Construction of a Preoperative Prediction Model for TACE Resistance in Primary Hepatocellular Carcinoma Based on...

Area of Science:

• Hepatocellular Carcinoma (HCC) Research
• Machine Learning in Oncology
• Medical Imaging Analysis

Background:

• Transcatheter arterial chemoembolization (TACE) is a key treatment for unresectable hepatocellular carcinoma (HCC).
• Treatment resistance to TACE significantly impacts patient prognosis.
• Accurate preoperative prediction of TACE resistance is crucial for treatment planning.

Purpose of the Study:

• To develop an interpretable machine learning (ML) model for preoperative prediction of TACE resistance in HCC.
• To utilize Shapley Additive Explanations (SHAP) for understanding feature importance in TACE resistance prediction.
• To identify key predictors of TACE resistance for improved patient stratification.

Main Methods:

• A retrospective analysis of 562 HCC patients who underwent TACE was conducted.
• Multi-modal data including blood tests, coagulation, biochemistry, and imaging features were integrated.
• Seven ML models were trained and evaluated, with feature importance analyzed using SHAP.

Main Results:

• The XGBoost model demonstrated superior performance, achieving an Area Under the Curve (AUC) of 0.942 in the training set and 0.898 in the validation set.
• Key predictors identified include Neutrophil-to-Lymphocyte Ratio (NLR) and tumor capsule integrity.
• The model achieved an F1 score of 0.741, indicating robust predictive capability.

Conclusions:

• An interpretable ML model effectively predicts TACE resistance using multi-modal data with high accuracy (AUC > 0.8).
• This model serves as a valuable preoperative tool for identifying patients at high risk of TACE resistance.
• The findings support the clinical translation of ML for optimizing HCC treatment strategies.