Automated Detection of Gallstone Disease Using Machine Learning Techniques: A Data-Driven Approach

Abstract

Gallstone disease is a very general gastrointestinal condition characterized with origination of stones within the gallbladder, which can lead to inflammation, abdominal pain, and serious complications. Early detection of this disease is helpful for effective treatment. This research presents a comprehensive framework to evaluate the performance of stacked ensemble classifiers in supervised learning tasks. The presented approach employs three heterogeneous model ensembles: (i) XGBoost, LightGBM, and CatBoost; (ii) Random Forest, Extra Trees, and Gradient Boosting; and (iii) K-Nearest Neighbors, Support Vector Classifier, and Decision Tree. These models were integrated with a meta-learning strategy using Logistic Regression within a Stacking-Classifier architecture, with pass through enabled to enrich the meta-model with original input features alongside base learner predictions. Stratified five-fold cross-validation was applied to evaluate the performance and ensure the robustness of the presented model across varying class distributions.  This multi-level validation approach enabled a rigorous comparison across diverse classifier families. The proposed model achieves 97.0\% accuracy, demonstrating a significant improvement over existing research. The experimental results demonstrate that the proposed method has significant potential to support decision-making, reduce diagnostic errors, and improve patient outcomes.