Machine learning-based prediction of short-term recurrence of colorectal adenomatous polyps following EMR: model development and validation study.

OBJECTIVE: This study aims to develop and validate a model for predicting the 1-year recurrence of adenomatous polyps following endoscopic mucosal resection (EMR), and explore associated risk factors. METHODS: Patients who underwent their first EMR for colorectal polyps at the Affiliated Hospital of Xuzhou Medical University from September 2018 to September 2023 were retrospectively enrolled. The dataset was randomly divided into training and testing sets at a ratio of 7:3. Additional patient data from October 2023 to April 2025 from the same center were utilized as the internal validation set, while an external validation set was obtained from Xuzhou Central Hospital. Feature variables were selected via least absolute shrinkage and selection operator (LASSO) regression. Five machine learning (ML) algorithms, including logistic regression (LR), random forest (RF), support vector machine (SVM), gradient boosting machine (GBM), and eXtreme gradient boosting (XGBoost), were used to build predictive models. Model performance was examined via receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Feature importance was interpreted via SHapley Additive exPlanations (SHAP). RESULTS: According to LASSO regression, 10 predictive variables were identified. The performance of the GBM model was the highest, with an area under the curve (AUC) of 0.824 (95 % CI: 0.764-0.884) in the testing set, 0.793 (95 % CI: 0.718-0.868) in the internal validation set, and 0.800 (95 % CI: 0.725-0.875) in the external validation set. Calibration curves indicated close agreement between predicted and observed results. Decision curve analysis demonstrated a satisfactory net benefit across a wide spectrum of threshold probabilities. CONCLUSION: The ML-based model developed in this study accurately predicts the short-term recurrence of adenomatous polyps after EMR and may assist clinicians in personalizing follow-up strategies.