Agreement of VNC and Spectral Photon-Counting CT in the Assessment of Normal to Minimal Hepatic Fat in Comparison with MRI

RATIONALE AND OBJECTIVES: To evaluate the agreement of different liver fat quantification methods using Photon-Counting-Detector (PCD)-CT via virtual non-contrast (VNC) reconstructions and dedicated spectral post-processing algorithms, with MRI-derived fat fractions (FF) serving as the reference standard. MATERIALS AND METHODS: This retrospective study included consecutive patients who underwent abdominal PCD-CT in arterial (art) and portal-venous (pv) phases and abdominal multiparametric MRI with chemical shift imaging between February 2022 and May 2024. Patients were excluded if no arterial phase was available or if chronic or acute liver disease was present. Liver FF was quantified using two methods in both contrast phases: via Hounsfield units from VNC images using an established formula (HU-FF), and directly using a spectral post-processing algorithm (SPP-FF). These were compared among themselves and to MRI-FF using Pearson's correlation and mean absolute error (MAE). RESULTS: The final cohort included 42 patients (mean age, 66.9 ± 11.4 years; 50% male). Median MRI-FF was 1.49% (Interquartile Range [IQR], 0.72%-4.57%). HU-FF and SPP-FF showed moderate correlation with MRI-FF (r = 0.40-0.71, P < .05). Correlation between HU-FF and SPP-FF was high (r = 0.72-0.80), with excellent intra-method agreement (HU-FF: r = 0.94; SPP-FF: r = 0.93). MAE was lowest for SPP-FF-art (3.30%) and highest for HU-FF-pv (4.82%). CONCLUSION: PCD-CT offers several methods for liver fat quantification, yet, in a cohort with low liver fat, agreement with MRI varied considerably.