Multi-Day Reliability of High-Density Electromyography for Continuous Control of Assistive Devices

Area of Science:

• Biomedical Engineering
• Neurorehabilitation
• Human-Computer Interaction

Background:

• Chronic hand deficits post-stroke significantly impair daily tasks and quality of life.
• Assistive devices offer potential benefits but require precise user input for complex control.
• High-density surface electromyography (HD-sEMG) is a promising input method but faces challenges with day-to-day signal variability.

Purpose of the Study:

• To assess the multi-day reliability of high-density surface electromyography (HD-sEMG) control signals.
• To determine if HD-sEMG based models can consistently predict user force output across separate sessions.
• To evaluate the feasibility of using HD-sEMG for user input in assistive devices without frequent recalibration.

Main Methods:

• Eight neurotypical subjects performed isometric finger flexion and extension tasks.
• High-density surface electromyography (HD-sEMG) data (192 channels) and fingertip force were recorded over three sessions.
• Models correlating motor unit firing rate to force were developed from session 1 data and applied to sessions 2 and 3.

Main Results:

• Models predicting fingertip force from HD-sEMG showed modest increases in error across sessions.
• Average prediction error was 11.2% MVF (session 1), 14.4% MVF (session 2), and 14.6% MVF (session 3).
• Flexion direction showed a significant, yet small, increase in model fit error in later sessions.

Conclusions:

• HD-sEMG based motor unit firing models demonstrate acceptable multi-day reliability for controlling assistive devices.
• This approach may allow for consistent user input for hand function assistance without daily recalibration.
• The findings support the clinical relevance of HD-sEMG for enabling independent assistance to improve hand function in individuals post-stroke.