The pen hovered, then moved. On the page appeared two characters — 中国. The woman holding the pen had not seen a face, a word, or the sky in more than twenty years.
What guided her hand was not her eyes. It was a thin array of electrodes sitting against the back of her retina, turning light from a camera into electrical pulses her brain could finally read as shape.
This is the retinal-path visual brain–computer interface (脑机接口), and in 2026 it moved from animal lab to human ward in China — quietly, and far ahead of the hype cycle.
Why "retinal-path" and not a camera in the brain
Most people imagine visual restoration as a chip wired straight into the visual cortex. The Chinese teams took a different, gentler route.
The retina is the eye's own neural sheet. In diseases like retinitis pigmentosa, the light-sensing cells die but many of the output neurons — the ganglion cells — survive. A retinal-path implant targets those survivors:
- A small external camera captures the scene.
- An algorithm converts the image into patterned electrical signals.
- A high-density electrode array stimulates the remaining retinal neurons.
- The natural optic nerve carries the signal to the visual cortex, the way it always did.
Because the optic nerve and brain are left intact, the patient's own visual pipeline does the hard decoding. The device only has to speak the retina's language.
A 320-channel first, and a 40-year-old patient
The most documented case comes from Nuanxinja (暖芯迦), a Hangzhou-area company, working with Zhejiang Provincial People's Hospital (浙江省人民医院).
The first human implant of its 320-channel retinal-path visual brain–computer interface was completed in May 2026. The recipient was a 40-year-old woman with retinitis pigmentosa lasting over twenty years — only faint light perception remained, no shapes, no faces.
After activation she could:
- clearly recognize and neatly write the characters "中国";
- precisely grasp everyday objects such as a computer mouse and a drinks bottle;
- walk independently in a straight indoor line and make right-angle turns.
A second site, the Hangzhou branch of the Eye Hospital of Wenzhou Medical University, opened its own trial in June 2026. Nuanxinja has said it plans about 30 human implants that year and is building a follow-up 1,280-channel chip for sharper vision.
A parallel route from a different company
Nuanxinja is not alone. Wisecare (微智医疗) built the IMIE smart retina, a 256-channel flexible electrode placed on the macula to stimulate surviving ganglion cells. In April 2026 Xiangya Hospital of Central South University (中南大学湘雅医院) implanted it in a fully blind subject; by June the team reported the patient could read characters on an eye chart and walk through doorways indoors.
Both products share the same bet: restore functional, low-resolution vision — enough to navigate a room, find a cup, see a face's outline — rather than promise 20/20 sight. That restraint is what makes the engineering believable.
How the implant goes in
Unlike cortical devices that require opening the skull, a retinal-path implant is placed through the eye, in a procedure clinicians compare to advanced cataract surgery plus an electrode insertion. There is no craniotomy, no brain incision, and the patient later wears a specialized spectacle unit that carries the camera and the wireless power link. That surgical simplicity matters as much as the chip: it is what could let ordinary eye hospitals, not only elite neurosurgical centers, perform and maintain the procedure.
The price gap is the real story for patients
Visual prostheses have existed in the West for years, but at prices that put them out of reach. Nuanxinja's founder has contrasted overseas devices at roughly US$200,000–250,000 with a domestic target of about ¥100,000–150,000 (≈ US$14,000–21,000 / HK$110,000–165,000). The company has also said it began filing with the US FDA, a sign the "made in China" visual implant is being aimed at a global market, not just a domestic one.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the speed of these first implants shows domestic neurotech shifting from "can we stimulate the nerve" to "can a patient actually use it at home" — the harder and more useful question.
What readers can do now
- If a family member has retinitis pigmentosa or macular degeneration with intact optic nerve, ask retinal specialists about visual brain–computer interface trials rather than assuming only cochlear-style implants exist.
- Check whether the optic nerve is still functional — these devices need it; they do not help if the nerve or visual cortex is destroyed.
- Treat any "full sight restored" headline skeptically; current systems target contour, light, and large-object perception, not fine reading.
Honest limitations
Primary facts here come from district government releases (Yuhang, Hangzhou), health-news reporting, and company statements; independent peer-reviewed journals were not the main source for the "write 中国" and grasping results, and long-term durability beyond the reported months is not yet published. The two companies (Nuanxinja, Wisecare) use different electrode counts and surgical paths, and head-to-head data do not exist. Resolution remains low, and cost figures are company-stated, not reimbursed prices. These are early clinical cases, not approved mass-market products, and outcomes vary by remaining retinal structure.