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

Area of Science:

• Endocrinology and Metabolism
• Neurology
• Data Science and Machine Learning

Background:

• Diabetic peripheral neuropathy (DPN) is a prevalent and disabling complication of type 2 diabetes mellitus (T2DM).
• DPN significantly diminishes patients' quality of life and escalates healthcare costs.
• Timely prediction and intervention are essential for mitigating the adverse effects of DPN.

Purpose of the Study:

• To develop and validate a machine learning-based predictive model for early detection of DPN in T2DM patients.
• To identify key clinical and biochemical predictors associated with DPN risk.
• To compare the performance of various machine learning algorithms for DPN risk stratification.

Main Methods:

• A cohort of 1,544 T2DM patients was analyzed, split into training (n=1,082) and testing (n=462) sets.
• Boruta and LASSO algorithms were employed for feature selection, identifying eight key predictors.
• Nine machine learning models were trained and evaluated for DPN risk prediction.

Main Results:

• Stochastic Gradient Boosting (SGBT) achieved the highest predictive performance, with an AUC of 0.933 in training and 0.811 in testing.
• Key predictors identified include diabetes duration, uric acid, HbA1c, NLR, smoking status, SCR, LDH, and hypertension.
• SHAP analysis confirmed the clinical significance of diabetes duration and HbA1c in DPN prediction.

Conclusions:

• The developed SGBT model offers a robust tool for the early prediction of DPN in T2DM.
• This predictive model can support the development of enhanced prevention and management strategies for DPN.
• The findings underscore the utility of machine learning in identifying individuals at high risk for diabetic complications.