A Novel and Efficient Surface Electromyography Decomposition Algorithm Using Local Spatial Information.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) decomposition into motor unit spike trains (MUSTs) is vital for neural interfaces, particularly upper limb prosthetics.
• Existing decomposition methods struggle with small, low-energy forearm muscles due to their design for larger muscles.
• Efficient and accurate decomposition is critical for advancing non-invasive neural control systems.

Purpose of the Study:

• To develop a novel, efficient, and accurate sEMG decomposition method for forearm muscles using local spatial information.
• To overcome limitations of conventional methods that are less effective for small muscles with low global energy.
• To improve the identification of low-energy motor units (MUs) for better prosthetic control.

Main Methods:

• Proposed a fast spatial spike detection method to replace iterative blind source separation (BSS) processes.
• Leveraged spatial distribution characteristics of motor unit action potentials for pre-classification and initial template creation.
• Utilized local spatial information to avoid convergence to high-energy MUs and improve detection of low-energy MUs.

Main Results:

• The novel approach successfully identified low-energy MUs in small muscles more effectively than conventional BSS algorithms.
• Achieved identification of over 40% more reliable MUs.
• Reduced processing time by 30% compared to existing methods.
• Demonstrated effectiveness on both simulated and experimental sEMG data.

Conclusions:

• The proposed method offers a significant advancement in sEMG decomposition for forearm muscles.
• This technique enhances the efficiency and accuracy of identifying motor units, crucial for non-invasive neural interfaces.
• The findings pave the way for improved control of prosthetic devices using MUST-based technology.