Machine learning predicts duodenal stump leakage in gastric cancer.

A machine learning model was developed to predict duodenal stump leakage (DSL) in 1,107 gastric cancer patients post-gastrectomy. The model utilized 189 clinical features and evaluated six algorithms, with extreme gradient boosting (XGB) achieving the highest area under the receiver operating characteristic curve (AUROC) score of 0.880. Random forest followed with a score of 0.858. Inclusion of additional postoperative data significantly enhanced predictive accuracy, making predictions more reliable over time.