Machine learning-based prediction of diabetic peripheral neuropathy: model development and clinical validation

Abstract

Background: Diabetic peripheral neuropathy (DPN) is a common and debilitating complication of type 2 diabetes mellitus (T2DM), significantly impacting patients' quality of life and increasing healthcare burdens. Early prediction and intervention are critical to mitigating its impact.

Methods: This study analyzed 1,544 diabetic patients from the First Affiliated Hospital of Shandong First Medical University, who were randomly divided into a training cohort (n = 1,082) and a testing cohort (n = 462) using a 7:3 split ratio. Feature selection was performed using both Boruta and LASSO algorithms, and the intersection of the selected variables was used as the final predictor set. Eight key predictors were identified from 23 variables, including diabetes duration, uric acid, HbA1c, NLR, smoking status, SCR, LDH, and hypertension. Nine machine learning models were developed and compared for DPN risk prediction.

Results: Stochastic Gradient Boosting (SGBT) demonstrated the best performance (training AUC: 0.933, 95% CI: 0.921-0.946; testing AUC: 0.811, 95% CI: 0.776-0.843). Shapley Additive Explanations (SHAP) analysis provided interpretability, highlighting the clinical importance of diabetes duration and HbA1c among other predictors.

Conclusion: This study establishes a robust predictive tool for early DPN detection, laying the foundation for improved prevention and management strategies.

Keywords: clinical data; diabetic peripheral neuropathy; interpretable; machine learning; risk prediction model.

Figure 1

Overview of data processing and machine learning workflow. (A) Database: Participants were selected from The First Affiliated Hospital of Shandong First Medical University (2023–2024). Inclusion criteria were age ≥18 years and T2DM diagnosis. Exclusion criteria included other neuropathies, malignancies, severe infections, organ dysfunction, and metabolic disorders. The final study cohort comprised 1,544 individuals. (B) Feature selection: From an initial set of 23 variables, key variables were identified using the Boruta and LASSO methods. Selected features included diabetes duration, serum creatinine (SCR), hypertension, neutrophil-to-lymphocyte ratio (NLR), smoking status, uric acid, lactate dehydrogenase (LDH), and HbA1c. (C) Model training and testing: Nine machine learning algorithms, including Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Naive Bayes (NB), XGBoost (XGB), Stochastic Gradient Boosting Trees (SGBT), and Neural Network (NNET), were applied, with hyperparameter optimization performed using 10×10-fold cross-validation. Model evaluation metrics included Receiver Operating Characteristic (ROC), Area Under the Curve (AUC), F1-score, calibration curves, and decision curves. Model interpretation was conducted using SHAP analysis for feature importance.

Figure 2

Predictor screening results. (A) Boruta feature selection. (B) LASSO screening with lambda values indicated by dashed lines. (C) Variable trajectories in the LASSO model. (D) Common predictors identified by Boruta and LASSO, including diabetes duration, HbA1c, NLR, SCR, hypertension, smoking status, LDH, and uric acid.

Figure 3

Performance and comparison of nine predictive models. ROC curves for the training set (A) and the test set (B). Evaluation metrics for the training set (C) and the test set (D), including accuracy, sensitivity, specificity, PPV, NPV, F1 score, and kappa value. Decision Curve Analysis (DCA) for the training set (E) and the test set (F).

Figure 4

SHAP analysis for feature interpretability. (A) SHAP dendrogram of features for the SGBT model. (B) Feature importance ranking for the logistic regression model. (C) Decision plot of feature contributions to the model outputs.

Figure 5

Web-based calculator for predicting the risk of diabetic peripheral neuropathy (DPN) in patients with diabetes using the developed model. By entering values for diabetes duration, uric acid, HbA1c, NLR, smoking status, serum creatinine (SCR), lactate dehydrogenase (LDH), and hypertension, an individualized DPN risk prediction can be obtained.