An AI bridge between brain and spine helped a paralysed man walk with greater control

Gert-Jan Oskam regained more natural control of standing and walking through an electronic connection between his brain and spinal cord. The system, reported in Nature in May 2023, decoded his movement intentions and converted them into stimulation that activated his leg muscles. It helped him climb steps and adapt his movements to the ground beneath him.[2]

Oskam had an incomplete cervical spinal-cord injury from a cycling accident. Earlier spinal stimulation and rehabilitation had already enabled basic walking with a walker. The new brain connection improved when and how he could move, including walking with crutches over more complex terrain. This was a one-person clinical demonstration with walking aids.[1]

Giving an intention a route around the injury

The central engineering idea is to connect a signal that can still be produced to a movement system that can still respond. CEA’s description explains that attempted or imagined movement produces usable electrical activity in the motor cortex. Recording and decoding that activity provides commands for a spinal stimulator, which helps recruit the muscles involved in the intended movement.[3]

The decoder did more than detect a command to start walking. One component estimated which joint Oskam intended to move; another estimated the movement’s direction and size. The resulting commands adjusted stimulation. That continuous relationship helped him vary his steps instead of relying only on a preprogrammed walking sequence.[1]

Years of engineering behind a visible step

The recording technology had a separate earlier demonstration. In 2019, CEA reported that a participant with tetraplegia controlled a suspended four-limb exoskeleton using brain signals. He practised with a virtual version at home and the physical machine at Clinatec. That experiment established control of an external robot through the implanted recording system.[4]

CEA describes WIMAGINE as a 64-electrode recording device placed inside the skull, with electrodes contacting the protective membrane over the brain. Its electronics digitise the signals and send them wirelessly to an external station. Machine-learning software then estimates the deliberately imagined movement. The hardware and decoder had to work together reliably before those commands could become useful movement.[4]

Connecting this technology to spinal stimulation changed the destination of the command. The person’s own legs became the moving parts. EPFL’s team combined the brain recordings, adaptive AI and stimulation into a wireless system, bringing Oskam’s intentions into closer correspondence with the action he wanted to perform.[2]

Rehabilitation continued beyond switching it on

After 40 rehabilitation sessions, the researchers also measured improvements without stimulation, including better voluntary control of hip muscles. Oskam could use crutches independently. The team proposed that training with the bridge helped reorganise surviving neural pathways; it did not demonstrate that the injured spinal cord had simply regrown.[1]

The study also recorded an infection requiring removal of one brain implant; it was later replaced. The other implant continued functioning, allowing rehabilitation and use to resume after recovery.[1]

The practical outcome included ordinary activities such as standing with friends. That is part of the significance of the work: useful mobility consists of transitions, pauses and adaptations as well as a sequence of steps across a laboratory floor.[2]

The next stage is reaching more people

On 9 September 2026, ONWARD Medical reported that seven people had received its investigational ARC-BCI therapy after two further implants. The company pairs a brain–computer interface with an implanted spinal-stimulation system in continuing feasibility studies. The count describes participation in the development programme; it does not establish that all seven achieved Oskam’s walking results.[5]

ARC-BCI remains investigational and unavailable for commercial use. The programme is nevertheless moving beyond its original demonstration. Its next valuable contribution will be to show which people and movements benefit from the same basic idea: letting an intended action guide assistance to the body in real time.[5]

Report an error or suggest a correction