Interpretable Machine Learning Model for Predicting and Risk Assessment of Diabetic Nephropathy (Preprint)

Abstract

UNSTRUCTURED Introduction: Diabetic Nephropathy (DN), severe complications of diabetes, is characterized by proteinuria, hypertension, and progressive renal function decline, potentially leading to end-stage renal disease (ESRD). DN's pathogenesis involves high glucose levels, oxidative stress, inflammation, and fibrosis, resulting in kidney changes such as glomerular basement membrane thickening and glomerulosclerosis. The International Diabetes Federation projects that by 2045, 783 million people will have diabetes, with 30%-40% of them developing DN. Early detection and intervention are crucial for preserving renal function, improving quality of life, eliminating cardiovascular complications, and reducing healthcare costs. Methods: This study utilized machine learning (ML) techniques to develop and validate a predictive model for DN, focusing on both high predictive accuracy and model interpretability. Data from 1,000 Type-2 diabetes patients, including 444 with DN and 556 without, were analyzed. Various ML algorithms, including decision trees, random forests, Extra Trees, AdaBoost, XGBoost, and LightGBM, were employed. Multiple imputation was used for handling missing data, and the Synthetic Minority Over-sampling Technique (SMOTE) addressed data imbalance. Model performance was evaluated with metrics such as accuracy, precision, recall, F1 score, specificity, and area under the curve (AUC). Explainable Machine Learning (XML) techniques like LIME and SHAP were used to enhance model transparency and interpretability. Results: XGBoost and LightGBM demonstrated superior performance, with XGBoost achieving the highest accuracy of 86.87%, a precision of 88.90%, a recall of 84.40%, an F1 score of 86.44%, and a specificity of 89.12%. LIME and SHAP analyses provided insights into the contribution of individual features to the prediction outcomes, identifying serum creatinine, C-peptide, albumin, and lipoproteins as significant predictors. Conclusion: The developed ML model not only provides a robust predictive tool for early diagnosis and risk assessment of DN but also ensures transparency and interpretability, crucial for clinical integration. By enabling early intervention and personalized treatment strategies, this model has the potential to improve patient outcomes and optimize healthcare resource utilization.