Decoding Diabetic Neuropathy: Tap into Postural Sway for Optimal Feature Detection based on Machine Learning Approach

Diabetic neuropathy, a condition resulting from diabetes-induced nerve damage, affects various parts of the body, leading to complications. This study aimed to classify individuals into healthy, diabetic, or neurologically ill categories by analyzing changes in their center of pressure (COP) data collected from standing on a Wii Balance Board. Machine learning (ML)-based classification models were employed to identify factors influencing diabetic neuropathy and enable early intervention. Quantitative characteristics and continuous information extracted from COP data were utilized for classification. Explainable AI techniques such as Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP) were employed to interpret classifier decisions. Various classification methods including Random Forest, Logistic Regression, Naive Bayes, Gradient Boosting, and K-Nearest Neighbors were evaluated, with Random Forest and KNN yielding the best performance. These models achieved high accuracy (96-97%), precision, recall, F1 scores, and AUC-ROC values (0.99). The study demonstrates the promising potential of machine learning in detecting postural sway disorders from COP data and underscores the importance of explainable AI techniques for understanding classification decisions. This research provides empirical evidence supporting the feasibility and utility of classification using time series data in postural test analysis.