Bo Hong did not set out to build a headline-grabbing device. A neuroscientist by training, he spent years inside Tsinghua University's biomedical engineering program asking a quieter question: how do you listen to the brain without hurting it?
The answer his team landed on is a small implant placed just outside the brain's protective membrane — no wires through the skin, no battery sealed inside the skull.
That device, called NEO (Neural Electronic Opportunity), is a brain-computer interface (脑机接口) that has moved further than most from the lab bench into real hospitals. Hong, a professor of biomedical engineering at Tsinghua and dean of the university's Wei Xian honor college, leads the group behind it.
The bet on "halfway in"
For two decades, brain-computer interfaces have split into two camps. One puts sensors on the scalp — safe, but the signal is muffled by bone and skin, and control is rough. The other pushes electrodes deep into brain tissue for crisp signals, but at the cost of surgical risk, scarring, and long-term stability problems.
Hong's team picked a middle path. Instead of piercing the brain, NEO's electrodes sit in the epidural space, the thin layer just outside the dura mater that wraps the brain. Think of listening through a wall rather than walking into the room: you hear enough, and you do not disturb anyone inside.
"The critical question is what lies beyond the laboratory," Hong told Nature in 2026. "The challenge is not simply demonstrating what is possible, but ensuring that devices are safe, reliable, and usable over the long term." He frames the work as a series of pragmatic choices rather than a single breakthrough — usability and safety, he argues, must come before spectacle. Where groups such as Neuralink and the older BrainGate program thread electrodes into the cortex for maximum signal, NEO trades some resolution for a device that, in principle, can stay put for years.
A coin-sized translator
NEO is about 25 millimeters across, roughly the size of a coin, and sits flush within the skull over the sensorimotor cortex. Its job is to catch the faint electrical intent of movement from patients whose spinal cords no longer carry those signals — people with severe paralysis from injury or disease.
The harder engineering problems were power and decoding.
- No battery, no plug. Running a cable through the skin invites infection; sealing a battery inside the body means repeat surgeries when it dies. NEO uses near-field wireless coupling: a coil outside the scalp sends power across the skull to a second coil in the implant, and the same magnetic link carries neural data back out. "We use the magnetic field to send energy into the body, and use that same link to power a recording device and broadcast neural signals out," Hong explains.
- Weak signals, sharpened. Epidural signals are fainter than those from deep electrodes, but the team found they still carry useful information up to around 200 Hz. By blending several frequency bands and building what they call "virtual channels," they improved decoding enough for patients to move a cursor, drive a wheelchair, and — most consequentially — close a hand.
Beyond motor recovery
NEO is the visible tip of a longer research program. Hong's lab also studies how the brain handles speech and language, mapping how auditory, language, and motor regions cooperate in time to process tones and meaning. That work has appeared in journals including Nature Communications, Nature Neuroscience, and PNAS, and it feeds the same decoding toolkit the implant relies on.
The throughline is a preference for measured, clinically grounded engineering over dramatic demos. The group spent more than ten years working with hospitals such as Tsinghua's Yuquan campus affiliation before a single patient was implanted, deliberately choosing a route that clinicians — not just researchers — could defend.
From one patient to a registry
The first human implant took place in October 2023 at Xuanwu Hospital (宣武医院) in Beijing. Two more followed at Tiantan (天坛) and Huashan (华山) hospitals. By 2025 the program had expanded into a multi-center registered clinical trial across roughly a dozen hospitals, with more than 30 patients who had cervical spinal cord injuries.
According to Tsinghua's own accounts, every implanted patient reached brain-controlled grip. One person paralyzed for sixteen years was able to hold a grandchild; another flew on a commercial flight while using the system. These are small, specific freedoms, and the team treats them as the real measure of success rather than a benchmark score.
In early 2026, China's National Medical Products Administration approved NEO — formally the "implantable brain-computer interface hand motor function compensation system" — for clinical use. Multiple Tsinghua and state-media accounts describe it as the world's first approved invasive brain-computer interface medical device. Nature had already flagged the project as one of its events to watch in 2025.
Trend reading
According to Li An, chief scientist at BrainNet (脑机网), the harder problem for brain-computer interfaces is not the first implant but the long road from a single clinical success to routine, trusted care that hospitals and families can rely on.
Honest limitations
Most of the factual backbone here comes from Tsinghua University's official pages and a 2026 Nature Portfolio explainer built around Hong's own comments, supplemented by state-media reporting in People's Daily and Guangming Daily. A few numbers drift between official sources — one Tsinghua page cites eleven hospitals, another about a dozen, and the approval is described as both "early 2026" and "March 2026" — so treat the hospital count and exact approval month as approximate. What remains genuinely unknown is long-term durability beyond the trial window, the device's performance across a wider range of disabilities, and how NEO will be priced and scaled outside a research setting.
What readers can do now
- If you follow medical technology, track NEO's multi-center trial publications rather than launch headlines — they will show whether gains hold past the first year.
- If you work in rehabilitation or neurology in China, watch which hospitals join the approved program; that list signals where patients can actually access the system.
- If you are a researcher or investor, read the Nature Portfolio piece alongside Tsinghua's materials to separate the team's claims from independent description.