--- title: "A 4-Hour Surgery, Two Weeks to Walk: China's Brain-Spine Interface Bet" date: 2026-11-02 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-brain-spine-interface-walk-again language: en --- # A 4-Hour Surgery, Two Weeks to Walk: China's Brain-Spine Interface Bet > Chinese teams are building a "nerve bridge" across severed spinal cords — and unlike classic brain-computer interfaces, the goal is not to move a robot arm but to move the patient's own legs. Ye stood between two parallel bars in a rehabilitation room in Shanghai, six months after the accident that took the use of his legs. Nothing was attached to his head. A cap picked up the electrical chatter of his motor cortex; a chip already sitting on his spinal cord answered it; an exoskeleton around his thighs did the last mechanical work. When he thought about standing, he stood. Nobody in the room called it a brain-computer interface (脑机接口). The team at Tongji Hospital had fused three separate technologies — non-invasive EEG, spinal cord stimulation, and a lower-limb exoskeleton — and the point of the exercise was that Ye's own body did the moving. That distinction is the whole story of the brain-spine interface (脑脊接口), a branch of neurotech that China has pushed harder than almost anyone. Classic brain-computer interface (脑机接口) work reads intention and hands it to a machine. The brain-spine interface tries to bridge the break and hand it back to the patient. ## Key takeaways - A brain-spine interface (脑脊接口) decodes movement intent from the brain and converts it into timed electrical stimulation of spinal nerve roots, bypassing the injury rather than outsourcing movement to a device. - Fudan University's team completed the world's first four proof-of-concept surgeries of a "three-in-one" brain-spine interface; the first patient walked more than 5 metres on his own intent on day 14 after surgery. - A separate team in Beijing combined an implanted brain-computer interface (脑机接口) with spinal stimulation and an exoskeleton, and reported a patient moving from a complete to an incomplete spinal cord injury grade at one-year follow-up. - The limiting factor is no longer the decoder. It is surgical workflow, hospital rehabilitation capacity, and the fact that none of this is a registered medical device yet. ## What a brain-spine interface actually does The spinal cord is the highway between the brain and the limbs. When it is damaged, the road breaks: the brain still issues the command, the legs still have muscle, but nothing arrives. A brain-spine interface rebuilds the missing stretch. Electrodes on or in the motor cortex pick up the signal that means "lift the left leg." A decoder — a large model (大模型) is increasingly what teams mean here — translates it within milliseconds, and a stimulator delivers precisely timed pulses to the lumbosacral nerve roots below the injury. The muscles contract. The loop closes. The difference from a conventional implant is philosophical. A robot arm is an external appliance a patient learns to drive. A nerve bridge is internal, and — if early observations hold — partly self-effacing: several Fudan patients regained some voluntary control of paralysed muscles within two weeks even with stimulation off, hinting at rewiring rather than permanent hardware dependence. ## The four surgeries that changed the argument Fudan's Institute of Science and Technology for Brain-Inspired Intelligence made its case in early 2025. Working with Fudan's Zhongshan Hospital, then Huashan Hospital, the group led by associate researcher Jia Fumin (加福民) completed four proof-of-concept operations. The engineering claims are specific: - Brain and spinal electrodes placed in a single operation lasting about four hours. - Electrode chips roughly 1 millimetre across, replacing the older paddle electrodes around 50 millimetres wide, which cuts surgical trauma. - Lightweight decoding software running fast enough that a delay of seconds could mean a fall — the team says real-time intent decoding took about three years to solve. - Leg movement returning within 24 hours of surgery, and independent leg control and stepping within two weeks. The patient the team talks about most is a 34-year-old man from Guangdong, paralysed for two years after a fall. On day 14 he walked more than five metres under his own intent, and at his first follow-up reported warmth and tingling in his feet, leg muscle contraction, and returning bladder and bowel sensation. Fudan puts the global population paralysed by spinal cord injury at roughly 20 million; Jia's framing is blunt — two hospitals, four surgeries, a technique that appears reproducible. ## A second route: implant, stimulation, exoskeleton Beijing is running a different combination. In May 2025, a team led by Zhao Guoguang and Duan Wanru at Xuanwu Hospital implanted the "Beinao-1" (北脑一号) brain-computer interface together with a timed spinal cord stimulation system in a man paralysed from the chest down for five years. One year later, the team reported that his neurological grade had moved from a complete injury (ASIA A) to an incomplete one (ASIA C) — a shift that breaks the old assumption that a complete injury's losses are fixed. Details from that follow-up are unusually concrete: bladder emptying time fell from five minutes to two, and residual urine volume went from 190 ml to zero within six months. With braces and an exoskeleton he can kick a ball and walk with assistance. Tongji Hospital's version, performed in March 2026, is non-invasive on the brain side — EEG through a cap — which lowers the surgical barrier and the signal quality with it. Three routes, one idea. ## Why the hospital sets the pace, not the algorithm None of these systems is an approved product. Fudan's work was proof-of-concept surgery; the Beijing and Shanghai cases are clinical research. Getting from four patients to four hundred needs a device manufactured identically every time, a surgical protocol another hospital can perform, and rehabilitation measured in months per patient. That is where the money has gone. Shenfu Jianxing (神复健行), spun out of Fudan to commercialise the brain-spine interface, started building a research and production headquarters in Shanghai's Waigaoqiao district with roughly 100 million yuan (≈ US$14M / HK$109M) of planned investment, and its device entered the US FDA's breakthrough therapy designation (突破性疗法通道) — an unusual export of a Chinese neurotech asset into an American regulatory fast lane. According to Li An, chief scientist at BrainNet (脑机网), the real test for brain-spine interfaces is whether a mid-tier city hospital can reproduce a team's results without the inventors in the room. ## Honest limitations This piece leans on institutional publications from Fudan University, hospital releases and state broadcasters. Four proof-of-concept surgeries are not evidence of efficacy: no control group, no blinding, no published peer-reviewed trial in the sources used here. "Walking more than five metres" and "ASIA A to ASIA C" are single-patient outcomes, and grade improvements can reflect rehabilitation intensity as much as the device. None of these combinations holds a Class III medical device (三类医疗器械) licence in China. I have not covered cost, complication rates, or who pays to revise a device every few years. ## What readers can do now - If you or a family member has a spinal cord injury, ask the treating physician specifically whether an injury is complete or incomplete, and ask for the ASIA grade in writing — it determines which trial protocols you can even be screened for. - Look up registered studies before chasing viral videos: search the indication and "clinical trial (临床试验)" on a registry and check the recruitment status and the hospital named as the site. - Give any implant programme a time horizon: ask its one-year and three-year follow-up data, and how many centres have performed the operation independently. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-brain-spine-interface-walk-again Free to quote with attribution and a link to the original.