Lymph node metastasis in patients with hepatocellular carcinoma using machine learning: a population-based study.

Area of Science:

• Oncology
• Machine Learning
• Medical Informatics

Background:

• Hepatocellular carcinoma (HCC) is a major global health concern.
• Accurate prediction of lymph node metastasis (LNM) in HCC is crucial for patient management and prognosis.
• Current methods for LNM prediction in HCC may lack precision, necessitating improved tools.

Purpose of the Study:

• To develop and validate a population-adapted machine learning model for predicting LNM in HCC.
• To identify key indicators associated with LNM risk in HCC patients.
• To provide clinicians with a reliable tool for identifying high-risk HCC patients requiring intensive surveillance.

Main Methods:

• Analysis of a large cohort (23,511 patients) from the SEER database and a smaller cohort (57 patients) from a local hospital.
• Utilized seven identified LNM risk indicators.
• Developed prediction models using Decision Tree (DT), Logistic Regression (LR), Multilayer Perceptron (MLP), and Extreme Gradient Boosting (XGBoost) algorithms.
• Evaluated model performance using Area Under the Curve (AUC), accuracy, sensitivity, and specificity.

Main Results:

• Lymph node metastasis (LNM) was present in 7.14% of the SEER cohort.
• Independent predictors of LNM included race, sequence number, tumor size, T stage, and AFP levels.
• The Logistic Regression (LR) model demonstrated optimal performance with an AUC of 0.751 in the SEER cohort.
• External validation in the local cohort showed robust generalizability for the LR model (AUC: 0.73).

Conclusions:

• The LR-based machine learning model exhibits superior predictive capability for LNM in HCC.
• This model serves as a valuable clinical tool for guiding personalized therapeutic strategies in HCC management.
• The findings support the use of machine learning for enhancing precision in HCC patient stratification.