A paralyzed man walked again with a digital bridge

Researchers at EPFL and Lausanne University Hospital connected a brain implant to a spinal stimulator with software, and a man paralyzed twelve years earlier walked again.

A cord that stopped carrying the message

Walking is produced in the lumbosacral spinal cord, the lowest segment of the cord, under instruction from the brain. A spinal cord injury cuts the line between the two. The region that generates stepping is often intact below the lesion, and the cortex above it still forms the intention to move, but the signal no longer arrives. That is the problem the Lausanne group set out to solve by other means than repairing the cord itself.

The participant in the 2023 Nature paper was a 38-year-old man who had sustained an incomplete cervical spinal cord injury at the fifth and sixth cervical vertebrae in a cycling accident ten years before he enrolled. The injury was incomplete, meaning some nerve fibres still crossed the lesion, and the paper describes his condition as chronic tetraplegia, with all four limbs affected.

What stimulation alone had already done

The brain implant was not the first step. In 2018, Wagner, Grégoire Courtine, Bloch and colleagues at EPFL and Lausanne University Hospital reported in Nature that targeted epidural electrical stimulation of the lumbosacral cord, delivered in trains timed to the intended movement, re-established control of paralysed leg muscles during overground walking within one week in three people injured more than four years earlier. After a few months of training, the participants regained some voluntary control of those muscles with the stimulation switched off.

The man in this story was enrolled in that programme, the STIMO trial, and completed five months of neurorehabilitation supported by spinal stimulation. It allowed him to step with a front-wheel walker. He then used the stimulation at home for approximately three years. Over that period his neurological recovery reached a plateau, which the authors give as his reason for entering the next trial.

A second paper from the same group in 2022, in Nature Medicine, described a redesigned electrode paddle that targeted the dorsal roots serving the legs and trunk. In three people with complete sensorimotor paralysis, activity-specific stimulation programmes enabled standing, walking, cycling, swimming and trunk control within a single day. The stimulation worked. What it lacked was a way for the person to steer it with intention rather than with preset programmes and external triggers.

Two implants and a decoder

The brain-spine interface, reported by Lorach and colleagues in May 2023, closed that gap. Two cortical implants, each an eight-by-eight grid of 64 electrodes housed in a 50-millimetre titanium case the thickness of the skull, were placed over the sensorimotor cortex, outside the dura, to record electrocorticographic signals. The implants were built by the French research agency CEA. A processing unit decoded the recorded signals into predicted movement intentions and translated those predictions, wirelessly and in real time, into changes in the epidural stimulation delivered through a paddle lead positioned over the dorsal root entry zones of the lumbosacral cord. The trial is registered as STIMO-BSI, NCT04632290.

The stimulation side rests on a principle the group had established earlier: electrodes positioned over the dorsal root entry zones can steer electric fields toward particular pools of motor neurons, which allows extension and flexion to be regulated at each joint. The team configured a library of stimulation programmes, each a combination of anodes, cathodes, frequencies and amplitudes, that mobilized the hip, knee and ankle on both sides. The decoder's job was to select and grade those programmes from the cortical signal.

The authors report that the interface could be calibrated within a few minutes. In confusion matrices, the decoder identified the intended joint with an accuracy of 74 percent, plus or minus 7 percent standard error, and the stimulation activated the targeted muscle group with an accuracy of 83 percent, plus or minus 6. After five minutes of calibration, the participant's hip flexor muscle activity during attempted hip flexion was fivefold higher with the interface than without it.

Standing, walking, stairs

The study comprised functional assessments before implantation, the surgery, a six-week period in which decoders were calibrated and stimulation libraries assembled, and a 15-week programme of neurorehabilitation with physiotherapists, 40 sessions in total. The participant used the system at home. The authors report that signal quality and decoding accuracy during walking remained globally unchanged over nearly one year of use, and that the system allowed him to stand, walk, climb stairs and cross complex terrain under what he described as natural control of his legs. Although the decoders were recalibrated from time to time, the authors report that the cortical features they relied on remained stable, and that the depth of their modulation increased progressively, which they read as a gradual improvement in the participant's ability to drive the system.

In the EPFL and CHUV press release of May 24, 2023, the participant, Gert-Jan, then 40, said that standing at a bar with friends was a simple pleasure that represented a significant change in his life. Courtine described the system as a wireless interface between the brain and the spinal cord that transforms thought into action.

Recovery with the bridge switched off

The second finding is the one the authors treat as most consequential. Over the two trials, the participant's score on the Walking Index for Spinal Cord Injury II rose from 6 before STIMO to 16 after STIMO-BSI. He gained four points on the light-touch sensory score. The six-minute walk test, weight-bearing capacity, timed up-and-go and Berg Balance Scale all improved. After neurorehabilitation supported by the interface, he could walk more than ten metres with crutches without any assistance and without stimulation. The authors interpret this as neurological recovery promoted by the training, with the interface supplying the link between intention and movement during rehabilitation.

What the story does and does not show

The 2023 paper describes one participant. His injury was incomplete, and he had already completed three years of stimulation-supported training before the brain implants were placed, so the gains reported sit on top of that earlier recovery and cannot be separated from it by this design. There is no control group. Walking took place with crutches or a walker, not unaided. The decoder correctly identified the intended joint 74 percent of the time.

The story's headline figure of twelve years refers to the interval between the accident and the paper's publication in 2023; the paper itself places the injury ten years before enrolment. The claim that the cord rewired itself is an inference from improved scores with the system off, not a direct observation of new connections. What the paper does show is that a man who had plateaued after years of stimulation regained measurable function when the stimulation was placed under the control of his own cortical signals, and that some of that function persisted without the device. Whether it generalizes to complete injuries, or to more than one person, is the subject of the ongoing trial.

Thought to movement, with the injury bypassed by software.