A Biophysical-Model-Informed Source Separation Framework For EMG Decomposition.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Surface electromyography (sEMG) is crucial for neuromuscular diagnostics.
• Traditional blind source separation (BSS) methods for motor unit (MU) decomposition lack biophysical constraints, limiting accuracy.
• Accurate MU decomposition is vital for understanding neural drive and developing advanced human-computer interfaces.

Purpose of the Study:

• To introduce a novel Biophysical-Model-Informed Source Separation (BMISS) framework for MU decomposition.
• To integrate anatomically accurate forward EMG models into the decomposition process.
• To enable unsupervised estimation of neural drive and motor neuron properties using MRI-based anatomical data.

Main Methods:

• Developed a BMISS framework incorporating MRI-based anatomical reconstructions.
• Utilized generative modeling for direct inversion of a biophysically accurate forward EMG model.
• Employed an unsupervised learning approach for decomposition.

Main Results:

• BMISS achieved higher fidelity in motor unit estimation compared to traditional methods.
• The framework significantly reduced computational cost.
• Validated the approach in a controlled simulated setting.

Conclusions:

• BMISS offers a more accurate and computationally efficient method for MU decomposition.
• The framework enables non-invasive, personalized neuromuscular assessments.
• Potential applications include clinical diagnostics, prosthetic control, and neurorehabilitation.