Key result
Stacking ensemble machine learning model predicts IVIG resistance in Kawasaki disease with ~99% accuracy.
Why the study?
Does an explainable ensemble machine learning framework accurately predict IVIG resistance in patients with Kawasaki disease?
Does an explainable ensemble machine learning framework accurately predict IVIG resistance in patients with Kawasaki disease?
Answers are grounded in this paper and cite their sources.
An explainable stacking ensemble machine learning model can accurately predict IVIG resistance in Kawasaki Disease, offering clinicians an interpretable tool for early identification.
As a major reason for serious coronary complications if not treated appropriately, acute pediatric inflammatory conditions like Kawasaki Disease and Intravenous immunoglobulin (IVIG) resistance necessitate precise analytical assistance systems. With the objective of improving early identification of IVIG-resistant patients, the explainable machine learning architecture presented in this study was developed using routinely available clinical and laboratory information to address the complex and heterogeneous nature of the disease, as early identification of IVIG-resistant patients is clinically important yet remains challenging. The method followed involved Random Forest-based feature selection, and SMOTEENN sampling were used as preprocessing techniques to improve feature significance while addressing disparity. Several ensemble techniques, such as voting, blending, and stacking to enhance predictive performance, and eleven classifiers, including Logistic Regression, Decision Tree, Random Forest, Extra Trees, XGBoost, CatBoost, LightGBM, AdaBoost, Support Vector Machine, K Nearest Neighbors, and Gaussian Naive Bayes, were examined. The obtained results showed that the stacking ensemble model combining Random Forest, Decision Tree, and Extra Trees performed best, achieved 99.26% accuracy, 99.39% F1 score, and an AUC score of 99.94%. We have conducted an ablation study to determine the specific contribution of every component, sampling, features, and tuning of the proposed model's overall performance. Calibration in our research ensures that the proposed model's predicted probabilities match real-world frequencies. External validation was conducted to justify our proposed model's generalizability on an independent dataset that was not used during model development. We have used Explainable AI SHAP and LIME to further validate our findings and show models’ transparency. Thus, the interpretability of the clinical features can be made by the clinicians. In conclusion, we can suggest that the strategy we have followed gives us an accurate result for the identification of IVIG-resistant in Kawasaki Disease (KD) patients, which is also interpretable.
No takes yet. Share an insight, caveat, or question.
Khan et al. (2026) studied Kawasaki Disease with Intravenous immunoglobulin (IVIG) resistance. Stacking ensemble machine learning model vs. Other machine learning classifiers was evaluated on Prediction of IVIG resistance. A stacking ensemble model combining Random Forest, Decision Tree, and Extra Trees accurately predicted IVIG resistance in Kawasaki disease, achieving 99.26% accuracy and an AUC of 99.94%.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: