Neural Interfaces: Medical Applications and Industry

Neural interfaces — two main domains:

1. Medical applications: neurorehabilitation, motor augmentation, neural control of movement

2. Movement neuroscience: biosignal processing, neurophysiology, neural models + AI

Key signal: EMG (Electromyography)

• Records electrical activity produced by skeletal muscles
• HD-sEMG (High-Density surface EMG) — array of many electrodes, provides fine-grained spatial resolution
• Used via electrode sleeves on the forearm

What a neural interface enables:

• Neuroprosthetic control for tetraplegia (e.g., controlling a robotic arm via BCI)
• Simultaneous control of multiple functions of a bionic hand prosthesis
• Decoding attempted hand movements in paralyzed patients
• Control of >20 degrees of freedom of the human hand in real time

Targeted Muscle Reinnervation (TMR):

• Surgical technique: amputated nerve endings are rerouted to remaining chest/forearm muscles
• Enables amputees to control a prosthetic hand via EMG through habitual movements

Spinal Cord Injury (SCI) — key findings:

• Even in motor-complete paralysis, spared motor neurons can be detected and decoded
• A direct spinal cord–computer interface enabled paralyzed patients to control a virtual hand with >10 degrees of freedom (Oliveira et al., Brain 2024)

Functional Electrical Stimulation (FES):

• Electrically stimulating forearm muscles to restore hand function
• Controlled intuitively via remaining EMG signals (from contra- or ipsilateral side)
• Restores hand function in tetraplegics

PlayAgain: pediatric neuroorthosis to restore grasping in children with a paralyzed hand

• Exploits the brain's high neuroplasticity during childhood

ML/AI role in neural interfaces:

• CNN maps HD-sEMG signals → high-dimensional latent space → hand kinematics (MLP)
• Real-time decoding within neuromuscular delays (~10-15 ms)