Machine Learning-Based Prediction Model for Type 2 Diabetic Peripheral Neuropathy: Role of Bilateral Brachial-Ankle...

Area of Science:

• Endocrinology and Metabolism
• Neurology
• Artificial Intelligence in Medicine

Background:

• Diabetic peripheral neuropathy (DPN) is a common complication of type 2 diabetes mellitus (T2DM).
• Early prediction and risk stratification of DPN are crucial for effective management.
• Novel predictive markers and machine learning approaches are needed to improve DPN detection.

Purpose of the Study:

• To establish a machine learning model for predicting DPN in T2DM patients.
• To investigate the predictive roles of bilateral brachial-ankle pulse wave velocity (baPWV) and anthropometric indices in DPN.
• To identify key risk and protective factors for DPN using advanced analytical methods.

Main Methods:

• Retrospective analysis of clinical data from 966 T2DM patients.
• Classification of patients into DPN and non-DPN groups based on nerve conduction tests.
• Utilizing BorutaShap for feature selection and establishing nine machine learning models.
• Employing SHAP (SHapley Additive exPlanations) for model interpretability and factor importance ranking.

Main Results:

• The Support Vector Machine (SVM) model demonstrated the best performance in DPN prediction (AUC 0.82).
• Left brachial-ankle pulse wave velocity (LBAPWV) emerged as the most significant predictor (SHAP=0.70), followed by gender, Glucose 0min, fT3, diabetes duration, and hip circumference.
• Identified risk factors include LBAPWV, gender, and diabetes duration; protective factors include fT3 and hip circumference.

Conclusions:

• Machine learning models, particularly SVM, offer robust prediction of DPN in T2DM.
• Asymmetric predictive power was observed between LBAPWV and RBAPWV, with LBAPWV being more influential.
• Findings enhance DPN risk stratification by highlighting key predictors like LBAPWV and protective factors such as hip circumference.