Making connections: Designing a new neural interface module

July 5, 2019

The 96-channel Utah Electrode Array (UAE) used to record neural information. Photo: Greger Research lab, ASU

About 300,000 individuals in the United States alone are living with a spinal cord injury, based on data gathered by the National Spinal Cord Injury Statistical Center, with nearly 18,000 new cases occurring each year. The injuries can be devastating, having a major impact on daily life, including on bowel and bladder control, the ability to walk, and on the ability to control trunk, arm, and hand movements. In addition to their high costs, spinal cord injuries also often lead to shorter life expectancy, repeat hospital stays and high costs of care.

Restoring arm and hand function has been identified as a top priority among individuals with quadriplegia, and functional electrical stimulation (FES) may offer hope to some patients, according to Autumn Bullard, a 2019 doctoral graduate in the lab of BME Associate Professor Cindy Chestek.

Bullard’s work on a new module for an existing FES system was recently published in Bioelectronic Medicine in the article, “Design and testing of a 96-channel neural interface module for the Networked Neuroprosthesis system.

In FES, small bursts of electricity are used to directly stimulate the nerve or muscle to move, providing a detour around previous signaling pathways interrupted by the injury. But current FES systems are less than intuitive for injured individuals to use. In addition, they allow few degrees of freedom of movement, and they require at least some residual movement in the shoulder or wrist in order to provide input about desired movement and to control the system.

“This makes them unusable by people who are completely paralyzed,” says Bullard. “This is the gap we’re trying to address — to develop a solution that can help patients with the most severe injuries.”

“This is the gap we’re trying to address — to develop a solution that can help patients with the most severe injuries.”Autumn Bullard

Other technology, based on brain-machine interfaces, or BMIs, can predict a user’s motor intention directly using signals from the brain. The signals are recorded by implanted electrodes, which connect to an external recording device. The recorded signals are decoded by machine learning algorithms and used to predict the patient’s intention with respect to desired, rather than actual, movement. This in turn controls the prosthetic device.

But there are shortcomings with BMIs as well. The systems are not easily portable, and they carry a risk of infection for the patients using them. To overcome the challenges, a fully implantable BMI system would be required, and this significantly raises power requirements.

“We wanted to build on this existing system to expedite the route to patient use,” says Bullard, who earned her PhD in March.

Autumn Bullard. Photo: Joseph Xu, Michigan Engineering Communications & Marketing

Bullard and collaborators developed an additional neural recording module for the NNP to record signals from the brain and to facilitate brain-controlled FES in a fully implantable system. The signals, once decoded, predict motor intent rather than actual residual movement.

Bullard built the 96-channel neural recording module from readily available, off-the-shelf components to fit within the packaging and power specifications of the existing NNP.

The team was able to extract signal power at a low-frequency band and transmit 96 channels at 2 kilo-Samples per second, consuming only 33.6 milliwatts, allowing enough remaining capacity for the other modules to function as well.

The research group also tested its module to ensure it could successfully communicate over the Controller Area Network (CAN) and receive power from the NNP power module.

Previously, Bullard and collaborators had designed a 16-channel implantable device that can record signals and transmit them to a computer for processing. This was done by extracting signal power at a low-frequency band and using those signals to predict motor intent. She used the same technique to develop the 96-channel device.

“Increasing the number of channels to 96 allows us to interface with the Utah electrode array, which also has 96 channels. This is the array used in all human brain-machine interface experiments and allows us to record the signals in the brain necessary for predicting motor intent,” she says.

The greater number of channels also means more signal-related information can be recorded, which can help increase the degrees of freedom of movement and improve the level of control of a prosthetic device.

“This is the first attempt at combining designs from two different groups,” notes Bullard, who hopes the work will be used by other groups as a guide to create additional modules to address other functions lost by spinal cord injury patients.

“By collaborating and building on an existing system instead of creating an entire system from the ground up, we accelerated the pathway to a system that can be used clinically,” she says.

“By collaborating and building on an existing system instead of creating an entire system from the ground up, we accelerated the pathway to a system that can be used clinically.” Autumn Bullard

Currently, the NNP device, with the new module, can record signals and put them in the format required by the machine learning algorithms to predict motor intent. The next step is to program the device itself to use the algorithm on the signals it has recorded.

The team also is putting together an investigational device exemption application for the U.S. Food and Drug Administration so that it can start an early feasibility study in human patients.

The pathway to clinical translation is long, and questions about safety as well as battery life still must be addressed. According to Bullard, “These questions will be factors for potential patients to consider in order to determine if the clinical benefit outweighs the risk of surgical implantation.”

Bullard and Chestek collaborated with Kevin Kilgore, P. Hunter Peckham, Alex Campean, and Brian Smith from Case Western Reserve University. The work was funded by the Craig H. Neilsen Foundation, and Bullard was awarded a National Science Foundation Graduate Research Fellowship.

Bullard et al. Bioelectronic Medicine (2019) 5:3.