Development and Evaluation of a Multiparametric Ultrasound Approach for Quantifying Hepatic Steatosis

Abstract

Objective

Accessible and noninvasive methods are crucial for the detection and monitoring of steatosis. This study assessed the performance of ultrasound via three ultrasound markers—the attenuation coefficient, the backscattering coefficient and speckle statistics—modeled by a Nakagami distribution, using a dual intercostal and subcostal approach. Combining these three ultrasound markers obtained from the inter- and subcostal acquisitions generated the ultrasound-based fat fraction (UFF). The PDFF measured with MRI was used as the gold standard.

Materials and methods

One hundred thirty-three patients were included in this prospective, bicentric study. For both the intercostal and subcostal approaches, the three ultrasound markers were compared with the PDFF and their coefficients of determination were estimated by means of a second-order polynomial fit. The correlation between the UFF and PDFF was assessed using the Pearson correlation coefficient. ROC curves were used to study the performance of UFF.

Results

Independently, all three ultrasound markers were well correlated with the PDFF: attenuation coefficient r² = 0.72 (intercostal) and r² = 0.65 (subcostal), backscattering coefficient r² = 0.68 (intercostal) and r² = 0.69 (subcostal), and Nakagami m parameter r² = 0.51 (intercostal) and r² = 0.52 (subcostal). The UFF score was strongly correlated with the PDFF (r² = 0.79). The areas under the ROC curves (AUROCs) were 0.94 to distinguish between grade S0 and grades ≥ S1 and 0.98 to distinguish between ≤ S1 and ≥ S2.

Conclusion

UFF is strongly correlated with PDFF for steatosis quantification. Combining both intercostal and subcostal measurements improves this correlation with PDFF.