Accuracy of ultrasonography in the assessment of liver fat compared with MRI

Abstract

To investigate the accuracy of ultrasonography in the assessment of hepatic steatosis using magnetic resonance imaging (MRI) as standard of reference and to explore the influence of additional hepatic iron overload. A total of 2,783 volunteers (1,442 women, 1,341 men; mean age, 52.3±13.8 years) underwent confounder-corrected chemical-shift-encoded MRI of the liver at 1.5 T. Proton-density fat fraction (PDFF) and transverse relaxation rate (R2*) were calculated to estimate hepatic steatosis and liver iron overload, respectively. In addition, the presence of hepatic steatosis was assessed by B-mode ultrasonography. The sensitivity, specificity, and accuracy of hepatic ultrasonography were determined for different degrees of hepatic steatosis and different amounts of liver iron. MRI revealed hepatic steatosis in 40% of participants (n=1,112), which was mild in 68.9% (n=766), moderate in 26.7% (n=297), and severe in 4.4% (n=49) of patients. Ultrasonography detected hepatic steatosis in 37.8% (n=1,052), corresponding to 74.5% sensitivity and 86.6% specificity. The sensitivity of ultrasound increased with the amount of hepatic fat present and was 65.1%, 95%, and 96% for low, moderate, and high fat content; whereas the specificity was constantly high at 86.6%. The diagnostic accuracy of ultrasound for detection of hepatic steatosis did not vary significantly with the amount of liver iron present. Ultrasonography is an excellent tool to assess hepatic steatosis in the clinical setting with some limitations in patients with a low liver fat content. The detection of hepatic steatosis by ultrasonography is not influenced by liver iron.