Machine learning algorithms outperform conventional regression models in predicting development of hepatocellular...

Area of Science:

• Hepatology
• Oncology
• Data Science

Background:

• Hepatocellular carcinoma (HCC) risk prediction in cirrhosis patients often lacks accuracy and validation.
• Machine learning (ML) offers a novel approach to enhance HCC prognostication.

Purpose of the Study:

• To develop and compare predictive models for HCC development in cirrhotic patients.
• Utilizing conventional regression analysis and ML algorithms.

Main Methods:

• 442 Child A/B cirrhosis patients (UM cohort) were followed prospectively.
• Models were developed using regression and ML, validated on the HALT-C Trial cohort.
• Model performance was assessed using c-statistics, net reclassification improvement, and integrated discrimination improvement.

Main Results:

• The ML algorithm achieved a c-statistic of 0.64 (95% CI 0.60-0.69) in the validation cohort.
• ML significantly outperformed the UM regression model (c-statistic 0.61) and the HALT-C model (c-statistic 0.60).
• ML demonstrated superior diagnostic accuracy via net reclassification improvement (P<0.001) and integrated discrimination improvement (P=0.04).

Conclusions:

• Machine learning algorithms enhance the accuracy of risk stratification for HCC in cirrhosis patients.
• ML models can accurately identify high-risk individuals for HCC development.