Advancing Technology for Probing theNervous System
Deciphering the body’s neural code to examine biomechanics, movement control, and assistive technologies.
The Challenge
Our challenge is to utilize new motor unit recording technologies to advance such theories.
Our Vision
We use the latest non-invasive technologies for analyzing motor unit properties developed by our team to study.
Altec has revisited classic theories of motor control that were developed primarily through animal and mathematical models, evoked potentials, or highly constrained movements. Tools have come a long way to allow empirical measurements of motor units in human subjects to investigate how well these theories hold up when acquired during functional voluntary movements."
Fundamental Control of Movement
Our brains initiate and coordinate actions through “highways” of cells, called neurons, that can receive, transform, and relay commands.
These neurons are responsible for transmitting information, such as commands for movement to our muscles, through electrical signals.
Unfortunately, neuromuscular diseases affect tens of thousands of individuals each year—including muscular dystrophy, cerebral palsy, and spinal cord injury, among others—to degrade this brain-body connection and, in turn, prevent our muscles from receiving commands to move or function.
Neural Processing Problem
It has long been known that a specialized sensor placed on the surface of the skin can “read” the electrical activity of underlying muscles by amplifying related neural signals. This technique is called surface electromyography (sEMG).
The sEMG signal can contain contributions from dozens or even hundreds of motor units, making it difficult to detect individual motor units by the naked eye.
For decades, scientists have sought methods to reliably identify individual motor unit content within an sEMG signal. This process has been termed sEMG decomposition.
NeuroMap
After decades of R&D, our team recently created a new architecture called NeuroMap, which can autonomously decompose an sEMG signal in real time using nothing but a PC and a non-invasive, wireless sEMG sensor.
Our system is robust, and most importantly, generalizes to most types of normal human activities and movements.
Methods
We leverage an unsupervised learning framework to detect unique motor units via extracting a series of sEMG-based features, then performing custom feature vector clustering to separate motor units. Based on the motor units found in this training step, our algorithms rapidly explore the combinatoric search space to be able to track each motor unit in real time.
Results
Our findings demonstrate the feasibility of a real-time construct for recognizing motor unit action potentials within an sEMG signal, averaging over 85% MUAP shape identification and MU tracking accuracy. Processing times exceeded real-time potential, averaging 11.7 (±10.0) ms per 31.5 ms-window across over 64,060 windows tested.
How does it work?
Our architecture solves the decomposition problem by splitting it into two steps:
Step 1
Identifying unique motor units
Step 2
Tracking each detected motor unit throughout the recorded sEMG signal
Move the slider left or right to see the recorded sEMG signal and detected motor units!
Looking to the Future
This technology unlocks a host of new possibilities, and research is already underway.
Our team is exploring the utility of NeuroMap for tackling novel ways to control a prosthesis, applications in clinical settings, and new ways of visualizing muscle performance in human movement and exercise.
