Comparing ultrasound-derived fat fraction and CT for diagnosing hepatic steatosis: an MRI-PDFF reference study

This study directly compares the diagnostic performance of two routinely available, non-invasive techniques-ultrasound-derived fat fraction (UDFF) and non-contrast computed tomography (NCCT)-for screening and grading hepatic steatosis (HS). In addition, it exploratorily assesses the utility of liver stiffness measurement (LSM) for quantifying liver fat content.

In this prospective single-center study, 78 patients with chronic liver disease (CLD) were recruited and underwent NCCT, ultrasound [UDFF and Automatic Point Shear Wave Elastography (Auto-pSWE)], and magnetic resonance imaging-proton density fat fraction (MRI-PDFF) within 1 week. The same cohort was analyzed using two different MRI-PDFF threshold criteria: the primary set and the alternative set. In the primary set, Spearman's correlation coefficients assessed relationships among UDFF, CT liver-to-spleen attenuation ratio [CT(L/S)], LSM, and MRI-PDFF. The diagnostic performance of UDFF and CT(L/S) for grading HS was evaluated using ROC curves with bootstrap AUC comparisons. Agreement and reliability were assessed using ICC and kappa statistics, and influencing factors were examined by logistic regression. In the alternative set, the diagnostic performance of UDFF and CT(L/S) for grading steatosis was similarly evaluated using ROC curves with bootstrap AUC comparisons.

After excluding three patients, 75 were included in the final analysis. In the primary set, for HS ≥ S1, UDFF (AUC = 0.926; cutoff 8.5%) outperformed CT(L/S) (AUC = 0.826; cutoff 0.95; p < 0.05). For HS ≥ S2, both methods performed similarly [UDFF AUC = 0.961; CT(L/S) AUC = 0.943; p > 0.05]. Grading diagnosis agreement was good between UDFF and MRI-PDFF (κ = 0.740), but only moderate for CT(L/S) (κ = 0.517). UDFF demonstrated higher inter-operator (ICC = 0.949) and intra-operator (ICC = 0.967) reliability than CT(L/S). Triglyceride level was the determinant factor for UDFF (OR = 8.510, p < 0.05). In the Alternative Set, UDFF also outperformed CT(L/S) for HS ≥ S1 (p < 0.05), with no significant difference for HS ≥ S2 (p > 0.05).

UDFF outperforms CT(L/S) in steatosis grading and reliability but is limited to the Siemens DAX probe, whereas CT(L/S) is equipment-independent but less reliable.