Introduction: Previous studies have shown that contrast enhanced computed tomography (CT) radiomics features of tumor can predict the microvascular invasion (MVI) of hepatocellular carcinoma (HCC), but which area of the HCC contributes the most remains unclear. This study aimed to study the spatial heterogeneity of HCC through contrast-enhanced CT radiomics, and to explore whether the tumor invasive front can better predict MVI. Methods: A cohort of 155 patients with primary HCC ≥ 2 cm were retrospectively analyzed. Radiomics features were extracted from the portal phase of contrast enhanced CT including the entire tumor, tumor invasive front (0.5cm inside the tumor edge), and tumor center (1cm outside the tumor edge). The LASSO regression was applied for features selection, and the obtained radiomics features were further used to construct the MVI prediction model. The predictive capability of the models was evaluated by the area under the receiver operating characteristic curves (AUCs). Results: Six, three and zero radiomics features were obtained from 851 features in the entire tumor, tumor invasive front, and tumor center, respectively. The AUCs of the models constructed with the radiomics features of the entire tumor and tumor invasive front were 0.7765 and 0.7647, respectively, and there was no statistically significant difference in predictive power between the two models. Conclusions: The tumor invasive front radiomics features of contrast enhanced CT for HCC can better predict MVI. The tumor invasive front has a greater contribution to MVI, and the degree of malignancy in this area may be higher.