Roughly 200 million people worldwide live with age-related macular degeneration, a condition that erodes central vision and, in its advanced dry form, has no approved treatment that can reverse the damage. A research team now reports a device that could change the calculus: an implantable artificial retina that converts near-infrared light into electrical signals strong enough to activate the eye’s output neurons, potentially restoring useful sight and, in a twist no conventional therapy offers, opening a perceptual channel into wavelengths humans have never been able to see.
The device is detailed in a paper published in early 2026 in Nature Electronics. The study, whose authors include researchers specializing in flexible bioelectronics and retinal prosthesis design, combines a phototransistor array, three-dimensional liquid-metal micropillar electrodes, and a near-infrared (NIR) transmission filter into an epiretinal implant thin enough to rest against the inner surface of the retina. In animal experiments, the implant translated NIR photons into electrical stimuli that reliably drove retinal ganglion cell activity, the critical first relay on the path to the brain. The team also deposited its device characterization data, animal-experiment results, and analysis code in a publicly accessible Figshare repository referenced in the paper’s data-availability statement, inviting independent verification. (Neither a standalone DOI for the repository nor full author names are reproduced here because the sourcing available as of May 2026 does not supply them independently of the paper itself.)
Building on proven electrode technology
The implant’s electrode architecture did not appear from nowhere. It draws on a fabrication strategy validated in an earlier Nature Nanotechnology study focused on liquid-metal microelectrodes for ultrathin retinal prostheses. That work established that soft, gallium-based electrode pillars could maintain biocompatibility and mechanical flexibility when placed against living retinal tissue, solving a long-standing problem: rigid electrodes tend to damage the delicate cell layers they are meant to stimulate. By inheriting that tested interface, the newer NIR device avoids re-litigating basic safety questions about its contact with neurons.
Infrared vision is no longer theoretical
The idea of granting mammals infrared perception has been demonstrated through more than one independent approach. In 2019, a team led by researchers at the University of Science and Technology of China reported in Cell that injectable upconversion nanoantennae allowed mice to form images from NIR light. The nanoparticles, injected beneath the retina, converted invisible infrared photons into visible wavelengths that rod and cone cells could detect. Mice fitted with the particles navigated mazes and distinguished patterns using only infrared cues, confirming that their brains interpreted the new input as coherent images rather than noise.
A 2020 Science paper took a complementary route, demonstrating tunable NIR-responsive sensors that restored light sensitivity in damaged retinas. By engineering photosensitive components that respond selectively to NIR wavelengths, the researchers showed that retinal tissue with severely reduced native function could still drive downstream neural activity when stimulated with the right infrared signals. Together with the nanoantennae work, the study confirmed that NIR-based retinal perception is experimentally established in animal models across multiple hardware and chemical strategies.
Can artificial and natural vision coexist?
For patients who still have partial sight, the critical question is whether an electronic implant will clash with whatever biological vision remains. A 2022 Nature Communications study addressed this directly, documenting simultaneous prosthetic and natural vision in patients with atrophic AMD who received a subretinal photovoltaic prosthesis. Participants reported that artificial signals added contrast and shape information without erasing their residual natural perception. The finding supports the broader premise behind the NIR device: that an infrared channel could be layered on top of existing sight rather than replacing it.
What has not been proven yet
Despite the encouraging preclinical results, the strongest evidence for the NIR epiretinal device still comes from animal experiments. No long-term biocompatibility data from human implants have been published for this specific configuration. The earlier liquid-metal electrode work and the nanoantennae studies were also conducted in animal models, meaning the entire chain of evidence supporting the “new vision channel” concept has yet to clear the bar of sustained human testing. Questions about scar-tissue formation, immune responses, and gradual changes in retinal architecture over years remain unanswered.
A related clinical effort, the PRIMAvera trial (NCT04676854), is evaluating a photovoltaic subretinal system for central vision restoration in atrophic AMD patients across multiple European sites. The trial’s registry entry confirms pre-specified endpoints and enrollment criteria, but no full peer-reviewed outcome analysis has been published as of May 2026. Until those results appear, claims about real-world benefits of photovoltaic retinal prostheses should be treated as provisional, and direct comparisons with NIR-based systems remain premature.
A gap also separates the two main technical strategies. The injectable nanoparticle approach is minimally invasive and piggybacks on existing photoreceptors, but raises concerns about long-term particle stability, clearance, and potential toxicity. The hardware implant centralizes complexity in a device that can, in principle, be removed or upgraded, but requires surgery and precise positioning. Whether the strengths of both methods can be combined, or whether one will prove superior for scalable human use, is an open question no published study has resolved.
Engineering hurdles remain formidable as well. Power delivery, heat dissipation inside the eye, mechanical robustness under constant eye movements, and long-term stability in the saline environment of the vitreous cavity are all challenges that bench-top and short-duration animal experiments do not fully address. The tunable NIR sensor work from 2020, for instance, established proof of concept at the sensor level but did not tackle the packaging and power problems that a daily-wear implant would face.
How the brain might handle a new spectrum
Perhaps the most fascinating unknown is perceptual. Animal studies show that neural circuits can adapt to new spectral inputs, but human perception is shaped by decades of experience with a specific range of wavelengths. It is unclear whether patients would experience NIR input as a distinct color, as enhanced brightness, or as a qualitatively different sensation altogether. Training protocols, rehabilitation strategies, and user-interface decisions, such as which NIR bands to emphasize for navigation versus reading, will all shape how usable the new channel becomes in daily life.
Current treatments for AMD underscore why the stakes are high. Anti-VEGF injections can slow wet AMD but do nothing for the dry, atrophic form that accounts for the majority of advanced cases. Gene therapies like voretigene neparvovec (Luxturna) target a narrow genetic subtype of inherited retinal dystrophy, not AMD. For the tens of millions of people whose central vision is fading with no pharmacological recourse, a device that can bypass damaged photoreceptors entirely and talk directly to ganglion cells represents a fundamentally different kind of intervention.
Why the next step is a first-in-human trial of the NIR implant
The strongest pieces of evidence are the peer-reviewed primary studies: the Nature Electronics paper describing the device, the Nature Nanotechnology paper validating the electrode architecture, the Cell paper proving NIR image-based vision in mammals, the Science paper demonstrating tunable NIR retinal sensors, and the Nature Communications paper showing prosthetic-natural vision coexistence in AMD patients. Each addresses a distinct link in the chain from electrode design to infrared perception to clinical compatibility. Taken together, they make it plausible, though not guaranteed, that an implant could one day provide both restored and expanded vision.
What remains to be shown is whether these devices can deliver stable, meaningful gains in daily life for people with degenerative eye disease, and whether the promise of seeing beyond the visible spectrum will translate into practical advantages worth the surgical risk. Until first-in-human data for the NIR epiretinal implant arrives, the technology sits at a compelling but familiar frontier in biomedical engineering: proven in the lab, unproven in the clinic, and tantalizing enough to keep pushing forward.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
