Prediction Model of Ocular Metastases in Gastric Adenocarcinoma: Machine Learning-Based Development and...

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Gastric adenocarcinoma (GA) with ocular metastasis (OM) signifies advanced disease.
• Ocular metastasis is a rare but serious complication of gastric cancer.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting the risk of ocular metastasis in gastric adenocarcinoma patients.
• To identify key clinical and biochemical risk factors associated with GA-related OM.

Main Methods:

• Retrospective cohort study of 3532 GA patients.
• Utilized machine learning algorithms, including gradient boosting machine (GBM), and Shapley additive interpretation (SHAP) for risk factor identification.
• Model performance evaluated using area under the receiver operating characteristic curve (AUC).

Main Results:

• The GBM model achieved an AUC of 0.997 in the test set, demonstrating high predictive accuracy.
• Identified key risk factors for OM: LDL, CA724, CEA, AFP, CA125, Hb, CA153, and Ca2+.
• A functional online risk prediction calculator was developed.

Conclusions:

• Machine learning, particularly GBM, provides a robust method for predicting GA-related OM.
• The developed model and calculator can aid in early identification of patients at risk for OM, facilitating prompt treatment.