--- title: "China's first approved brain–computer interface lets paralyzed patients grasp again — no wire through the skull" date: 2026-09-29 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-brain-tsinghua-neo-wireless language: en --- # China's first approved brain–computer interface lets paralyzed patients grasp again — no wire through the skull > In March 2026, China's drug regulator cleared the NEO system — a wireless, semi-invasive brain–computer interface (脑机接口) from Tsinghua University and Bourntech — for clinical use, the first device of its kind approved anywhere. Built with epidural electrodes and no in-body battery, it has already helped dozens of spinal-cord-injury patients regain brain-controlled hand grasp. The bigger story is what "approved" changes for the field. A man who had not moved his own hand in years picked up a cup by thinking about it. The signal did not travel down a broken spinal cord. It jumped the gap through a coin-sized implant and a wireless link. What separates this from the brain–computer interface (脑机接口) headlines of the past is that it is no longer a lab demo. In March 2026, China's National Medical Products Administration approved the NEO system for clinical use — the first such device of its kind cleared anywhere to move from trial to hospital. ## A translator that sits outside the brain NEO stands for the implantable brain–computer interface hand-motor-function compensation system. It was built by the team of Professor Hong Bo (洪波) at Tsinghua University's School of Biomedical Engineering, together with Bourntech (博睿康), a Beijing-based neurotech firm spun out of the same university lab. The design choice is the interesting part. Most famous brain–computer interfaces go fully inside the brain, threading electrodes into the cortex for the clearest signal — at the cost of long-term tissue risk. NEO takes a middle road that Professor Hong's group proposed back in 2013: semi-invasive (半侵入式). A coin-sized electrode array is placed just under the skull, on top of the dura mater, recording brain signals from the outside without touching neurons. That placement avoids the infection and electrode-drift problems of deep implants while still picking up a usable signal. The team found that epidural signals carry useful information up to about 200 Hz, and by combining frequency bands they build what they call "virtual channels" — eight physical electrodes end up resolving more than 100 signal channels. ## How the numbers actually look The figures that matter are not about transistor counts. They are about whether a paralyzed person can reliably move something. - **Over 90% grasp-decoding accuracy** from those eight electrodes, with decoding latency held to a few hundred milliseconds — fast enough to feel like intent, not a delay. - **No battery inside the body.** NEO uses near-field wireless power and communication, so there is no implanted cell to fail or require a second surgery to replace. - **A single surgery and home use.** Patients are discharged days after implantation and train at home with an external controller driving an air-powered glove or wheelchair. One patient, paralyzed from a car accident for years, learned to operate a brain-controlled exoskeleton glove and even completed a brain-controlled wheelchair test after implantation. ## From one patient to thirty-two The path to approval was unusually public. The first human implant happened in October 2023 at Beijing Xuanwu Hospital. Follow-on feasibility cases ran at Tiantan Hospital, Huashan Hospital in Shanghai, and Jiangsu Provincial People's Hospital. By December 2025, the system had been implanted in **32 patients with cervical spinal-cord injury across 11 hospitals** in a national multicenter trial. According to the disclosed trial summary, **every patient achieved brain-controlled grasping**, and the main clinical endpoint was met by 100% of participants. In 22 of those 32 cases, six months of brain training produced a measurable improvement in the patients' own autonomous hand-motor scores — an unexpected hint that the interface may do more than route around the injury. Nature listed the project among its worth-watching science events for 2025, a rare nod for a Chinese clinical neurotech effort. ## Why "semi-invasive" matters The approval matters less as a single product than as a regulatory precedent. China now has a cleared, domestically developed brain–computer interface (脑机接口) with fully local software, core hardware, and key materials. For a country watching import controls on advanced chips and instruments, a made-at-home neural implant is a sovereignty argument as much as a medical one. It also reframes the global race. US efforts like Neuralink pursue fully invasive implants; non-invasive headsets avoid surgery but read weak signals through the skull. NEO's epidural route is a distinct third path, and being first to clinical approval gives Chinese researchers a real-world dataset their competitors lack. ## What it cannot do yet Two caveats are worth saying plainly. First, the device restores a narrow function — hand grasp and basic manipulation for specific injuries. It is not a cure for spinal-cord damage, and Professor Hong has repeatedly said the work solves one concrete clinical problem, not the whole field of paralysis, stroke, or neurodegeneration. Second, long-term durability and broader indications remain open science problems. The trial shows safety and a consistent effect, but "approved" in China still means early-stage clinical use, not a device sitting in every rehab ward. Scaling to other functions — speech decoding, finer limb control — is the stated next step for a 2.0 version. ## What readers can do now - **If you work in medtech or rehabilitation**, treat NEO as a signal that semi-invasive (半侵入式) BCI is now a clinically cleared category in China — worth tracking for procurement, partnership, or competitive benchmarking against invasive and non-invasive approaches. - **If you follow neurotech investing**, watch the 2.0 roadmap (language and finer-limb decoding) and Bourntech's regulatory and commercialization steps more than the headline implant count. - **If you build with brain–computer interface (脑机接口) tech**, separate the wireless-power breakthrough (the part that removes the battery-risk surgery) from the decoding claims, and validate each independently before relying on vendor numbers. ## Honest limitations Core facts (NMPA clinical approval in March 2026, specifically 13 March per Xinhua; developer Tsinghua University's Hong Bo team with Bourntech; semi-invasive/epidural route proposed 2013; eight electrodes yielding 100+ virtual channels; over 90% grasp-decoding accuracy; hundreds-of-milliseconds latency; near-field wireless power with no in-body battery; first human implant October 2023 at Beijing Xuanwu Hospital; 32 cervical spinal-cord-injury patients across 11 hospitals by December 2025 with 100% brain-controlled grasping and 22 showing autonomous hand-function improvement; Nature's 2025 worth-watching listing) come from Chinese state and official outlets: Xinhua, China Youth Daily, China Education Daily, China Daily, and China.org.cn, citing NMPA and the developer. The "world's first approved invasive-class BCI medical device" framing follows Chinese regulator and state-media language; NEO is technically semi-invasive (半侵入式), a distinct category from fully cortical implants. The decoding-accuracy and endpoint figures are clinical-trial and vendor/developer disclosures, not independently benchmarked by a third party. No RMB amounts appear in this article, so no currency conversion applies. Analysis is current to 29 September 2026 and is not medical or investment advice. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-brain-tsinghua-neo-wireless Free to quote with attribution and a link to the original.