A stroke patient who cannot move a finger sits still, eyes on a screen. No keyboard, no switch — just a thin cap of electrodes and a thought. Across the room, a robotic "extra finger" curls to pick up a ball. The signal that moved it was generated entirely inside the patient's head. This is not a lab demo from a far-off future; it is a Tuesday at Ming Dong's (明东) lab in Tianjin.
Most people have never heard of him. In brain-computer interface (脑机接口) circles, he is hard to avoid.
The researcher who stayed in the classroom
Ming was born in July 1976 in Yantai, Shandong. He earned both his bachelor's (1999) and PhD (2004) in biomedical engineering at Tianjin University, then joined the faculty in 2006 and became a full professor in 2011. Today he is vice president of Tianjin University, dean of its School of Medicine, and director of the Haihe Laboratory of Brain-Computer Interface (脑机接口) — a flagship Tianjin research lab built precisely for this field.
He is also a member of the National Committee of the Chinese People's Political Consultative Conference and a leader in the Jiusan Society's science and technology committee, which means he spends as much time arguing for the field's future in policy rooms as in cleanrooms.
Why non-invasive, not invasive
Brain-computer interface (脑机接口) splits into two camps. Invasive systems implant electrodes in or on the brain for crisp signals but carry real surgical and long-term safety risk. Non-invasive systems read brain activity through the scalp — lower bandwidth, far safer.
Ming's bet, repeated in interviews, is blunt: a future industry cannot be built by putting chips in everyone's heads. High-performance non-invasive brain-computer interface (脑机接口) is the only path that "flies into ordinary homes." He has also been willing to cool the hype, warning openly about bubble,炒作 (hype), and toy-like demos — a rare note of restraint in a field drunk on headlines.
From "thought control" to a rehab clinic
Under Ming, the lab built the "神工" (Divine Craftsman) series — artificial neural robot systems for stroke rehabilitation. The first full-limb stroke-rehab neural robot, "神工一号," demonstrated that a patient's own intention could drive external assistance to retrain weakened limbs.
Concrete outputs the lab and its partners cite:
- A non-invasive intracranial-pressure monitor accurate to 0.05 mmHg — no craniotomy needed for severe patients.
- The first in-space brain-computer interface (脑机接口) experiment, run on the Tiangong-2 space station.
- A 2023 non-invasive system with an ultra-large instruction set and high transmission rate, unveiled at the World Intelligence Congress.
- A 2025 collaboration with Tsinghua University on a memristor-based neuromorphic non-invasive "evolving" BCI system, published in Nature Electronics.
- The world's first on-chip BCI intelligent interaction system (2024), growing brain-like tissue on an electrode chip — work done with Southern University of Science and Technology.
The clinic is the point
For Ming, the metric that matters is not a flashy demo but a hospital bed. In April 2025 the lab and Tianjin Huanhu Hospital began building China's first comprehensive BCI clinical trial ward; it opened on 1 October 2025, organizing solutions for stroke, cerebral palsy, Parkinson's, and epilepsy.
Policy caught up the same year. In March 2025, the National Healthcare Security Administration issued pricing guidelines that created separate billing items for brain-computer interface (脑机接口) — non-invasive adaptation, invasive implantation, and removal. In plain terms: once the tech is ready, hospitals will have a sanctioned way to charge for it.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the non-invasive path Ming champions is where the country's BCI field can realistically scale before invasive implants clear long safety and ethics hurdles.
Where large models (大模型) fit in
Ming is explicit that brain-computer interface (脑机接口) and large models (大模型) are two-way fuel. An emotional BCI can read a user's state and help a large model (大模型) write more human text; a motor BCI can hand a model's decisions to a prosthetic or wheelchair. The catch, he notes, is data: BCI datasets sit around 10^4 scale, hopelessly short of the 10^9 a large model (大模型) needs to train well. Closing that gap is the next hard problem.
Tianjin has even spun structure around the field: a registered-capital ¥1 billion (≈ US$141M / HK$1.1B) brain-computer interface industry group, plus the country's first BCI undergraduate major and first BCI first-level doctoral discipline.
Honest limitations
Readers should calibrate enthusiasm against real constraints:
- Much of this is early-stage. "Firsts" here are lab, peer-reviewed, or pilot-clinic claims — not mass-market products. The Huanhu ward is a trial site, not a nationwide rollout.
- Non-invasive means lower bandwidth. Scalp signals are noisier and coarser than implanted ones; for some applications invasive BCI will remain superior for years.
- Independent replication is limited. Many milestones are reported via the lab, the university, and state media; we have not seen large multi-center trials confirming long-term efficacy at scale.
- Regulation is new. 2025's pricing and ethics guidelines are foundations, not proof of safe, durable clinical use.
- Timelines are aspirational. "Flying into ordinary homes" depends on cost, reliability, and reimbursement that are not yet settled.
What readers can do now
- Watch the Huanhu Hospital BCI ward for published outcomes — patient counts, adverse events, and rehab gains are the real scoreboard.
- Follow National Healthcare Security Administration billing updates; reimbursable BCI codes are the gate to hospital adoption.
- Try the open-source stack. The lab's MetaBCI is an open Python platform — a concrete way to see what non-invasive BCI can and cannot do today.
- Track the ethics guidelines (2024 BCI research ethics steer, 2025 standardization roadmap) as the guardrails that will decide how fast this field is allowed to move.