Machine learning-based early survival prediction in early-onset hepatocellular carcinoma: a SEER-based multi-model...

Area of Science:

• Oncology
• Machine Learning in Medicine
• Biostatistics

Background:

• Early-onset hepatocellular carcinoma (eHCC) presents unique clinical challenges.
• Prognostic factors for eHCC survival are not well understood.
• There is a need for improved predictive models for eHCC.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting early survival in eHCC patients.
• To identify key prognostic factors influencing eHCC survival.
• To compare ML model performance against the AJCC staging system.

Main Methods:

• Utilized the SEER database (2010-2021) for eHCC patient data.
• Developed and compared five ML algorithms (MLP, LR, SVM, RF, XGBoost) and an AJCC staging model.
• Assessed model performance using AUC, accuracy, precision, recall, F1 score, and SHAP for variable importance.

Main Results:

• The Random Forest (RF) model achieved an AUC of 0.80 in the validation cohort, outperforming the AJCC model (AUC 0.66).
• RF model metrics included 0.72 accuracy, 0.71 precision, 0.81 recall, and 0.75 F1 score.
• SHAP analysis identified surgical intervention as the most significant predictor in the RF model.

Conclusions:

• The validated RF model demonstrates strong predictive ability for early survival in eHCC.
• This ML-based approach can enhance risk stratification for eHCC patients.
• The findings support the use of ML models to guide individualized treatment decisions in eHCC.