Objective  To investigate the value of 3D-Dixon magnetic resonance imaging (MRI) based on chemical shift effects and diffusion-weighted imaging (DWI) in preoperative pathological grading of hepatocellular carcinoma (HCC).

  Methods  We retrospectively analyzed preoperative MRI data from 109 patients with HCC confirmed by pathological examination from January 2016 to December 2017 in Second Affiliated Hospital of Nanchang University. The intratumoral fat fraction (FF) and minimum apparent diffusion coefficient (ADCmin) values were measured using chemical shift with in/opposed-phase (IOP) and DWI sequences, respectively. The differences in the FF and ADCmin values of the solid components in HCC tumors with different pathological differentiation levels were compared, and the correlations between FF and ADCmin values and pathological grade were analyzed. Receiver operating characteristic (ROC) curves were used to determine the differential value, sensitivity, and specificity of FF and ADCmin values for different pathological grades.

  Results  A total of 109 patients were included in this study, 24 in the poorly differentiated group, 69 in the moderately differentiated group, and 16 in the highly differentiated group. The FF and ADCmin value differences between each pathological differentiation group were statistically significant. There was a positive correlation between FF and pathological grade (r=0.566, P < 0.01). An FF of 8.71% was determined to be the cutoff for distinguishing between poorly differentiated and moderately differentiated tumors (sensitivity, 75.0%; specificity, 67.19%; area under the ROC curve, 0.738). The cutoff for distinguishing between moderately and highly differentiated tumors was 5.53% (sensitivity, 71.87%; specificity, 75.0%; area under the ROC curve, 0.843). There was a negative correlation between ADCmin values and pathological grade (r=-0.674, P < 0.01). The ADCmin cutoff value for distinguishing between poorly and moderately differentiated tumors was 0.379 × 10^-3 mm²/s (sensitivity, 79.17%; specificity, 84.37%; area under the ROC curve, 0.890). The cutoff value for distinguishing between moderate and highly differentiated tumors was 0.532 × 10^-3 mm²/s (sensitivity, 89.06%; specificity, 62.50%; area under the ROC curve, 0.812).

  Conclusions  3D-Dixon and DWI findings play an important role in determining the diagnosis and prognosis of HCC. The FF and minimum ADC values are significantly correlated with HCC pathological grading and can be used as noninvasive biological indicators to evaluate histopathological factors of HCC.