A Machine Learning-Based Model for Cirrhosis Risk Stratification Incorporating Noninvasive Markers and Clinical...

Area of Science:

• Hepatology
• Medical Informatics
• Machine Learning in Medicine

Background:

• Chronic liver disease (CLD) poses a significant health burden.
• Accurate cirrhosis risk stratification is crucial for timely intervention.
• Existing methods may lack comprehensive integration of diverse predictive markers.

Purpose of the Study:

• To develop and validate a nomogram model for cirrhosis risk stratification in CLD patients.
• To integrate routine laboratory parameters, non-invasive fibrosis markers, and liver elastography.
• To compare the performance of machine learning models against traditional methods.

Main Methods:

• Retrospective enrollment of 344 CLD patients, divided into training (n=241) and validation (n=103) sets.
• Identification of independent predictors using univariate, LASSO, and multivariate logistic regression.
• Construction and internal validation of Random Forest, Support Vector Machine, and Logistic Regression models using Bootstrap method.

Main Results:

• Key predictors identified: age, platelet count, AST/ALT ratio, total bilirubin, abnormal INR, and liver stiffness measurement.
• Random Forest model achieved AUCs of 0.835 (training) and 0.745 (validation).
• Good calibration and superior net clinical benefit of the Random Forest model were observed.

Conclusions:

• A validated cirrhosis risk prediction model integrating diverse parameters was developed.
• The Random Forest model demonstrated strong predictive performance.
• The nomogram offers a simple, interpretable tool for bedside clinical application in CLD risk stratification.