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

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Huntington Disease (HD) is a genetic neurodegenerative disorder impacting motor function, characterized by involuntary movements and balance issues.
• Current gait analysis for HD severity relies on macroscopic features like stride length, neglecting detailed foot pressure data.
• Individual footstep pressure patterns offer a potentially rich, underutilized data source for assessing HD progression.

Purpose of the Study:

• To investigate the efficacy of deep learning models in classifying Huntington Disease patient severity using individual footstep pressure data.
• To explore the potential of VGG16 and similar architectures for analyzing plantar pressure dynamics in HD.
• To compare the performance of deep learning models against traditional machine learning methods for HD gait assessment.

Main Methods:

• Utilized footstep pressure sensor mat data (e.g., Zeno Walkway) to capture individual footfall pressure distributions.
• Applied deep learning techniques, specifically VGG16 and related modules, for feature extraction and classification.
• Employed image pre-processing techniques to optimize the input data for the deep learning models.
• Validated model performance against the Motor Subscale of the Unified HD Rating Scale (UHDRS) as the ground truth.

Main Results:

• A VGG16-based deep learning model achieved a classification accuracy of 89% for Huntington Disease severity.
• This accuracy surpasses that of a 3D Convolutional Neural Network (3D CNN) at 82% and Support Vector Machine (SVM) at 86.9%.
• Image pre-processing was identified as a critical factor influencing model performance.

Conclusions:

• Individual footstep pressure data, when analyzed with deep learning, provides a highly accurate method for classifying Huntington Disease severity.
• Deep learning models, particularly VGG16, demonstrate superior performance compared to traditional methods for this specific application.
• This research highlights a promising new avenue for objective, non-invasive monitoring and assessment of Huntington Disease progression.