi-PCCT simulator and benchmarked its imaging performance with respect to a prototype. We imaged a computational Gammex phantom with fat fractions (FF) ranging from 0% to 100%, along with five XCAT human models with liver FF ranging from 1% to 50%, using an abdominal CT protocol. The resulting spectral sinograms were processed using a material decomposition (MD) technique. We calculated HU-based Proton Density Fat Fraction (PDFF) from single-energy images in XCAT models and compared it against the MD-derived FF. The MD-derived FF of both datasets was assessed against the digitally defined ground truth values. RESULTS: We observed a strong correlation (R 2 = 0.98) between MD-derived, HU-based PDFF, and ground-truth FF in a Gammex and XCAT models. There was no statistically significant difference (P = .52) in FF quantification accuracy between Gammex and the XCAT human models. The root mean square errors were 4.7% for Gammex and 2.7% for XCAT. Bland-Altman analysis further confirmed good agreement between the ground truth and MD-derived FF, with differences in FF ranging from -6.9% to 7% for Gammex and -3.0% to 37.6% for XCAT. CONCLUSION: The results indicate that dSi-PCCT could enable accurate liver fat quantification across a wide range of FFs in multiple objects. These findings suggest that the potential utility of dSi-PCCT for accurate liver fat assessment should be explored in vivo.
