The interreader agreement and validation of contrast-enhanced ultrasound liver imaging reporting and data system

Abstract

PurposeThis study explored the interreader agreement and diagnostic performance of contrast-enhanced ultrasound liver imaging reporting and data system (CEUS LI-RADS). MethodsBetween January 2014 and December 2017, 1366 patients at risk for hepatocellular carcinoma (HCC) who underwent CEUS were included in this retrospective study. Four ultrasound physicians rated the HCC likelihood of focal liver lesions (FLLs) using CEUS LI-RADS v2017. Interreader agreement on CEUS LI-RADS categories and major features (arterial phase hyperenhancement (APHE), washout appearance) were assessed using weighted kappa statistics (κ). Diagnostic performance was described by sensitivity, specificity, PPV and NPV, +LR, -LR. ResultsThe interreader agreement (κ) for CEUS LI-RADS categories, APHE, and washout appearance ranged from 0.61 to 0.73, 0.65 to 0.83, and 0.58 to 0.71, respectively. Interreader agreement for LI-RADS categories and APHE were almost substantial between FLLs <2 cm, ≥2 cm, <5 cm; interreader agreement for major features were fair to substantial for FLLs ≥5 cm. The accuracy, PPV and + LR for HCC and malignancy in FLLs <2 cm, ≥2 cm, <5 cm, ≥5 cm were high, with values of 84.7% to 91.9%, 90.2% to 94.2%, and 2.2 to 8.0, respectively. CEUS LI-RADS had the highest specificity for HCC (90.2%) and malignancy (90.9%) diagnosis for FLLs <2 cm and <5 cm, respectively; specificity was lowest for HCC (54.7%) and malignancy (68.3%) diagnosis for FLLs ≥5 cm. ConclusionsCEUS LI-RADS is a good standardized categorization system for high-risk patients, and the combination of two or three LR-M features may improve the true-negative classification of HCC diagnosis.