To evaluate the value of gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) enhanced MRI in differential diagnosis of dysplastic nodules (DN) and small well-differentiated hepatocellular carcinoma (sWDHCC) induced by liver cirrhosis, and to construct a differential diagnosis model.

Clinical imaging data of 116 patients pathologically diagnosed with DN or sWDHCC in the Third Affiliated Hospital of Sun Yat-sen University from January 2015 to December 2020 were retrospectively analyzed. The informed consents of all patients were obtained and the local ethical committee approval was received. Among them, 98 patients were male and 18 female, aged from 31 to 74 years, with a median age of 59 years. All patients received Gd-EOB-DTPA enhanced MRI to evaluate the imaging features of the lesions in in-phase and out-of-phase MRI, fat-suppression T1WI, fat-suppression T2WI, diffusion-weighted imaging (DWI), arterial phase, venous phase and hepatobiliary phase, and to assess whether the margins of the lesions were clear. Imaging characteristics between two groups were compared by Chi-square test. The independent influencing factors of differential diagnosis were identified by multivariate Logistic regression analysis and a regression model was constructed. The differential diagnostic efficiency of this model for sWDHCC and DN was evaluated by the ROC curve.

Gd-EOB-DTPA enhanced MRI showed that 116 lesions from 116 patients included 44 DN and 72 sWDHCC. Univariate analysis showed that the proportion of patients with arterial phase hyperenhancement (APHE), DWI hyperintensity and T2WI hyperintensity in the sWDHCC group was significantly higher than that in the DN group (χ²=68.280, 7.408, 5.500; all P<0.05). Multivariate Logistic regression analysis demonstrated that APHE (OR=0.001, 95%CI: 0.000-0.009) and washout in venous phase (OR=0.058, 95%CI: 0.010-0.353) were the independent influencing factors in differentiating sWDHCC from DN (both P<0.05). The regression model was constructed: Y=3.836-6.588× arterial phase features-2.840×venous phase features. The area under the ROC curve (AUC), sensitivity and specificity of this model in differentiating sWDHCC from DN were 0.94, 0.90 and 0.89, respectively. The AUC of APHE, DWI hyperintensity and T2WI hyperintensity was 0.86, 0.63 and 0.61, respectively. The AUC of the regression model was significantly larger than that of APHE, DWI hyperintensity and T2WI hyperintensity (Z=3.103, 6.176, 6.795; all P<0.05).

Washout during APHE and venous phase of Gd-EOB-DTPA enhanced MRI are the imaging features for differential diagnosis between sWDHCC and DN. This regression model based on imaging features possesses high differential diagnostic value.