Independent component analysis based algorithms for high-density electromyogram decomposition: Experimental...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Motor unit firing activity is crucial for understanding neural control of skeletal muscles.
• Reliable extraction of motor unit activity from surface electromyography (EMG) signals remains a significant challenge in signal processing.
• High-density EMG offers a rich source of data for detailed motor unit analysis.

Purpose of the Study:

• To quantify and compare the performance of three independent component analysis (ICA)-based decomposition algorithms: Infomax, FastICA, and RobustICA.
• To evaluate the accuracy of these algorithms in decomposing high-density EMG signals from upper extremity muscles.
• To assess the agreement in motor unit discharge timings and the number of concurrently identified motor units across different algorithms.

Main Methods:

• Utilized high-density EMG signals recorded from biceps brachii and extensor digitorum communis muscles at various contraction levels.
• Applied three ICA decomposition algorithms: Infomax, FastICA, and RobustICA.
• Evaluated decomposition performance using the separation index (silhouette distance) and rate of agreement (RoA) metrics, focusing on discharge timing and common motor unit identification.

Main Results:

• Achieved a high rate of agreement (80%-90%) between the different ICA algorithms, consistent across varying contraction levels.
• RobustICA demonstrated a tendency for higher RoA with other algorithms, particularly Infomax.
• FastICA and Infomax algorithms tended to identify a greater number of common motor units concurrently.

Conclusions:

• The study provides valuable insights into the utility of different ICA algorithms for motor unit decomposition from upper extremity EMG signals.
• High agreement across algorithms suggests robustness in motor unit activity extraction, even with varying contraction intensities.
• Findings aid in selecting appropriate algorithms and filter criteria for accurate motor unit analysis in EMG signal processing.