Gait feature analysis of polyneuropathy patients

Abstract

Polyneuropathy (PNP) and aging both bring changes to the walking pattern of elderly people. However, the identification methods of PNP from gait patterns were not sufficiently investigated from a technical perspective. In this study an automated classification method was developed to discriminate the neuropathic gait from both young healthy and old healthy gait using artificial neural network (ANN). A robust markerless gait detection system was employed and experiments were conducted in normal clinical conditions on 10 young, 10 old and 10 neuropathy patients. Four types of gait features, namely temporal features, kinematic joint trajectories in time domain, the Fourier transform of joint angles in frequency domain, and the symmetry indexes, were extracted. One-way analysis of variance (ANOVA) was employed as a statistical analysis tool and feature selection method. Each type of features and the selected features obtained from ANOVA were served as the input of a two-layer-feed-forward neural network separately. A twofold cross validation method with enhanced generalization was utilized to evaluate the accuracy of classification. The ground truth information for the result validation was provided by the medical experts involved in the study. The outcome of individual feature set showed that the kinematic features in time domain reached the highest classification accuracies of 94.2%, 94.8% and 94.8% for three classes, while the symmetric features yielded the lowest. Combining two sets of features can improve the performance slightly and the best performance was achieved by using the selected significant features with accuracies of 96.2%, 97.0% and 96.9% respectively.