Machine Learning for Lymph Node Metastasis Prediction in Early Gastric Cancer: A Comparative Analysis.

Area of Science:

• Oncology
• Medical Informatics
• Surgical Oncology

Background:

• Lymph node metastasis (LNM) is critical for staging and treatment of early gastric cancer (EGC).
• Accurate prediction of LNM is essential for guiding surgical and adjuvant therapy decisions in EGC.
• Current methods for LNM prediction in EGC may benefit from advanced analytical approaches.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for predicting lymph node metastasis (LNM) in early gastric cancer (EGC).
• To identify key predictors of LNM in EGC patients to inform clinical decision-making.
• To compare the performance of various machine learning models in predicting LNM in EGC.

Main Methods:

• Data from 1085 EGC patients undergoing gastrectomy with lymph node resection were analyzed.
• Seven machine learning algorithms were trained and validated, with hyperparameter tuning.
• Model performance was assessed using accuracy, Brier class, and Area Under the Curve (AUC).

Main Results:

• Random Forest, Extreme Gradient Boosting, and Neural Network models showed strong performance on the validation set (AUCs 0.796, 0.788, 0.779).
• Subgroup analyses in T1a and T1b stages revealed varying model performances (e.g., Logistics Models and RF for T1a).
• SHAP analysis identified distinct variable importance for LNM prediction in EGC and its subgroups.

Conclusions:

• Machine learning models demonstrate potential for predicting LNM in EGC, enhancing treatment strategy development.
• Identified risk factors provide valuable insights for personalized management of EGC patients.
• These predictive models can support clinical decision-making for EGC with LNM.