Longitudinal high-density EMG classification: Case study in a glenohumeral TMR subject

Area of Science:

• Biomedical Engineering
• Neuroprosthetics
• Rehabilitation Medicine

Background:

• Targeted muscle reinnervation (TMR) offers advanced prosthetic control for upper-limb amputees but is limited by the number of reinnervation sites.
• Pattern recognition algorithms using electromyography (EMG) signals show promise in overcoming these limitations.
• High-density EMG (HD-EMG) recordings have demonstrated higher classification accuracy than conventional methods in previous myocontrol studies.

Observation:

• This case study longitudinally assessed HD-EMG classification in a glenohumeral amputee over 17 months post-TMR surgery.
• Five experimental sessions were conducted during a standard rehabilitation protocol, including therapy and myoprosthesis use.
• EMG activity was first detected in the second session, with classification accuracy reaching 76% by the third session and approximately 95% in the final two sessions.

Findings:

• Classification accuracy for 12 distinct EMG signal classes improved substantially throughout the rehabilitation period.
• Short-term (1-hour interval) training/testing sets initially yielded lower accuracy (32%), but improved with prosthesis usage (67% in session 5).
• The study demonstrates a clear improvement in EMG classification accuracy correlating with the TMR rehabilitation process.

Implications:

• HD-EMG classification holds significant potential for enhancing the dexterity and functionality of myoelectric prostheses.
• Longitudinal monitoring of EMG signal quality and classification accuracy is crucial for optimizing prosthetic control post-TMR.
• This approach may enable more intuitive and sophisticated control strategies for individuals with high-level limb loss.