Liver Imaging Reporting and Data System: Discordance Between Computed Tomography and Gadoxetate-Enhanced Magnetic Resonance Imaging for Detection of Hepatocellular Carcinoma Major Features

Abstract

Purpose: The goal of this study was to compare agreement between computed tomography (CT) and magnetic resonance imaging (MRI) in the evaluation of the major Liver Imaging Reporting and Data System (LI-RADS) features used in assessment of hepatocellular carcinoma: arterial phase hyperenhancement (APHE), portal venous phase washout (WO), capsule appearance (capsule), and largest diameter (diameter).

Methods: Patients with liver protocol CT and gadoxetate-enhanced MRI within 1 month of each other and at least 1 discrete untreated liver lesion were included. Two readers independently reviewed hepatic arterial phase and portal venous phase of each lesion on both CT and MRI, presented at random. The APHE, WO, capsule, and diameter were assessed for each lesion on CT and MRI. The LI-RADS category was assigned based on the recorded major features. Interobserver agreements between the readers for both imaging modalities and for each of the major features were assessed using κ statistics. Agreement between CT and MRI for each reader and for each feature was assessed using κ statistics. Agreement was interpreted based on κ as follows: 0.20 or less, slight agreement; 0.21 to 0.40, fair agreement; 0.41 to 0.60, moderate agreement; 0.61 to 0.80, substantial agreement; and 0.81 to 1.00, almost perfect agreement. Intraclass correlation coefficient was used to assess concordance of diameter measurements.

Results: There were 42 patients (mean age, 62.2 ± 7.0 years; 33 men [78.6%]) with 50 lesions. On CT, the interobserver agreement between the readers was almost perfect for APHE (κ = 0.85), WO (κ = 0.83), and capsule (κ = 0.86). On MRI, the interobserver agreement between the readers was almost perfect for APHE (κ = 0.86) and WO (κ = 0.83) and moderate for capsule (κ = 0.59). Intraclass correlation coefficient for diameter measurement was 0.99 for CT and 0.98 for MRI. For reader 1, the agreement between CT and MRI was fair for APHE (κ = 0.39) and capsule (κ = 0.26) and moderate for WO (κ = 0.49). For reader 2, the agreement between CT and MRI was moderate for APHE (κ = 0.43) and capsule (κ = 0.43) and fair (κ = 0.38) for WO. Agreement between readers for final LI-RADS category was substantial for CT (κ = 0.79) and moderate for MRI (κ = 0.60). Agreement for final LI-RADS categories between MRI and CT was fair for both reader 1 (κ = 0.33) and reader 2 (κ = 0.39).

Conclusions: Interobserver agreement for the major LI-RADS features varies from moderate to almost perfect, for both CT and MRI. However, the agreement between CT and MRI for each of the major LI-RADS features is poor, ranging from fair to moderate. This poor agreement contributes to substantial differences between final LI-RADS category assigned on CT versus MRI.