Clinical decision system for chronic kidney disease staging using machine learning.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Nephrology

Background:

• Chronic Kidney Disease (CKD) affects many individuals, necessitating stage-specific treatment plans.
• Machine Learning (ML) and Generative AI show potential for predicting CKD progression.
• Current ML models face challenges in generalizability, interpretability, and resource demands.

Purpose of the Study:

• To develop a clinical support system for accurate CKD stage classification using ML.
• To enhance prediction accuracy and guide personalized treatment through feature selection and model evaluation.

Main Methods:

• Employed ML algorithms: Gradient Boosting, XGBoost, CatBoost, and GAN AML.
• Utilized feature selection techniques: Recursive Feature Elimination, chi-square test, and SHAP.
• Evaluated models using precision, recall, F1-score, accuracy, and AUC-ROC.

Main Results:

• CatBoost and GAN AML demonstrated high effectiveness in classifying CKD stages.
• Expert knowledge in feature selection significantly improved ML model performance.
• The study identified key features for accurate CKD staging.

Conclusions:

• CatBoost and GAN AML are effective tools for CKD stage classification.
• Strategic feature selection is vital for optimizing ML models in nephrology.
• Future work should focus on dataset diversity, clinical integration, and model interpretability.