Prediction of Microvascular Invasion of Hepatocellular Carcinoma by Pre-Operative CT Imaging: A Retrospective Analysis

Background: Hepatocellular carcinoma (HCC) is the most common primary liver cancer, curative hepatectomy is a good treatment option, but its results remain unsatisfactory due to high recurrence rates related to microvascular invasion. Objectives: This study aimed to evaluate preoperative computed tomography (CT) features for predicting microvascular invasion (MVI) in HCC. Materials and methods: This retrospective study analyzed 57 patients with HCC who underwent curative hepatectomy. Preoperative triphasic CT assessed tumor size, margin (smooth/non-smooth), capsule status (present/absent), peritumoral enhancement (present/absent) and two-trait predictor of venous invasion (present/absent). Post-operative histopathologic detection of MVI served as the gold reference standard. Results: Age and operation were statistically insignificantly different between present and absent Histopathological microvascular invasion groups. Sex (female) was statistically significantly higher in present microvascular invasion group than absent microvascular invasion group (P value=0.047). Number of FLs and internal arteries (intratumoral arteries) were statistically insignificantly different between present and absent histopathological microvascular invasion groups. Size of tumor, non-smooth tumor margin and peritumoral enhancement were statistically significantly higher in present microvascular invasion group than in absent microvascular invasion group (P value<0.05). Tumor capsule and continuous hypo-attenuating halo was statistically significantly lower in present microvascular invasion group than absent microvascular invasion group (P value < 0.001 and 0.002 respectively). In univariate regression, size of tumor, non-smooth tumor margin and continuous hypo-attenuating halo were independent predictors of microvascular invasion while sex, tumor capsule and peritumoral enhancement were not. In multivariate regression, size of tumor, non-smooth tumor margin and continuous hypo-attenuating halo were independent predictors of microvascular invasion. Conclusion: Non-smooth margin can predict microvascular invasion with 86.11 % sensitivity, 90.48 % Specificity, 93.9% PPV and 79.2% NPV (P value<0.001). Absent continuous hypo-attenuating halo can predict microvascular invasion with 91.67% sensitivity, 42.86 % specificity, 73.3% PPV and 75.00 % NPV (P value=0.028).