Objective:To explore the differences of the diagnostic performance between the most recent 2018 version of liver imaging reporting and data system (LI-RADS v2018) and 2017 version (LI-RADS v2017) based on gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) enhanced MRI to diagnose hepatocellular carcinoma (HCC) in high-risk patients.Methods:The clinical data and imaging findings of 237 patients in high-risk of HCC who underwent Gd-EOB-DTPA enhanced MRI and obtained postoperative or biopsy pathological results within one month from June 2016 to December 2019 in Tianjin Third Central Hospital were collected retrospectively. A total of 282 observations were obtained as study objects. Two independent radiologists blindly reviewed the preoperative MRI of all patients. The observations were categorized according to LI-RADS v2018 and v2017 respectively. The inter-observer agreement of the categorization between the two radiologists was tested by kappa analysis. With the LR-5 and LR-4+5 as the diagnosis of HCC, the sensitivity, specificity, accuracy, and Youden index of the LI-RADS v2017 and LI-RADS v2018 were evaluated with postoperative histopathological results as references. The McNemar test was used to compare the diagnostic performance between the two versions.Results:The two physicians had good consistency in the categorization of observations, with kappa values between 0.536 and 0.793. Using LR-5 as the criterion for HCC diagnosis, the Youden index (0.687) of LI-RADS v2018 was higher than that of v2017 (0.612). The sensitivity [80.6% (166/206)] and accuracy [82.6% (233/282)] were both higher than those of LI-RADS v2017 [70.4% (145/206) and 75.9%(214/282)] (χ2=19.048, 14.087, both P<0.001). The specificity was slightly lower [88.2%(67/76) and 90.8%(69/76), respectively], but there was no statistical difference (χ2=0.500, P=0.500). With LR-4+5 as the diagnosis of HCC, the diagnostic performance of the two versions was the same. The sensitivity [91.3% (188/206)] and accuracy [87.6% (247/282)] were higher, and the specificity [77.6% (59/76)] were lower than the LR-5 standard of LI-RADS v2018 (χ2=20.045, P<0.001; χ2=5.633, P=0.018; χ2=16.056, P<0.001), and the Youden index (0.689) was also higher than the LR-5 standard of LI-RADS v2018. Conclusions:Based on Gd-EOB-DTPA enhanced MRI, the LI-RADS v2018 has higher sensitivity and accuracy in diagnosing HCC than v2017. Correct use can provide more objective diagnostic evidence for the clinic.