It’s estimated that over 17,500 new spinal cord injuries occur each year in the U.S., meaning 54 people out of every million Americans experience one. Recovering from those injuries predictably isn’t easy. Patients regain the majority of function within the first six months after the injury, but any remaining loss present after a year is much more likely to become permanent. (Over 30% of spinal cord injuries result in partial or complete paralysis.)

That’s why engineers and neurosurgeons at Brown University, Rhode Island Hospital, Intel, and Micro-Leads Medical have committed to spending the next two years designing and building an Intelligent Spine Interface (ISI), which they say will record motor and sensory signals from spinal cord neurons and learn to stimulate those neurons using AI. Now, with a $6.3 million grant from the U.S. Defense Advanced Research Projects Agency (DARPA), they’re preparing to implant electrodes in a test to see if severed nerves can be made to communicate.

Their work expands on that of teams in Switzerland and at the Kentucky Spinal Cord Injury Research Center at the University of Louisville, which found that spinal cord stimulation can help restore voluntary muscle control post-injury. And it coincides with a program spearheaded by the U.S.-based Center for Sensorimotor Neural Engineering (CSNE), which aims to refine algorithms and hardware designed to restore limb function to those with spinal lesions. Yet another effort — the ByAxon project, a four-year, four-nation initiative funded by the European Commission’s Horizon 2020 Future and Emerging Technologies — will similarly prototype implants that reconnect nerves on either side of an injury.

“We know that circuits around a spinal lesion often remain active and functional,” said project lead David Borton, an assistant professor at Brown’s School of Engineering and a researcher at the University’s Carney Institute for Brain Science. “The hope is that by using information from either side of a lesion in a bidirectional way, we could make a significant impact on the treatment of spinal cord injuries. This exploratory study aims to build the toolset — the mix of hardware, software, and functional understanding of the spinal cord — to make such a system possible.”

Intel spine AI

Borton and colleagues plan to use an experimental interface to record signals traveling down the spinal cord above an injury site, and to drive electrical stimulation below the lesion while injecting signals above it using information coming up the cord. It’ll mimic natural signaling processes more closely than previous devices like it, Borton says, potentially aiding in (or even accelerating) the recovery of limb and muscle sensations below the damaged nerves.

AI and machine learning tools will play a critical role in decoding signals recorded from the spine, according to Naveen Rao, Intel corporate vice president and general manager of the AI Products Group. Intel will supply them, along with the necessary AI accelerator hardware, software, and expertise (excepting the spinal cord stimulation technology, which Micro-Leads will furnish). And scientists at the chipmaker will work with Thomas Serre, an associate professor of cognitive, linguistic and psychological sciences at Brown, to devise analogs for the spinal cord’s biological systems.

“As a Ph.D. student at Brown, I investigated how to interface the brain with machines as an application,” said Rao. “Now at Intel, we’re combining our AI expertise with Brown University’s cutting-edge medical research to help solve a critical medical problem: how to reconnect the brain and spine after a major spinal injury.”

Over the next two years, the team of researchers will recruit volunteers with spinal cord injuries to participate in physical therapy while implanted with the interface for up to 29 days. The initial focus will be on signals related to leg control for walking and standing, as well as signals having to do with bladder control (the latter of which, they note. is a top concern of people with spinal cord injuries). An external PC will decode signals in the first phase of the two-year project, but the team hopes it’ll serve as a blueprint for a fully implantable device to come.

“What’s new in this project is taking information from the spinal cord itself and to use that to drive stimulation to another part of the spinal cord,” Borton continued. “In this way, we’re taking advantage of as much intact tissue as we can, which we think could lead open the door to wider therapeutic application of spinal cord stimulation, for example, bladder control.”