Objectives: To assess the inter-observer agreement of imaging features of LI-RADS using contrast-enhanced CT (CECT) and gadoxetic-acid enhanced MR (GAeMR) between expert and novice radiologists
Materials and Methods: From January 2013 to December 2015, sixty-three patients (M:F=54:9, mean age: 55.8±9.5 years) which were confirmed with HCC pathologically and undergoing CECT and GAeMR were included in this study. Two abdominal special radiologists analyzed CT and MR images with consensus and other two novice radiologists analyzed CT and MR images with consensus. The size of the tumor, which is an important diagnostic criterion in the LI-RADS lexicon, was compared to the pathologically measured size to assess the measurement error in each imaging modality and in each phase. We also evaluated inter-observer agreement between the readers of arterial hyperenhancement, washout, and capsule appearance, which are the major imaging findings of LI-RADS in HCC. In addition, inter-observer agreement of final LI-RADS categorization was evaluated.
Results: A total of 63 patients were included in the study. When comparing the difference between the measured value of each tumor and the size of pathologically measured tumor as a reference value, it was confirmed that the measurement error was smaller than that of CT by MRI. However, there was no statistical difference in the measurement error between expert radiologists and novice radiologists (p=0.29, p=0.62). When we assessed the inter-observer agreement of the HCC imaging features of LI-RADS, the arterial hyperenhancement was moderate in both CT and MR (к=0.566, 0.475). Inter-observer agreement on washout appearance was fair in CT (к= 0.364) and substantial in MR (к=0.630). In the case of capsule appearance, inter-observer agreement was moderate in CT (к=0.491) and fair in MR (к=0.303). There was no significant difference in the proportion of the LI-RADS category between experts and novices. When evaluating the value of intraclass correlation coefficients (ICC), it was found that the inter-reader agreement of MR was lower than that of CT (ICC = 0.574 in CT, 0.318 in MR).
Conclusion: MR was slightly more accurate than CT in measuring the size of HCC, especially for beginners. Inter-reader agreement was higher in CT than MR in arterial hyperenhancement and capsule appearance, whereas in terms of washout appearance, that was higher in MR than in CT. LI-RADS showed fair to moderate agreement for diagnosis of HCC on CT and GAeMR.
