Increasing the sensitivity of LI-RADS v2018 for diagnosis of small (10-19 mm) HCC on extracellular contrast-enhanced MRI

摘要

Purpose To evaluate whether the LI-RADS v2018 LR-5 criteria can be modified to increase sensitivity without reducing specificity for diagnosing small (10-19 mm) HCC. Methods 167 consecutive high-risk patients with 174 small observations reported clinically on extracellular contrast-enhanced MRI from 2014 to 2018 were retrospectively studied. The best available reference standard was applied for each observation. Blinded to the reference standard, two radiologists scored LI-RADS imaging features retrospectively and assigned each observation a LI-RADS category using LI-RADS v2018 and each of four modified LI-RADS versions (mLI-RADS I to IV) with successively more expansive LR-5 criteria. Per-observation sensitivity and specificity of LR-5 for small HCC using each version were assessed. Each modified version was compared to v2018 (McNemar test). Results The 174 observations included 135 HCC, 8 non-HCC malignancies, and 31 benign entities. Using LI-RADS v2018, LR-5 provided 70% (both readers) sensitivity and 95% (both readers) specificity for small HCC. Expanding the LR-5 criteria to include nonrim APHE plus at least one additional major feature (mLI-RADS I) or no APHE plus at least two additional major features (mLI-RADS II) significantly increased sensitivity (reader 1/reader 2: 75%/75% vs. 70%,p = 0.016/0.031; 78%/79% vs. 70%,p = 0.001/0.001) without significantly reducing specificity (reader 1/reader 2: 90%/92% vs. 95%,p = 0.500/1.000 for both). mLI-RADS III and IV further increased sensitivity (reader 1/reader 2: 80%/81% vs. 70%,p < 0.001/< 0.001; 94%/92% vs. 70,p < 0.001/< 0.001) but with trend-level (reader 1/reader 2: 85%/80% vs. 95%,p = 0.125/0.063) or significant (reader 1/reader 2: 64%/62% vs. 95%,p < 0.001/< 0.001) specificity reductions. Conclusions Expanding the v2018 LR-5 criteria to include nonrim APHE plus at least one additional major feature or no APHE plus at least two additional major features significantly increases sensitivity without significantly reducing specificity for small HCC. Confirmation is warranted in multi-center prospective studies.