Risk factors for hepatocellular carcinoma rupture: multicentre retrospective study

Area of Science:

• Hepatology and oncology research.
• Application of machine learning in clinical prediction.
• Medical data analysis and predictive modeling.

Background:

• Hepatocellular carcinoma (HCC) rupture is a severe complication with poor outcomes.
• Accurate risk stratification is crucial for timely intervention.
• Existing predictive tools require enhancement for clinical utility.

Purpose of the Study:

• To identify significant risk factors for HCC rupture.
• To develop and validate a predictive model for HCC rupture.
• To integrate machine learning for enhanced risk prediction accuracy.

Main Methods:

• Retrospective analysis of 5952 HCC patients.
• Propensity score matching to balance patient groups.
• Development of random forest and deep learning models.
• Evaluation using AUC, precision, recall, and F1 score.

Main Results:

• Cirrhosis, protrusion ratio, and maximum tumor length identified as key risk factors.
• The CAPTure nomogram achieved AUCs of 0.857-0.840.
• Machine learning models (random forest, deep learning) showed AUCs of 0.870-0.872.

Conclusions:

• The CAPTure model offers a practical and accurate tool for HCC rupture risk assessment.
• Integration of traditional and machine learning methods improves predictive capabilities.
• Findings support early risk identification and optimized management of HCC rupture.