Compressed spike-triggered averaging in iterative decomposition of surface EMG

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor unit activity analysis is vital for diagnosing and managing neuromuscular disorders.
• Current high-density surface electromyography (HDsEMG) decomposition methods struggle with complete motor unit separation.

Purpose of the Study:

• To develop and validate a novel peel-off approach for automatic HDsEMG decomposition.
• To improve the accuracy and completeness of motor unit action potential (MUAP) train extraction.

Main Methods:

• A peel-off algorithm based on Fast Independent Component Analysis (FastICA) was developed.
• Techniques including Principal Component Analysis (PCA) and spike-triggered averaging were used for noise reduction and MUAP estimation.
• Motor unit spike trains were identified using density-based clustering, with a reliability measure to discard inaccurate estimates.

Main Results:

• The peel-off algorithm demonstrated superior performance over a standard FastICA approach in identifying motor units across various noise levels.
• Total recall increased by up to 33 percentage points and total precision by up to 24 percentage points when accounting for unidentified and duplicate motor units, respectively.
• Validation on experimental data showed a matched recall of 97% and precision of 85% compared to the reference algorithm.

Conclusions:

• The proposed peel-off approach significantly enhances the performance of HDsEMG decomposition.
• This advancement represents a critical step towards achieving complete motor unit decomposition and extracting valuable clinical information.