Predicting hepatocellular carcinoma survival with artificial intelligence.

Area of Science:

• Oncology
• Medical Informatics
• Bioinformatics

Background:

• Hepatocellular carcinoma (HCC) survival rates remain poor despite extensive research.
• Predictive models for HCC patient survival are crucial for treatment planning.

Purpose of the Study:

• To evaluate the efficacy of machine learning (ML) methods in predicting survival probability for HCC patients.
• To identify key predictors of mortality using feature selection techniques.

Main Methods:

• Retrospective analysis of 393 HCC patients (stages 1-4).
• Utilized demographic, clinical, pathological, and laboratory data.
• Employed various ML algorithms and feature selection methods for survival prediction.

Main Results:

• ML models achieved up to 91% recall for 6-month survival in early-stage HCC (stages 1-2).
• Models reached up to 92% accuracy for 3-year overall survival in advanced-stage HCC (stage 4).
• Weighted KNN and SVM models demonstrated high accuracy (87.5%-87.8%) in predicting patient outcomes.

Conclusions:

• Machine learning reliably predicts survival probabilities for HCC patients across all disease stages.
• AI models can accurately identify surviving individuals by analyzing clinical and pathological factors.