Building Trust in Clinical AI: A Web-Based Explainable Decision Support System for Chronic Kidney Disease.

Area of Science:

• Nephrology
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Chronic Kidney Disease (CKD) affects over 10% of the global population, necessitating early diagnosis for effective management.
• Machine learning (ML) presents a significant opportunity for advancing predictive diagnostics in healthcare settings.

Purpose of the Study:

• To develop and evaluate a Web-Based Clinical Decision Support System (CDSS) for predicting Chronic Kidney Disease (CKD).
• To integrate Explainable AI (XAI) methods, including SHAP and LIME, to enhance model transparency and trustworthiness.

Main Methods:

• Evaluation of multiple ML classifiers: KNN, Random Forest, AdaBoost, XGBoost, CatBoost, and Extra Trees for CKD prediction.
• Performance assessment using accuracy, confusion matrix statistics, and Area Under the Curve (AUC).
• Development of a real-time web-based application for practical implementation.

Main Results:

• The AdaBoost classifier achieved a perfect 100% accuracy rate in predicting CKD.
• All evaluated classifiers, except KNN, demonstrated perfect precision and sensitivity.
• The developed CDSS operationalizes ML models, improving accessibility for healthcare practitioners.

Conclusions:

• The study successfully developed a highly accurate and interpretable ML-based CDSS for CKD prediction.
• Explainable AI (XAI) integration enhances the clinical utility and trustworthiness of predictive models.
• The real-time application facilitates the adoption of advanced diagnostics in clinical practice.