A radiomics model for preoperative prediction of microvascular invasion as a factor in predicting prognosis of hepatocellular carcinoma after radiofrequency ablation.

[OBJECTIVE] To establish a non-invasive predictive radiomics model of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) and to explore the correlation between MVI status and prognosis of HCC patients after radiofrequency ablation (RFA).

[METHODS] A total of 142 HCC patients with or without MVI were randomly allocated into the training set (n = 99) and the validation set (n = 43). After preprocessing of tri-phase contrast-enhanced MRI, 1688 radiomics features were extracted per lesion. LASSO-Logistic regression was employed for feature selection and radiomics model construction to predict MVI in HCC. The model performance was evaluated using discrimination metrics. Additionally, 58 HCC patients who underwent RFA were enrolled. Univariate and multivariate logistic regression analyses were performed to investigate the impact of MVI on post-RFA prognosis of HCC.

[RESULTS] The radiomics model, especially that derived from the portal venous phase, exhibited effective performance in MVI prediction, achieving an area under the curve (AUC) of 0.888 in the training set and 0.769 in the validation set. Furthermore, the radiomics score was identified as an independent risk factor for recurrence after RFA.

[CONCLUSIONS] This non-invasive radiomics model enables preoperative identification of MVI and prediction of post-RFA recurrence risk in HCC patients, thereby providing valuable evidence for formulating individualized treatment strategies.