Restoration of complex movement in the paralyzed upper limb

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Functional electrical stimulation (FES) aids paralyzed individuals by artificially activating muscles.
• Current FES systems are limited to simple, preprogrammed movements due to difficulties in determining complex muscle activation patterns.

Purpose of the Study:

• To employ machine learning for identifying flexible muscle stimulation patterns to evoke a wide range of multi-joint arm movements.
• To overcome limitations in current FES technology for restoring motor function.

Main Methods:

• Recorded arm kinematics and electromyographic (EMG) activity from 29 muscles in a primate model.
• Trained an artificial neural network to predict muscle activity patterns for new movements.
• Converted predicted patterns into stimulus pulses delivered to upper limb muscles in paralyzed primates.

Main Results:

• Machine learning predictions of EMG were accurate within subjects but less so across subjects.
• FES-evoked movements showed good fidelity to desired movements in some instances.
• The study discussed methods to mitigate errors in FES-evoked movements.

Conclusions:

• The developed machine learning approach can generate a virtually unlimited range of movements.
• This system has the potential to significantly expand movement repertoire for individuals with high-level paralysis.
• Further refinement is needed to improve the accuracy and reliability of across-subject predictions.