Machine Learning for Anastomotic Leak Prediction: A Systematic Review and Experimental Validation

Binesh Sadanandan

Abstract: Anastomotic leak remains one of the most serious complications after gastrointestinal surgery, with reported incidence rates between 2.8% and 8.4% in colorectal procedures. Early identification of high-risk patients could enable targeted interventions and improve outcomes. We conducted a systematic review of machine learning approaches for anastomotic leak prediction, identifying 22 studies from PubMed and Embase. Most models achieved area under the curve (AUC) values between 0.65 and 0.89, with random forests and neural networks showing the strongest performance. However, few studies validated their models on external datasets, and reproducibility remains a significant concern. We also present an initial experiment using the MIMIC-IV database, achieving an AUC of 0.77 with gradient boosting on structured clinical features. Our findings highlight both the promise and current limitations of machine learning for surgical risk prediction, and we identify key directions for future research including the need for standardized benchmarks, external validation, and integration studies with clinical workflows.