Related Experiment Videos
Predicting MASLD in Gallstone Patients: An Interpretable Machine Learning Model with External Validation
1Department of General Surgery, Wuhan Third Hospital (Tongren Hospital of Wuhan University), Wuhan, Hubei, 430074, People's Republic of China.
Background:
Previous studies have shown a higher prevalence of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) in individuals with gallstones, as both conditions share common risk factors, including hyperlipidemia, obesity, physical inactivity, and diabetes. This study aimed to develop and validate a machine learning (ML)-based model for predicting MASLD risk in gallstone patients.
Methods:
A multicenter retrospective cross-sectional study was conducted, enrolling 390 gallstone patients from the Guanggu Branch of our hospital (January 2023-August 2025) as the internal cohort for model development. An external validation cohort of 134 subjects from the Shouyi Branch, matched in timeframe, was also included. The Synthetic Minority Oversampling Technique (SMOTE) was applied to address class imbalance in the internal training set. Feature selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression and the Boruta algorithm, followed by model development and validation using seven ML algorithms. Model performance was evaluated using calibration curves, and decision curves. The Support Vector Machines (SVM)-based diagnostic model was further interpreted using SHapley Additive exPlanations (SHAP).
Results:
In the study cohort, 15.46% of gallstone patients were diagnosed with MASLD. Seven variables were identified as significant predictors of comorbid MASLD in patients with gallstones. Among the seven ML models, the SVM algorithm achieved the highest AUC (0.902; 95% confidence interval: 0.857-0.946). The model's performance was confirmed in both the internal and external validation groups, with AUC values exceeding 0.7 in both cases. Calibration curves indicated strong model calibration. SHAP interpretation revealed that the top three predictive features were AIP, hyperlipidemia, and hemoglobin.
Conclusion:
A predictive model for concurrent MASLD risk in gallstone patients was developed and validated using seven clinical indicators.