A thin sheet of flexible electrodes, smaller than half a palm, rests against the outer wall of a patient's brain. No wires pierce the tissue underneath. For a person whose spinal cord was severed by injury, that sheet has become the missing bridge between a thought and a moving hand.
This is "Beinao-1" (北脑一号), built by Beijing NeuroWei Neural Technology (芯智达), a company spun out of the Beijing Institute for Brain Disorders with municipal and Zhongguancun backing. It sits in the middle lane of the brain–computer interface (脑机接口) race — not the skull-opening implants that reach deepest into tissue, and not the scalp caps that read only faint echoes.
The middle lane of the BCI race
Most coverage of brain–computer interfaces (脑机接口) splits the field into two camps. One puts electrodes inside the brain tissue for crisp signals and real risk. The other rests sensors on the scalp, safe but blurry. Beijing's team chose a third road: a semi-invasive (半侵入式) device that slips a flexible film into the epidural space, the narrow gap just outside the brain's protective membrane.
The design logic is simple to state and hard to execute. Place the electrode close enough to catch clean neural activity, but never let it touch live tissue, so there is no wound inside the skull to inflame or scar. Power and data move wirelessly, with no plug through the skin.
The current version carries 128 channels of a flexible electrode array and, according to Beijing municipal figures, had been implanted in at least 27 people by mid-June 2026. The longest implant had passed fifteen months, and the company reports the systems had accumulated more than 70,000 hours of safe operation inside patients.
What patients have done
The team describes two trajectories. One man with a cervical spinal injury and full limb paralysis received the implant in February of the prior year and, the company says, went on to control a robotic arm, a muscle-stimulation sleeve and a computer cursor, while his upper limb showed gradual rehabilitation. A second patient, paralyzed in the chest-and-below region for five years, was reported to have shifted from a complete to an incomplete injury classification three months after implantation — a change that, if durable, rewrites the old assumption that late-stage spinal injury function is fixed.
These are the company's own accounts, relayed through state media, and they describe direction more than proof. The harder question is whether the pattern holds across many patients and many months.
From lab to a regulated trial
NeuroWei (芯智达) is a 2023-founded platform company created by the Beijing municipal government together with Zhongguancun Development Group and anchored by the Beijing Brain Science Institute (北脑所). Its registration-grade multicenter trial opened at the end of March 2026. A company executive told Xinhua's Economic Information Daily that patient enrollment was already past the halfway mark and aimed to finish within the year, with a Class III medical-device application planned for 2027.
The trial spans multiple hospitals — Tiantan, Xuanwu and Peking Union among them — and recruits adults aged 18 to 65 with cervical spinal injury and tetraplegia persisting beyond twelve months. The study runs about twenty-six weeks with five in-person visits. That structure matters: a device is only a therapy once it can be tested the same way everywhere.
The same institute is already building the next step. "Beinao-2" (北脑二号) pushes to a 512-channel wireless fully implantable system and, per Beijing municipal disclosures, has entered large-animal testing with human validation targeted for later in 2026. A dedicated high-throughput EEG acquisition unit recently cleared a Class II registration and is moving toward hospital use.
Why a safer implant may win the ward
Beijing's Changping district, where much of this work sits, now hosts 44 brain-science and brain–computer interface companies, and the municipality has paired that cluster with a 300-million-yuan (≈ US$42m / HK$330m) brain-science fund. The policy bet is that volume will come from devices hospitals can actually adopt.
A half-invasive implant is easier to defend to a surgical committee than an open-brain procedure. The surgery reuses skills neurosurgeons already have, the device is reversible in principle, and wireless power removes the battery-replacement operation. Those unglamorous features are what turn a demonstration into a routinable treatment.
Trend reading
According to Li An, chief scientist at BrainNet (脑机网), the real bottleneck for semi-invasive systems is not the implant surgery but the decoding stability that must survive months inside a moving, healing brain — the algorithm, not the electrode, is where the program will be won or lost.
Honest limitations
- The implant count and safe-hours figures come from Beijing municipal and company statements, not a peer-reviewed publication; independent clinical data should confirm them.
- Patient recovery stories — including the reported shift from complete to incomplete injury — are company-described outcomes relayed through state media, based on a small, early cohort rather than a controlled trial.
- The 2027 registration timeline and the "Beinao-2" 2026 human-validation target are company plans, not approvals; regulatory outcomes remain uncertain.
- Long-term behavior of an epidural film over years — tissue response, signal drift, electrode adhesion — cannot be known from implants still measured in months.
What readers can do now
- Patients and families: ask rehabilitation and neurosurgery departments whether their hospital runs Beinao-1 (北脑一号) research screening; the trial recruits tetraplegic adults with injuries older than twelve months.
- Clinicians: track the multicenter registration trial's published endpoints rather than demonstration videos, and compare semi-invasive against open-brain and non-invasive routes on surgical burden and reversibility.
- General readers: watch the June 30, 2026 national device-classification rules and provincial pricing pilots — those, more than any single implant, will decide when such therapy becomes billable.