To explore the clinical value of a multi-parameter regression model based on preoperative microvascular flow imaging (MFI) combined with contrast-enhanced ultrasound (CEUS) in predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

A retrospective analysis was conducted on the clinical data of 186 HCC patients who had undergone postoperative pathological confirmation in Tangshan Third Hospital from January 2022 to August 2025. All patients underwent MFI and CEUS examinations before surgery. The postoperative pathological diagnosis of MVI was used as the gold standard to divide the patients into the MVI positive group (72 cases) and the MVI negative group (114 cases). Univariate analysis and LASSO regression were used to screen the predictive factors, and a combined predictive model was constructed through multivariate Logistic regression. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic efficacy of the model and each independent predictive factor.

The maximum diameter of tumors in the MVI positive group was significantly larger than that in the MVI negative group, and the proportion of patients with alpha-fetoprotein (AFP) >400 μg/L was also higher (both P<0.001). The proportions of high MFI blood flow grading, peritumoral nutrient vessels, heterogeneous enhancement, early clearance ≤60 seconds, and peritumoral enhancement in the MVI positive group were significantly higher than those in the MVI negative group (all P<0.001). Spearman correlation analysis showed that the MVI status was weakly to moderately positively correlated with MFI blood flow grading, peritumoral nutrient vessels, heterogeneous enhancement, early clearance, and peritumoral enhancement (r ranged from 0.173 to 0.419, all P<0.05). Multivariate Logistic regression analysis showed that the maximum tumor diameter (OR=2.182), AFP> 400 μg/L (OR=6.664), high MFI blood flow grading (OR=7.126), heterogeneous enhancement (OR=4.418), and peritumoral enhancement (OR=5.849) were independent risk factors for MVI (all P<0.05). ROC analysis showed that the area under the curve of the combined prediction model was 0.930, which was superior to any single indicator (all P<0.001), with a sensitivity of 0.764 and a specificity of 0.939.

The combined prediction model constructed based on preoperative MFI and CEUS can effectively predict the MVI status of HCC. Its diagnostic value is significantly superior to that of either imaging or serological indicators, and it is expected to provide an important reference for the preoperative individualized treatment decisions of HCC patients.