Development of prognostic models for advanced multiple hepatocellular carcinoma based on Cox regression, deep...

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Multiple hepatocellular carcinoma (MHCC) often presents at advanced stages, complicating treatment decisions.
• The AJCC-TNM staging system has limitations in accurately predicting prognosis for MHCC patients.
• There is a need for improved prognostic tools to guide clinical management and patient counseling.

Purpose of the Study:

• To identify key prognostic factors for multiple hepatocellular carcinoma (MHCC).
• To develop and validate a novel prognostic model for advanced MHCC.
• To quantify patient risk and survival probability more accurately than existing systems.

Main Methods:

• Utilized the Surveillance, Epidemiology, and End Results (SEER) database for patient data.
• Employed machine learning (ML) and deep learning (DL) algorithms, including Gradient Boosted Machine (GBM), for model development.
• Evaluated model performance using C-index, ROC curves, Brier score, and decision curve analysis, with SHAP for interpretability.

Main Results:

• The Gradient Boosted Machine (GBM) model demonstrated superior prognostic accuracy for advanced MHCC.
• The GBM model achieved a C-index of 0.73 and AUC values above 0.78 in the training cohort.
• The model showed robust performance in the test cohort and effectively stratified patient prognosis via Kaplan-Meier analysis.

Conclusions:

• The Gradient Boosted Machine (GBM) model is the most accurate machine learning approach for predicting prognosis in advanced MHCC.
• This new model offers a valuable tool for risk stratification and survival prediction in MHCC patients.
• The findings support the use of advanced ML techniques in oncology for improved patient outcomes.