What Is Spinal Cord Stimulation and How Is It Related to BCI?

He Was Told He'd Never Walk Again. Twelve Years Later, He Stood Up.

In February 2022, a man named Michel Roccati stood up from his wheelchair in a research lab in Lausanne, Switzerland. He took a step. Then another. Then another.

Michel had been completely paralyzed below the chest since a motorcycle accident twelve years earlier. His doctors had told him the damage to his spinal cord was total. The signal path from brain to legs was severed. No amount of physical therapy, no drug, no surgery could reconnect what had been destroyed.

They were right about the reconnection part. Nobody fixed his spinal cord.

Instead, a team at the Swiss Federal Institute of Technology (EPFL) did something more interesting. They implanted a thin array of electrodes on the surface of Michel's spinal cord, below the injury site. These electrodes delivered precisely timed pulses of electricity directly to the neural circuits responsible for leg movement. Those circuits, it turns out, were still there. Still intact. Still capable of producing the complex sequence of muscle activations needed for walking. They just needed someone to wake them up.

And here's where it gets truly remarkable. A separate implant, placed on Michel's brain, read his intention to move. When he thought "walk," the brain implant decoded that signal and told the spinal stimulator exactly which pattern to fire, and when.

His brain was still giving the orders. The damaged spinal cord just wasn't delivering them. So the researchers built a digital postal service around the break.

This is the story of spinal cord stimulation and brain-computer interfaces. And it's one of the most extraordinary convergences in modern neuroscience.

The Spinal Cord Is Not Just a Wire

Before we get into the technology, we need to correct a misconception that almost everyone carries around. Most people think of the spinal cord as a cable. A biological wire that connects the brain to the body. Signals go down, muscles move. Signals go up, you feel things. Simple.

This is wrong. Profoundly wrong. And understanding why it's wrong is the key to understanding everything that follows.

Your spinal cord is not a passive cable. It's a computer.

The lumbar region of your spinal cord, the section in your lower back, contains neural circuits called central pattern generators (CPGs) that can produce the rhythmic, alternating patterns of muscle activation needed for walking entirely on their own. These aren't simple reflexes. They're sophisticated motor programs that coordinate dozens of muscles in precise temporal sequences.

Here's the evidence that makes this real: if you sever a cat's spinal cord completely, separating the brain from the legs entirely, and then support the cat's body weight on a treadmill, the legs will still walk. The lumbar spinal circuits generate the stepping pattern without any input from the brain at all. This was first demonstrated by Thomas Graham Brown in 1911, and it completely upended how scientists thought about movement.

Humans have these circuits too. Your brain doesn't micromanage every muscle contraction when you walk. It sends high-level commands, things like "go forward," "speed up," "turn left," and the spinal cord handles the details. Think of it like a CEO and a factory floor. The CEO (your brain) sets the strategy. The factory floor (your spinal cord) runs the assembly line.

So when a spinal cord injury cuts the connection between brain and body, the factory floor doesn't disappear. The workers are still there. The machinery still works. The assembly line is still perfectly capable of running. It just stopped receiving orders from headquarters.

Spinal cord stimulation is, in essence, a way to call the factory floor directly.

What Is Spinal Cord Stimulation?

Spinal cord stimulation (SCS) involves surgically implanting electrodes near the spinal cord and using them to deliver controlled electrical pulses. The concept has been around since 1967, when neurosurgeon Norman Shealy implanted the first spinal cord stimulator in a patient with chronic pain.

For most of its history, SCS has been a pain management tool. And it's a big one. Over 50,000 spinal cord stimulators are implanted every year in the United States alone. The basic idea is straightforward: electrical pulses delivered to the dorsal columns of the spinal cord interfere with pain signals traveling up to the brain. It's based on the gate control theory of pain, proposed by Melzack and Wall in 1965, which suggests that non-painful input can close the "gate" to painful input at the spinal level.

But the version of SCS that's making paralyzed people walk again is a fundamentally different beast. It's called epidural electrical stimulation (EES), and it works on completely different principles.

From Pain Management to Making People Walk

The discovery that spinal cord stimulation could do more than manage pain happened partly by accident, as the best discoveries often do.

In the early 2000s, researcher Susan Harkema at the University of Louisville was studying the effects of epidural stimulation on a young man named Rob Summers, who had been paralyzed from the chest down after a hit-and-run accident. The original goal was to study how stimulation affected blood pressure regulation and other autonomic functions.

