Investigation on explainable machine learning models to predict chronic kidney diseases.

Area of Science:

• Nephrology
• Artificial Intelligence
• Medical Informatics

Background:

• Chronic kidney disease (CKD) is a global health concern with significant morbidity and mortality.
• Early-stage CKD often lacks visible symptoms, hindering timely diagnosis and intervention.
• Effective early detection strategies are vital for managing complications and improving patient quality of life.

Purpose of the Study:

• To investigate the efficacy of an explainable artificial intelligence (XAI)-based strategy for predicting CKD using clinical data.
• To compare the performance of five machine learning (ML) models in CKD prediction.
• To enhance the interpretability of ML models for clinical decision support.

Main Methods:

• Collected clinical data from 491 patients (56 with CKD, 435 without CKD).
• Employed five ML models: logistic regression, random forest, decision tree, Naïve Bayes, and extreme gradient boosting (XGBoost).
• Utilized SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) for feature importance and model interpretability.

Main Results:

• The XGBoost model demonstrated superior performance with an AUC of 0.9689 and accuracy of 93.29%.
• Creatinine, glycosylated hemoglobin type A1C (HgbA1C), and age were identified as the most influential predictors in the XGBoost model.
• SHAP and LIME analyses provided insights into individualized CKD predictions and feature contributions.

Conclusions:

• An interpretable ML-based approach using XAI shows promise for early CKD prediction.
• XAI techniques like SHAP and LIME improve the transparency of ML models in healthcare.
• This approach can assist clinicians in understanding prediction rationales, facilitating better patient management.