Performance of Ultrasound Liver Fat Quantification for the Assessment of Hepatic Steatosis Using MRI Fat Fraction as a Reference Standard

Abstract

Objective: To evaluate the performance and reproducibility of ultrasound-based liver fat quantification (LFQ) for the assessment of hepatic steatosis. Methods: This prospective study included adult patients who underwent quantification of hepatic steatosis with ultrasound and magnetic resonance imaging (MRI) on the same day. Ultrasounds were acquired with LFQ to measure attenuation of the liver parenchyma, hepatorenal index and 2-D shear-wave elastography, based on the median of 10 measurements taken by two operators. The reference standard was based on fat fraction quantification obtained using the mDixonQuant sequence on a 3T-MRI. Areas under the receiver operator curve and optimal cutoffs with the Youden index were calculated. Inter- and intra-observer reliability was assessed using the intraclass correlation coefficient. Results: A total of 152 patients (77 males, median age 64.0 years) were enrolled. On MRI, 39 (25.7%), 11 (7.2%) and 10 (6.6%) patients had grade 1, 2 and 3 steatosis, respectively. The median LFQ was 0.58 dB/cm/MHz, with a 100% success rate. Intra- and inter-operator reliability was good, with an intra-class correlation coefficient of 0.88 and 0.85, respectively. The areas under the receiver operator curve of LFQ were 0.919 (cutoff >0.60, sensitivity 80.3%, specificity 92.4%) for grade >= 1 steatosis, 0.970 (cutoff >0.64, sensitivity 95.2%, specificity 85.5%) for grade >= 2 steatosis and 0.974 (cutoff >0.69, sensitivity 100%, specificity 93.7%) for grade 3 steatosis. For all steatosis grades, LFQ performance was significantly higher than the hepatorenal index. Conclusion: Ultrasound-based LFQ provides excellent performance for the diagnosis of hepatic steatosis with high reproducibility.