Then Rob Summers tried to move his toes. And they moved.

This was supposed to be impossible. Rob had a clinically complete spinal cord injury. There should have been zero voluntary motor signal getting through. But with the stimulator active, providing a background hum of electrical excitation to his lumbar spinal cord, the faint traces of signal that somehow survived the injury were suddenly enough to trigger movement.

The scientific paper, published in The Lancet in 2011, sent shockwaves through the neuroscience community. It suggested two things that nobody had dared to believe simultaneously. First, that "complete" spinal cord injuries might not actually be 100% complete. Some neural fibers, too few to produce movement on their own, might survive. Second, that epidural stimulation could amplify those surviving connections enough to restore voluntary control.

In the years that followed, the field accelerated dramatically.

That 2022 result from EPFL deserves special attention. Previous approaches required weeks or months of calibration before patients could walk with stimulation. The EPFL team's targeted approach, using electrode arrays designed to activate specific dorsal root entry zones in a precise spatial and temporal pattern, enabled standing and walking within a single day of implantation. The difference was in the specificity. Instead of bathing the whole spinal cord in electrical noise, they were delivering the right signal to the right neurons at the right time.

And then, in 2023, they connected it to the brain.

The Brain-Spine Interface: Where SCS Meets BCI

This is where two fields that developed mostly in isolation suddenly snapped together like puzzle pieces that nobody realized fit.

A [brain-computer interface](/guides/what-is-bci-brain-computer-interface) reads electrical signals from the brain and translates them into digital commands. Spinal cord stimulation delivers electrical patterns to the spinal cord to activate motor circuits. Put them together and you get something that the EPFL team calls a brain-spine interface (BSI): a system where the brain's own movement intentions directly control spinal stimulation in real time.

Here's how their system works, step by step.

A cortical implant (two small electrode arrays placed on the surface of the motor cortex) records neural activity while the person thinks about walking. Those signals are transmitted wirelessly to a portable processor. Machine learning algorithms decode the signals into predictions about which leg the person intends to move, and when. Those predictions are converted into stimulation commands and sent to the epidural electrode array on the lumbar spinal cord. The spinal stimulator fires the correct spatial and temporal pattern to activate the relevant motor neuron pools. The person's legs move.

The entire loop, from thought to movement, takes about 100 milliseconds.

The participant in the EPFL BSI study, a man named Gert-Jan Oskam who had been paralyzed for over a decade, described the experience in terms that should give you chills. He said the movement felt natural. Not like a machine was moving his legs, but like he was moving them himself. Because in a very real sense, he was. His motor cortex was generating the movement commands. The BSI was just ensuring those commands reached their destination.

The Neuroscience That Makes It Work

Why does any of this work at all? Why can you electrically stimulate the spinal cord and get coordinated walking instead of random muscle twitching?

The answer goes back to those central pattern generators. The motor circuits in the lumbar spinal cord are organized in a very specific architecture. Sensory neurons enter through the dorsal roots. Motor neurons exit through the ventral roots. Interneurons connect the two and create the oscillating circuits that generate rhythmic patterns.

Epidural electrical stimulation works primarily by activating the large-diameter sensory fibers in the dorsal roots. These fibers carry proprioceptive information (where your limbs are in space) and are physically the easiest neurons to recruit with external electrical fields because of their size and position. When you stimulate them, you're essentially injecting artificial sensory input into the spinal circuits, telling them that the leg is in a particular position and that it's time to fire the next phase of the step cycle.

The spinal circuits take that input and do what they've always been wired to do: produce coordinated motor output. The central pattern generators activate in sequence, the right muscles contract at the right times, and the leg moves.

It's a bit like push-starting a car. The engine (the spinal circuit) works fine. It just needs a kick to get going.

Now, tonic stimulation (constant, unchanging pulses) can activate these circuits in a general way. This is what the early experiments used, and it worked, but the movement was coarse. The breakthrough came when researchers figured out how to pattern the stimulation. By changing which electrodes are active and when, they could selectively recruit different motor neuron pools at different phases of the gait cycle. Left hip flexors fire at one moment. Right knee extensors fire 200 milliseconds later. The stimulation pattern mirrors the natural activation sequence that the brain would normally send through the spinal cord.

