sEMG-Based Trunk Compensation Detection in Rehabilitation Training

Area of Science:

• Rehabilitation Engineering
• Biomedical Signal Processing
• Neurorehabilitation

Background:

• Stroke survivors often exhibit trunk compensation during upper limb rehabilitation, hindering motor function recovery.
• Accurate detection of trunk compensation is crucial for optimizing rehabilitation training effectiveness.
• Existing detection methods using cameras or inertial sensors have limitations.

Purpose of the Study:

• To investigate the feasibility of a surface electromyography-based trunk compensation detection (sEMG-bTCD) method.
• To evaluate the performance of sEMG-bTCD in detecting common trunk compensations during rehabilitation exercises.

Main Methods:

• Collected surface electromyography (sEMG) signals from nine trunk muscles of healthy and stroke participants.
• Recorded sEMG during three upper limb rehabilitation tasks with and without trunk compensations (lean-forward, trunk rotation, shoulder elevation).
• Preprocessed sEMG data and extracted time-domain features, utilizing a support vector machine (SVM) classifier for detection.

Main Results:

• The sEMG-bTCD method achieved excellent detection accuracy in healthy participants (up to 100%) and good performance in stroke participants (up to 91.3%).
• Support vector machine (SVM) classifier demonstrated high accuracy, AUC, and F1 scores for all tested compensations in both groups.
• sEMG-bTCD outperformed camera and inertial sensor-based methods in detecting trunk compensations for both healthy and stroke subjects.

Conclusions:

• The surface electromyography-based trunk compensation detection (sEMG-bTCD) method is feasible and effective.
• This technology can guide stroke patients to correct compensatory postures, thereby enhancing the efficacy of upper limb rehabilitation.
• sEMG-bTCD offers a promising approach to improve rehabilitation outcomes for stroke survivors.