Deciphering Muscular Dynamics: A Dual-Attention Framework for Predicting Muscle Contraction From Activation Patterns.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Science

Background:

• Accurate muscle deformation assessment is vital for diagnosing muscle diseases like Facioscapulohumeral Dystrophy.
• Ultrasound imaging is effective but limited by device portability and wiring.
• Surface electromyography (sEMG) measures muscle activation, correlating with thickness changes.

Purpose of the Study:

• To develop a deep-learning method for inferring muscle deformation directly from sEMG signals.
• To eliminate the need for ultrasound imaging in muscle deformation monitoring.
• To enable portable, real-time muscle health assessment.

Main Methods:

• A deep learning model utilizing hierarchical self-attention and cross-attention mechanisms was employed.
• The model processed sEMG data to predict muscle deformation.
• Experiments were conducted on six healthy subjects.

Main Results:

• The developed approach accurately predicted muscle excursion from sEMG signals.
• An average precision of 0.923 ± 0.900 mm was achieved.
• The method demonstrated the feasibility of muscle deformation measurement using only sEMG.

Conclusions:

• This deep learning technique offers a portable and non-invasive method for muscle deformation monitoring.
• It integrates bioelectrical and biomechanical information for comprehensive muscle health insights.
• Potential applications include clinical diagnostics, sports science, and rehabilitation.