And when you let the brain itself control that pattern in real time through a BCI, the movement becomes fluid enough to look almost normal.

Brainwave data, captured at 256Hz across 8 channels, processed on-device. The Crown's open SDKs let developers build brain-responsive applications.

Explore the Crown

What the BCI Side of the Equation Actually Does

Let's zoom in on the brain-computer interface component, because it's doing something genuinely sophisticated.

The motor cortex, the strip of neural tissue along the top of your brain, is organized in a map. Different regions control different body parts. The area that controls legs sits at the top, along the medial wall of the brain (the inside surface between the two hemispheres). When you think about moving your right leg, neurons in the left medial motor cortex fire in a specific pattern.

In the EPFL brain-spine interface, two electrocorticography (ECoG) arrays sit on the surface of the motor cortex over the leg area. ECoG is a middle ground between invasive and non-invasive approaches. The electrodes are placed under the skull but on top of the brain's surface (the dura mater), not pushed into the tissue like the microelectrode arrays used in BrainGate or Neuralink. This gives much better signal quality than EEG while being less risky than penetrating electrodes.

The decoder, the algorithm that translates brain signals into stimulation commands, uses machine learning trained on the relationship between cortical activity and the person's stated movement intentions. During calibration, the participant thinks about specific movements while the algorithm learns which neural patterns correspond to which actions. Over time, the decoder becomes increasingly accurate, and the participant's brain adapts to the decoder's interpretation in a kind of neural-machine co-learning process.

This bi-directional adaptation is one of the most fascinating aspects of the entire system. The machine learns the brain, and the brain learns the machine. After weeks of use, Gert-Jan Oskam's brain had reorganized its neural patterns to produce cleaner, more distinct signals that the decoder could interpret more easily. Neuroplasticity, the brain's ability to rewire itself, was working in favor of the technology.

The Bigger Picture: A Growing Ecosystem

The brain-spine interface represents one point in a much larger space of brain-computer interface applications. The core principle, reading neural signals and using them to control something, applies whether you're controlling a spinal stimulator, a robotic arm, a computer cursor, or a music playlist.

The signal decoding challenges are fundamentally similar across the spectrum. Whether you're reading EEG through the skull or ECoG from the cortical surface, you're dealing with noisy biological signals that need to be cleaned, processed, decomposed into features, and classified by machine learning models. The BCI pipeline is the same. The resolution differs, but the engineering problems are cousins.

This is why consumer BCIs matter even if you personally don't need a spinal cord stimulator. Every developer who builds a brain-controlled application with the Neurosity Crown's SDK, every researcher who publishes a paper using consumer EEG data, every algorithm that learns to decode cognitive states from scalp-recorded signals, these all contribute to the broader field. The line between consumer neurotechnology and clinical neurotechnology is blurring, and it should be.

The Challenges That Remain

Let's be honest about what isn't solved yet, because the honest parts are just as important as the hopeful ones.

Surgical risk. Both the cortical implant and the spinal electrode array require surgery. The cortical component requires a craniotomy (removing a section of skull). The spinal component requires a laminectomy (accessing the spinal canal). These are real surgeries with real risks.

Hardware longevity. Implanted electronics face a hostile biological environment. The body's immune system treats implants as foreign invaders and forms scar tissue around them. Electrodes can shift position. Wires can break. Current systems need replacement or repair on timescales of years, not decades.

Decoding complexity. Walking on a flat, predictable surface in a lab is one thing. Navigating the chaos of everyday life, with stairs, curbs, uneven terrain, crowds, is vastly more complex. The decoder needs to handle a far wider range of movement intentions and environmental contexts than current systems support.

Energy. Brain implants, wireless transmitters, and spinal stimulators all need power. Battery technology limits how long the system can operate between charges. Researchers are exploring wireless power transfer and ultra-low-power electronics, but it remains a bottleneck.

Scale. As of 2026, the number of people who have received a combined BCI plus spinal stimulation system is in the single digits. Moving from "it works in the lab with a team of engineers supporting one participant" to "it works at home for thousands of patients" is an enormous engineering, regulatory, and manufacturing challenge.

Why This Matters Beyond Paralysis

Here's the "I had no idea" moment.

