Neuroprosthesis Restores Walking Ability in Patient with Complete Leg Paralysis
An epidural implant with AI-controlled spinal cord stimulation enabled a person with a cervical spinal cord injury to take natural steps and climb stairs.
Introduction
In June 2025, the world witnessed an event that experts called "the beginning of a new era in the treatment of motor disorders." The journal Nature published a study by Swiss researchers — the group of Grégoire Courtine and Jocelyne Bloch from NeuroRestore (a joint project of EPFL and Lausanne University Hospital) — on a brain-spine interface (BSI) that allowed a 38-year-old patient with an incomplete cervical spinal cord injury (C5/C6), paralyzed for ten years, to walk naturally again, climb stairs, and even navigate complex terrain.
This achievement is not just another success in neuroprosthetics. Unlike existing exoskeletons that require manual control or crutches, and unlike previous epidural stimulation systems that relied on motion detectors and gave patients the sensation of artificially initiated steps, the BSI creates a "digital bridge." It directly connects the motor cortex of the brain to the locomotion center in the spinal cord, restoring the natural connection interrupted by the injury. The patient, Gert-Jan (who allowed only his first name to be used), was able to control movements with his thoughts. "After 12 years of trying to get back on my feet, now I have learned to walk normally, naturally," he said.
Event Details and Timeline
Predecessors: From motion sensors to "thought." The Swiss work is the result of years of evolution. Initially, their spinal cord stimulation solutions relied on analysis of residual movements or wearable sensors. However, this created delay and artificiality. The real breakthrough came when the team decided to read the signal directly from the source — the brain.
Technological structure of the "digital bridge." The BSI system consists of two implants:
- Cortical implant (WIMAGINE): Two 64-electrode arrays (8x8) 50 mm in diameter, housed in a titanium casing as thick as the skull. The device is placed over the dura mater (epidurally), without penetrating the brain tissue, ensuring long-term signal stability and minimizing damage. The implant reads the electrocorticogram (ECoG) from the sensorimotor cortex area responsible for leg movement.
- Spinal implant: An electrode implanted epidurally in the lumbosacral region of the spinal cord, which is responsible for generating stepping movements.
How it works: The implant in the head captures patterns of brain activity associated with the intention to move a leg. An artificial intelligence (AI algorithm) decodes these signals in real time. They are then wirelessly transmitted to a portable computer (worn in a backpack), where they are converted into commands for the spinal stimulator. The stimulator sends precise electrical impulses to the corresponding areas of the spinal cord, causing leg muscles to contract. The entire loop — from thought to movement — takes fractions of a second.
Control quality and sensory feedback. The Swiss development was unique for a long time as it only provided motor control. However, in April 2026, researchers from the University of California, Irvine (UC Irvine) and Caltech took the next step: they created a bidirectional interface (BDBCI) that not only controls an exoskeleton but also provides feedback. A study participant (50 years old) using the system not only controlled steps but also "felt" the ground contact, thanks to stimulation of the somatosensory cortex. In a step-counting test, her accuracy reached 93%.
In parallel, Chinese scientists demonstrated a system based on an EEG cap (without brain implantation) for a post-stroke patient, and in Singapore, a clinical trial of AI-controlled transcutaneous (non-invasive) spinal cord stimulation (tSCS) was launched.
Impact and Significance
For science and neurorehabilitation. The Swiss work is a conceptual proof: a "digital bridge" between the brain and spinal cord is possible and effective. Patient Gert-Jan demonstrated natural movements, the ability to navigate obstacles, and climb stairs — tasks unattainable for users of classic exoskeletons.
The phenomenon of neuroplasticity. The most unexpected finding: even with the BSI turned off, the patient continued to show improvements. His walking index score (WISCI II) increased from 6 to 16 after 40 training sessions with the interface. This means the interface promotes long-term neuroplasticity — restoration of the patient's own neural pathways.
For patients. For millions of people with spinal cord injuries (about 300,000 patients in the US alone), this technology offers a radical improvement in quality of life. As UC Irvine neurology professor An Do notes: "Restoring the ability to walk is among the highest rehabilitation priorities for paralyzed individuals."
Limitations. Currently, the technology has been tested on a small number of patients (before 2025, only one; the Swiss group is preparing a study with three participants) and requires implantation of electrodes in both the brain and the spine. Additionally, the equipment is still bulky (wearable computer).
Reactions of Key Players
Swiss team (EPFL/CHUV). Grégoire Courtine and Jocelyne Bloch have already begun recruiting the next participants. Their goal is commercialization and miniaturization of the system to eliminate the "backpack" with the computer.
US scientists (UC Irvine/Nature). The American group focused on bidirectionality and portability, using an embedded computing system (3 microcontrollers at 48 MHz), eliminating the need for an external computer.
Scientific community. Reactions range from enthusiasm to conservative restraint. Neurosurgery professor Michael Fehlings (University of Toronto) called the bioengineering "truly outstanding" but pointed to the small number of patients in the study. Dr. Daniel Rubin (Harvard/MGH) noted that due to the non-invasive recording method, the signal from the brain surface may not be "clean" enough to control, for example, fine hand motor skills.
Alternative approaches (China, Italy). While debates about invasiveness continue, teams in China and Italy are actively promoting non-invasive EEG caps. Although signal quality is lower, the absence of surgical risks makes these technologies potentially more accessible.
Forecast and Conclusions
What we have as of May 2026. We stand on the threshold of an era of "thought-readable" prostheses. Invasive BSI interfaces provide high accuracy and naturalness of movement previously unattainable. Bidirectional BDBCI systems restore the lost sense of touch. In parallel, minimally invasive solutions are being developed for less severe patients.
Main challenges:
- Miniaturization and wireless communication. Current prototypes still have external units. The transition to fully implantable systems (as pursued at UC Irvine) is a matter of the next few years.
- Cost and accessibility. Electrode implantation, lengthy calibration, and AI training are expensive procedures. Will healthcare systems be able to afford them?
- Individuality of injuries. Most successes have been achieved with incomplete injuries (where some connections are preserved). Will the system work for patients with a complete anatomical transection of the spinal cord? Theoretically, yes, if the neurons below the injury site are alive, but this remains to be proven.
Forecast for 2030. Third-generation neuroprostheses will likely enter clinical practice as a standard of rehabilitation for certain categories of spinal cord injuries. They will become less invasive (possibly using ultra-thin carbon/diamond electrodes operating at the level of individual neurons) and will be characteristic of leading neurosurgical centers worldwide.
Conclusion. The brain-spine interface that restores a patient's ability to walk is not a science fiction plot but a documented fact of 2025-2026. Of course, the technology has a long way to go from laboratory prototypes to a mass-market product. But the main thing has happened: paralysis is no longer irreversible. The bridge between thought and movement has been built.
— Editorial Team
