Robust Liver Steatosis Quantification Using Photon-Counting CT in Comparison to PDFF: Validation across Protocols and Cohorts.

RATIONALE AND OBJECTIVES: To evaluate the robustness and grading performance of Photon-counting CT (PCCT)-derived fat fraction (CTFF) in liver fat quantification across different protocols and cohorts with varying liver conditions, using proton density fat fraction (PDFF) as the reference. MATERIALS AND METHODS: The prospective study enrolled 297 participants in the Volunteer cohort. Participants were randomly assigned to 120-kVp or 140-kVp groups and subdivided into high- and low-dose subgroups for technical validation. Clinical validation was performed in two cohorts: participants with metabolic dysfunction-associated steatotic liver disease (MASLD, n = 86) and heterogeneous liver diseases (HLD, n = 61). PCCT and PDFF (reference standard for grading) were performed within 1 month. Statistical analyses included intraclass correlation coefficients, Bland-Altman analysis, Mann-Whitney U and Kruskal-Wallis tests, and areas under the receiver operating characteristic (AUC-ROC) and precision-recall curves (AUC-PR). RESULTS: Overall, 444 participants were included (age 44.9 ± 14.3 years; 264 men). In the Volunteer cohort, CTFF showed excellent agreement with PDFF (all ICCs = 0.98; R2 ≥0.96) with a maximum bias of 0.80% (95% CI: 0.70-0.89%), across the whole-cohort and subgroups. No significant differences in the absolute error between PDFF and CTFF were observed across all subgroups (all p > 0.05). CTFF achieved high diagnostic accuracy for both the detection and grading of steatosis with all AUC-ROC and AUC-PR exceeding 0.99. The diagnostic accuracy of CTFF was verified in the MASLD and HLD cohorts. CONCLUSION: CTFF demonstrated excellent agreement with PDFF and consistent diagnostic performance across diverse scan settings and liver conditions.