Diagnostic Accuracy of Quantitative Ultrasound Liver Fat Estimation with MRI-PDFF as the Reference Standard

To assess the performance of 3 ultrasound-based liver fat quantification algorithms-tissue attenuation imaging (TAI), tissue scatter imaging (TSI; Nakagami parameter), and ultrasound-derived fat fraction (USFF)-using magnetic resonance imaging-derived proton density fat fraction (MRI-PDFF) as the reference standard.

In this prospective, IRB-approved, HIPAA-compliant study conducted at 2 North American centers, adults with or at risk for hepatic steatosis underwent quantitative ultrasound and same-day MRI-PDFF. The US examinations were performed with the Samsung RS85 system (Samsung Medison, Seoul, South Korea) using a C1-5S transducer. The software version utilized was 2.08.01.3538. Correlations were assessed using Spearman's rho. Diagnostic performance for steatosis grades S > 0, S > 1, and S > 2 was evaluated using AUROC analysis. Overall accuracy across grades was also assessed with the Obuchowski measure. Correlations and AUROCs were compared using Steiger's and DeLong's tests, respectively.

Among 181 participants (113 women; median age, 57 years), MRI-PDFF correlated strongly with TAI and USFF (rho, 0.77 and 0.79) and moderately with TSI (rho, 0.57). For S > 0, AUROCs were 0.92 for TAI, 0.78 for TSI, and 0.93 for USFF; for S > 1, 0.88, 0.77, and 0.89; and for S > 2, 0.81, 0.73, and 0.83, respectively. Obuchowski measures were 0.87, 0.76, and 0.88. TAI and USFF were significantly more accurate than TSI (p = .03 and p = .02), with no significant difference between them (p = .79).

TAI and USFF showed high, comparable performance for detecting hepatic steatosis and outperformed TSI. USFF was numerically superior to TAI, without significant incremental benefit. Discrimination between moderate and severe steatosis remained limited in the US evaluation.