New strategy of LI-RADS v2018 to improve the sensitivity for small hepatocellular carcinoma ≤ 3.0 cm on extracellular-contrast enhanced MRI

Introduction: We aimed to modify LI-RADS version 2018 to improve sensitivity and determine the value of the combination of high alpha-fetoprotein (AFP) levels for small HCC (sHCC, <= 30 mm) diagnosis. Methods: A total of 984 patients at high risk for HCC, with 1204 observations (including 997 small observations <= 30 mm), who underwent extracellular contrast-enhanced MRI were enrolled from five independent centers. Blinded readers evaluated the LI-RADS features and categorized each observation according to the LI-RADS v2018, modified LI-RADS and EASL. Odds ratios of LI-RADS major features (MFs) and several high AFP levels for sHCC diagnosis were analyzed using multivariable logistic regression. The modified LR-5 criteria was developed by including no APHE at any size with two MFs, and non-rim APHE with one MF (>= 10 mm) or with two MFs (< 10 mm). The diagnostic performance of each version of the LR-5 was compared using generalized estimating equations. Results: APHE, washout, enhancing capsule and five high AFP levels were independently associated with sHCC. In three datasets, the modified LI-RADS had higher sensitivities for sHCC (76.8 similar to 85.5 % vs. 73.7 similar to 75.9 %, P < 0.05) to the LR-5 v2018. The modified LI-RADS with AFP >= 200 ng/mL as an additional feature or as an alternative to threshold growth provided higher sensitivities for sHCC than LI-RADS v2018 (82.1 similar to 90.1 % vs. 73.7 similar to 75.9 %, all P < 0.05), modified LI-RADS (82.1 similar to 90.1 % vs. 76.8 similar to 85.5 %, all P <0.05) and EASL version 2018 (82.1 similar to 90.1 % vs. 73.3 similar to 74.7 %, all P < 0.05), with comparable specificities (all P > 0.05). Conclusion: The new strategy of LI-RADS v2018 provides significantly higher sensitivity and comparable specificity than those of LI-RADS v2018 for sHCC diagnosis on ECA-MRI.