Prediction of Diabetic Peripheral Neuropathy (DPN) using Plantar Pressure Analysis and Learning Models

Diabetes Mellitus is one of the world's most common diseases. A common complication of this disease is diabetic peripheral neuropathy (DPN), characterized by the loss of sensation in different parts of the foot. Several studies reported that DPN can be patterned from the plantar pressure data of a person. In this study, the plantar pressure data gathered using the Tekscan F-Scan Research Software were statistically analyzed and used to train different machine learning classifiers. Results show that the parametric differences between the left and right foot of the volunteers are indicative of asymmetric plantar loading. The Instant Maximum Force Time (IMxFT) and contact area (CA) on the right foot exhibited multiple significance on different regions namely: medial heel (MH), M2, M4, M5, and Midfoot (MF). The parameter length of contact time (LT) also showed a significant difference in the left foot of the diabetic groups. Classification algorithms used for the right foot dataset that posed the two highest accuracies are SVM, which yielded an accuracy of 91.91% and MLP with 89.82%. For the left foot, Gaussian Process Classifier (GPC) followed immediately with 90.11% accuracy while MLP, KNN, and Random Forest classifiers registered 87.42%, 84.66%, and 75.31%, respectively. Moreover, hyperparameter tuning using grid search of the top-performing classifiers for both the left and right datasets further improved the classifier performance to 92.91% and 91.74% for the SVM and MLP, respectively.