Abstract
<jats:p>Background. Metabolic dysfunction-associated steatotic liver disease (MASLD) is highly prevalent among patients with obesity and type 2 diabetes, yet hepatic fat distribution is spatially heterogeneous. This heterogeneity may influence the performance of noninvasive diagnostic tools, including ultrasound attenuation coefficient measurement (ACM), when benchmarked against magnetic resonance imaging-proton density fat fraction (MRI-PDFF). The purpose of the study was to quantify segmental heterogeneity of hepatic steatosis by MRI-PDFF and evaluate its impact on the diagnostic accuracy of ACM in MASLD. Materials and methods. A prospective cohort of 116 adults (42.2 % men; mean age 48.2 ± 13.6 years) underwent same-episode liver MRI-PDFF and ultrasound ACM. MRI-PDFF was quantified in all eight Couinaud segments to compute whole-liver PDFF (WL-PDFF) and intersegmental heterogeneity metrics. ACM values (dB/cm) were obtained under standardized quality criteria. Correlation, agreement (intraclass correlation coefficient, Bland-Altman), and ROC analysis were performed for clinically relevant WL-PDFF thresholds (≥ 5, ≥ 10, ≥ 20 %). Results. The mean WL-PDFF was 11.0 ± 6.4 %, while mean segmental PDFF was 11.3 ± 6.3 %. Intersegmental correlations with WL-PDFF were strongest in segments VIII (ρ = 0.99) and VII (ρ = 0.98) and lowest in segment I (ρ = 0.82). ACM correlated significantly with WL-PDFF (ρ = 0.63, p < 0.001), with the highest segmental correlation in segment VI (ρ = 0.68). ROC analysis showed excellent diagnostic performance: AUC = 0.809, 0.907, and 0.956 for PDFF ≥ 5, ≥ 10, and ≥ 20 %, respectively. Conclusions. Hepatic steatosis in MASLD exhibits clinically relevant segmental heterogeneity. ACM provides strong correlation with MRI-PDFF and excellent diagnostic performance for clinically significant fat thresholds, but localized variability may weaken concordance. Multiple ROI sampling, particularly in posterior-lateral right-lobe segments, enhances agreement with whole-liver PDFF and should be incorporated into standardized protocols.</jats:p>