CEUS ‐Based Microvascular Invasion Predictor in HCC : Improving Prognostic Stratification Following Thermal Ablation

Key Points

• This research aims to develop a deep learning model to assess microvascular invasion risk in hepatocellular carcinoma before surgery and evaluate its prognostic implications post-thermal ablation.
• Enrolled 688 patients with solitary hepatocellular carcinoma ≤ 5 cm undergoing contrast-enhanced ultrasound before surgery or thermal ablation.
• Extracted CEUS features using a vision transformer and integrated them with clinical data to create a microvascular invasion classifier.
• Compared recurrence-free survival between risk groups in the thermal ablation cohort using Cox regression analysis.
• The model achieved AUCs of 0.89 for internal validation and 0.81 for external validation in predicting microvascular invasion.
• Patients with high-risk microvascular invasion showed significantly lower recurrence-free survival rates: 75.5% at 1 year, 54.9% at 2 years, and 39.4% at 3 years (p < 0.001).
• High MVI risk predicted by the model had a hazard ratio of 2.90 (95% CI: 1.69-4.96) and was independently associated with recurrence-free survival.