Interpretable Machine Learning Model for Predicting and Assessing the Risk of Diabetic Nephropathy: Prediction Model...

Area of Science:

• Nephrology
• Artificial Intelligence
• Data Science

Background:

• Diabetic nephropathy (DN) affects 30-40% of diabetes patients, leading to kidney failure.
• Current diagnostic methods for DN lack sensitivity and specificity for early detection.
• Accurate, interpretable predictive models are crucial for timely intervention and improved patient care.

Purpose of the Study:

• Develop and validate a machine learning (ML) model for predicting DN in type 2 diabetes patients.
• Enhance model transparency and interpretability using explainable AI (XAI) techniques.
• Support early DN diagnosis, risk stratification, and personalized clinical decision-making.

Main Methods:

• Retrospective cohort study of 1000 type 2 diabetes patients (2015-2020).
• Utilized Extreme Gradient Boosting (XGBoost), CatBoost, and Light Gradient-Boosting Machine (LightGBM) algorithms.
• Applied Local Interpretable Model-Agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP) for interpretability.

Main Results:

• XGBoost and LightGBM showed superior performance in predicting DN.
• XGBoost achieved 86.87% accuracy, 88.90% precision, 84.40% recall, and 89.12% specificity.
• Serum creatinine, albumin, and lipoproteins identified as key predictors through LIME and SHAP analyses.

Conclusions:

• The developed ML model offers a robust and interpretable tool for early DN detection and risk assessment.
• The model's transparency is vital for clinical integration and trust.
• Potential to improve patient outcomes and optimize healthcare resource allocation through early intervention.