Can AI Predict Gangrenous Cholecystitis Before Surgery?: Frontline Medicine | September 2026

What Did the Study Find?

This study developed machine-learning models to identify gangrenous cholecystitis (GC) early using routinely available clinical, laboratory and imaging variables.

• 492 patients and 39 candidate clinical variables were evaluated.
• Seven algorithms were compared, including logistic regression, random forest, XGBoost, LightGBM and an artificial neural network (ANN).
• The ANN performed best, with AUCs of 0.921 internally and 0.867 on external validation.
• Five variables drove most of the prediction: NLR, WBC count, fibrinogen, neutrophil count and gallbladder-wall thickness.
• SHAP analysis was used to make individual predictions more clinically interpretable.
• The model was translated into a web-based risk calculator for potential bedside use.

Why Does It Matter?

Gangrenous cholecystitis can resemble uncomplicated acute cholecystitis before surgery, yet delayed recognition may lead to substantially greater morbidity. A model based largely on routine investigations could help identify patients requiring earlier surgical prioritization.

GastroAGI Takeaway

This is a good example of clinically useful AI built from simple data—not sophisticated imaging or expensive biomarkers. External validation is encouraging, but an AUC of 0.867 does not make the model ready for autonomous decision-making. Prospective multicenter validation and evidence that model-guided decisions improve outcomes are needed before routine implementation.