A Deep Learning-Based Approach for Gait Analysis in Huntington Disease.

Zhang S1, Poon SK2, Vuong K3, Sneddon A1, Loy CT3.

Author information

1: School of AMME, Faculty of Engineering & IT, The University of Sydney, Sydney, Australia.
2: School of Computer Science, The University of Sydney, Sydney, Australia.
3: Sydney School of Public Health, The University of Sydney, Sydney, Australia.

Abstract

Huntington Disease (HD) is a genetic neurodegenerative disease which leads to involuntary movements and impaired balance. These changes have been quantified using footstep pressure sensor mats such as Protokinetics' Zeno Walkway. Drawing from distances between recorded footsteps, patients' disease severity have been measured in terms of high level gait characteristics such as gait width and stride length. However, little attention has been paid to the pressure data collected during formation of individual footsteps. This work investigates the potential of classifying patient disease severity based on individual footstep pressure data using deep learning techniques. Using the Motor Subscale of the Unified HD Rating Scale (UHDRS) as the gold standard, our experiments showed that using VGG16 and similar modules can achieve classification accuracy of 89%. Image pre-processing are key steps for better model performance. This classification accuracy is compared to results based on 3D CNN (82%) and SVM (86.9%).