The combination of brain signal decoding and targeted electrical stimulation of neural tissue isn't just a paralysis treatment. It's a new category of medicine.

Think about what the brain-spine interface actually demonstrates at a fundamental level. It proves that you can read intention from one part of the nervous system, transmit it digitally, and deliver it as electrical stimulation to another part of the nervous system in real time. The spinal cord was the first target because the motor circuits there are relatively well understood and the clinical need is obvious.

But the same principle could apply anywhere the nervous system is damaged or degraded.

Stroke rehabilitation, where brain areas controlling speech or movement have been damaged, could benefit from BCIs that reroute signals around the lesion. Peripheral nerve injuries could be bridged with decoders that read motor cortex signals and stimulate nerves downstream of the damage. Bladder control, sexual function, autonomic regulation, all of these are controlled by spinal circuits that could theoretically be targeted with the same approach.

And then there's the non-clinical frontier. If you can decode the brain's intentions and deliver precise electrical stimulation to spinal circuits, you've essentially proved that the nervous system can be augmented with digital hardware. Today it's restoring lost function. Tomorrow, well, the line between restoration and enhancement has always been blurrier than we'd like to admit.

Where Consumer BCI Fits In

You might be reading all of this and thinking: this is amazing, but it's all implanted devices and surgical procedures. What does it have to do with a non-invasive EEG headset?

More than you'd think.

The fundamental science of brain signal decoding doesn't care whether the signal comes from an implanted electrode or a sensor sitting on your scalp. The math is the same. The machine learning is the same. The signal processing pipeline is the same. The signal quality differs, absolutely, but the intellectual infrastructure is shared.

Consumer BCIs like the Neurosity Crown serve the ecosystem in several concrete ways. They make brain signal data accessible to a global developer community. They provide platforms for testing decoding algorithms at scale. They let researchers prototype BCI paradigms without the cost and regulatory burden of implanted systems. And they demonstrate that brain-computer interfaces aren't just a clinical curiosity. They're a computing platform.

The Crown's 8 EEG channels at CP3, C3, F5, PO3, PO4, F6, C4, and CP4 cover the motor cortex, the frontal cortex, and the parietal and occipital regions. The same motor cortex signals that the EPFL brain-spine interface reads through implanted electrodes are also detectable, at lower resolution, through the Crown's C3 and C4 positions. Developers using the Crown's JavaScript or Python SDK are working with the same type of neural data, just captured non-invasively.

Every application built on consumer EEG advances the field. Every decoder trained on non-invasive data contributes knowledge that feeds back into clinical systems. The ecosystem is one ecosystem.

The Next Ten Years

What happens between now and 2036?

If the current trajectory holds, and there are strong reasons to believe it will, we'll see several things converge.

Spinal cord stimulation for motor recovery will move from research labs to specialized clinical centers. The early patients will be those with incomplete injuries, where some neural fibers survive and stimulation can amplify them. As the technology proves safe and effective at scale, it will expand to complete injuries using the brain-spine interface approach.

BCI decoders will get dramatically better. The combination of more training data, better algorithms (transformer architectures are already showing up in neural decoding papers), and more capable on-device processing will shrink the gap between invasive and non-invasive BCI accuracy. This won't eliminate the advantage of implanted electrodes, but it will make non-invasive BCIs useful for an expanding range of applications.

The line between clinical BCIs and consumer BCIs will continue to blur. The same company that builds a spinal cord stimulation controller might also build a consumer focus tracker. The same algorithms that decode walking intention from motor cortex might also decode concentration states from frontal EEG.

And somewhere in all of this, a generation of developers who first touched brain data through a consumer device like the Crown will build applications that nobody has imagined yet. That's how platforms work. The people who built the first iPhone apps couldn't have predicted TikTok. The people building the first BCI apps can't predict what brain-computer interfaces will become when millions of developers are working with neural data every day.

The spinal cord stimulation story is, in a way, the most dramatic proof of concept for this entire field. It proves that the nervous system is not a sealed system. It can be read, it can be written to, and the combination of reading and writing opens possibilities that neither alone can achieve.

Michel Roccati didn't just stand up from a wheelchair in a Swiss lab. He stood up as evidence that the boundary between biology and technology is more permeable than anyone assumed. That the signals in our nervous system are not locked behind biological walls but are accessible, decodable, and actionable.

And that changes everything.