On-Demand Gait-Synchronous Electrical Cueing in Parkinson's Disease Using Machine Learning and Edge Computing: A...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Wearable Technology

Background:

• Parkinson's disease (PD) significantly impacts mobility, causing gait impairment and Freezing of Gait (FoG).
• Current cueing technologies show promise but are hindered by smartphone reliance for data processing, affecting usability and detection speed.
• Edge computing offers a potential solution to enhance real-time detection and cueing systems.

Purpose of the Study:

• To develop and evaluate a novel wearable system for on-demand, gait-synchronous cueing that bypasses smartphone data processing.
• To improve the usability, robustness, and reduce the detection delay of Freezing of Gait (FoG) detection systems for Parkinson's disease patients.

Main Methods:

• A wearable system integrating an inertial sensor and a stimulator was developed, utilizing edge computing for direct on-device detection and cueing.
• Machine learning algorithms classified foot motion phases as normal or FoG-affected.
• System functionality and safety were assessed in two PD patients during Freezing of Gait assessments with video-annotated references.

Main Results:

• The system achieved over 94% detection rate for foot motions and 84% accuracy in classifying them as normal or FoG-affected.
• The Freezing of Gait detection delay was primarily determined by foot-motion duration, outperforming existing sliding-window approaches.
• The developed system demonstrated adequate gait phase and FoG detection performance.

Conclusions:

• Direct on-device computing on sensor and cueing hardware enables robust, synchronized Freezing of Gait detection and cueing.
• The proposed edge computing solution enhances mobility for Parkinson's disease patients by improving system usability and reducing detection latency.
• This adaptable solution can be integrated with various sensor and cueing modalities.