A Machine Learning Model to Predict Risk for Hepatocellular Carcinoma in Patients With Metabolic...

Area of Science:

• Hepatology
• Oncology
• Data Science

Background:

• Hepatocellular carcinoma (HCC) incidence is rising, particularly in patients with metabolic dysfunction-associated steatotic liver disease (MASLD).
• Patients with MASLD, even without advanced fibrosis, face higher risks of advanced HCC stages, reduced survival, and limited transplant eligibility.
• Machine learning (ML) offers a promising approach to analyze complex datasets for predicting individual HCC risk.

Purpose of the Study:

• To develop and validate a machine learning model for predicting HCC development in patients with MASLD.
• To identify key predictors of HCC risk within the MASLD population.
• To enable early risk stratification and guide targeted screening strategies.

Main Methods:

• Utilized Tableau and KNIME Analytics for data analysis and ML tasks.
• Developed ML models using standard clinical and laboratory parameters, employing Sci-kit learn algorithms.
• Trained a pilot model on data from UC Davis and validated it using an independent dataset from UC San Francisco.

Main Results:

• The predictive model achieved 92.06% accuracy in the validation cohort, with an AUC of 0.97.
• Liver fibrosis, assessed by the Fibrosis-4 score, emerged as the strongest predictor of HCC.
• The model demonstrated high performance with an F1-score of 0.84, 98.34% specificity, and 74.41% sensitivity.

Conclusions:

• ML models can significantly assist clinicians in early HCC risk assessment for MASLD patients.
• Further validation of these models can lead to cost-effective, personalized management strategies for at-risk individuals.
• This approach supports proactive healthcare decisions in MASLD management.