A 67-year-old man who once developed film in a photo shop sits in a clinic in Fujian. Six months earlier, a stroke silenced the wiring between his brain and his left arm. He thinks, hard, the single word "lift" — and this time the arm begins, slowly, to rise.
That moment is the headline result of work led by Duan Feng (段峰), a professor at Nankai University. His team did not crack open the skull to reach the brain. They slipped a brain–computer interface (脑机接口) in through a blood vessel instead.
The problem with reading the brain
Most attempts to connect a human mind to a machine fall into two camps. The "invasive" kind drills through the skull and plants electrodes directly in brain tissue — powerful, but risky and hard to reverse. The "non-invasive" kind rests a cap of sensors on the scalp, safe but fuzzy, because hair, skin, and bone smear the signal.
Duan Feng's group chose a third lane: the interventional brain–computer interface (介入式脑机接口). Instead of opening the head, doctors thread a catheter up a vein in the neck and place a tiny stent electrode against the wall of a vessel near the brain's motor cortex. A coin-sized wireless unit, buried under the chest skin, picks up the signal and beams it out.
A stent, a vein, no scalpel
In the Fujian trial, the team worked with doctors at Sanbo Fulong Brain Hospital and the Fujian Provincial Hospital of Traditional Chinese Medicine. The patient — a man left hemiplegic by a cerebral infarction — had the stent electrode delivered to a vessel beside his motor cortex. When he imagined moving, the device read the faint electrical intent and fired functional electrical stimulation to his paralysed muscles, in time with his own thought.
Crucially, the hardware stayed in the body only about two weeks, then was removed. The team reported no infection, no clot, no rejection. For a field where "permanent implant in the brain" is the scary default, a device you can take back out changes the risk conversation.
From sheep to monkey to a human
This was not a lucky first throw. The group spent years in the animal lab first:
- 2022 — first domestic animal test of an interventional BCI, in a sheep, proving the electrode could sit in a vessel and read brain signals.
- 2023 — first non-human primate test, where a monkey used the link to control a robotic arm and feed itself.
- 2024 — first retrieval test, pulling the sensor back out of a sheep's vessel to confirm the body tolerated both insertion and removal.
- 2025 — the first human trial, the stroke survivor in Fujian who moved his arm again.
The staged path mirrors how serious medical-device work is supposed to go: prove it is possible, prove it is useful, prove it is safe to remove, then try a person.
Why "interventional" matters
The approach leans on a surgery vascular doctors already perform thousands of times a day — placing stents and leads through veins, much like a pacemaker procedure. That familiarity is the whole point. Duan Feng has said the hardest part was not the engineering but finding a hospital willing to be first; by his account the team approached more than 30 hospitals before one said yes.
It also sits in a global race. The US firm Synchron has pursued a similar stent-based implant, the Stentrode, and ran its own animal tests. China's interventional route is one of several national bets on brain–computer interface (脑机接口) as a clinical and industrial field.
From lab to a real product
The research has a commercial arm. Tianjin Rongyuan Intelligent Technology (天津嵘元智能科技), set up in April 2023 at the Tiankai High-Tech Park, is the team's translation vehicle, working on interventional BCI hardware, domestically sourced neuro-intervention tools, and rehab robots. Trials are expanding into Fujian and Shandong, with a research ward for BCI patients, and the team targets medical-device registration within roughly three years.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the interventional route is "the most pragmatic path to first clinical use, because it reuses a surgery doctors already perform every day."
Honest limitations
This is still early. The human result rests on a single case, with the device in place only about two weeks — far from proof of long-term safety. Broader questions — how well it scales to leg movement, to speech, to degenerative disease — are unanswered. Figures here come from the university and the team's own disclosures; no independent multi-centre trial or regulator's sign-off has been published yet, and the three-year registration timeline is a plan, not a promise.
What readers can do now
- If you or a family member faces stroke rehab, ask neurologists at major centres about brain–computer interface (脑机接口) clinical trials rather than assuming nothing is available.
- If you study engineering or medicine, the convergence skills here are materials science, vascular intervention, and machine-learning signal decoding — a rare spot where three fields meet.
- If you follow policy or investment, watch the national medical-device approval pathway; a cleared interventional BCI would open a large rehab market almost overnight.