Liver fat quantification using deep silicon photon-counting CT: an in silico imaging study.

Area of Science:

• Medical Imaging
• Radiology
• Computational Phantoms

Background:

• Accurate liver fat quantification is crucial for diagnosing and managing fatty liver disease.
• Non-alcoholic fatty liver disease (NAFLD) is a growing global health concern requiring precise diagnostic tools.

Purpose of the Study:

• To evaluate the clinical utility of a novel deep silicon-based photon-counting CT (dSi-PCCT) for liver fat quantification.
• To assess dSi-PCCT's accuracy using in silico human models.

Main Methods:

• A dSi-PCCT simulator was developed and benchmarked.
• Computational Gammex and XCAT human phantoms with varying fat fractions were imaged.
• Material decomposition (MD) techniques were applied to spectral sinograms for fat fraction (FF) calculation.

Main Results:

• A strong correlation (R ² = 0.98) was observed between MD-derived FF, HU-based PDFF, and ground truth.
• No significant difference in FF quantification accuracy was found between phantom types (P = 0.52).
• Low root mean square errors (2.7% for XCAT, 4.7% for Gammex) indicate high accuracy.

Conclusions:

• dSi-PCCT enables accurate liver fat quantification across a wide range of fat fractions.
• The findings support further in vivo investigation of dSi-PCCT for liver fat assessment.