A proposal for a useful algorithm to diagnose small hepatocellular carcinoma on MRI.

Abstract

To assess MRI features for the diagnosis of small hepatocellular carcinomas (HCCs) and especially for nodules not showing both of the typical hallmarks. Three hundred and sixty-four cirrhotic patients underwent liver MRI for 10-30 mm nodules suggestive of HCC. The diagnostic performances of MRI features [T1, T2; diffusion-weighted (DW) imaging signal, enhancement, capsule, fat content] were tested, both individually and in association with both typical hallmarks and as substitutions for one hallmark. The diagnostic reference was obtained using a multifactorial algorithm ensuring high specificity (Sp). Four hundred and ninety-three nodules were analyzed. No alternative features, associations or substitutions outperformed the typical hallmarks for the diagnosis of HCC. For 10-20 mm nodules not displaying one of the typical hallmarks, hyperintensity on DW images was the most accurate substitutive sign, providing a sensitivity of 71.4% and Sp of 75% for nodules without arterial enhancement and sensitivity = 65.2% and Sp = 66% for nodules without washout on the portal or delayed phases. A new diagnostic algorithm, including typical hallmarks as a first step then the best-performing substitutive signs (capsule presence or DW hyperintensity) in combination with the nonmissing typical hallmark as a second step, enabled the correct classification of 77.7% of all nodules, regardless of size. Using MRI, the typical hallmarks remain the best criteria for the diagnosis of small HCCs. However, by incorporating other MRI features, it is possible to build a simple algorithm enabling the noninvasive diagnosis of HCCs displaying both or only one of the typical hallmarks.