New technologies promise sharper artificial vision for blind peopl
CHICAGO, ILLINOIS—In 2014, U.S. regulators approved a futuristic treatment for blindness. The device, called Argus II, sends signals from a glasses-mounted camera to a roughly 3-by-5-millimeter grid of electrodes at the back of eye. Its job: Replace signals from light-sensing cells lost in the genetic condition retinitis pigmentosa. The implant’s maker, Second Sight, estimates that about 350 people in the world now use it. Argus II offers a relatively crude form of artificial vision; users see diffuse spots of light called phosphenes. “None of the patients gave up their white cane or guide dog,” says Daniel Palanker, a physicist who works on visual prostheses at Stanford University in Palo Alto, California. “It’s a very low bar.”
But it was a start. He and others are now aiming to raise the bar with more precise ways of stimulating cells in the eye or brain. At the annual meeting of the Society for Neuroscience here last week, scientists shared progress from several such efforts. Some have already advanced to human trials—“a real, final test,” Palanker says. “It’s exciting times.”
Several common disorders steal vision by destroying photoreceptors, the first cells in a relay of information from the eye to the brain. The other players in the relay often remain intact: the so-called bipolar cells, which receive photoreceptors’ signals; the retinal ganglion cells, which form the optic nerve and carry those signals to the brain; and the multilayered visual cortex at the back of the brain, which organizes the information into meaningful sight.
Because adjacent points in space project onto adjacent points on the retina, and eventually activate neighboring points in an early processing area of the visual cortex, a visual scene can be mapped onto a spatial pattern of signals. But this spatial mapping gets more complex along the relay, so some researchers aim to activate cells as close to the start as possible.
Palanker’s team has designed a retinal implant of about 400 photodiodes or “pixels” that replace some of the retina’s spatial map. A video stream of the outside world is shown on the inside of a pair of glasses in near-infrared light, which the implant’s pixels convert into electrical signals to stimulate the retina’s bipolar cells. The Paris-based company Pixium Vision is testing the device in five people who have the photoreceptor-destroying disease macular degeneration. At last week’s meeting, Palanker presented videos showing that participants who had been implanted with the prosthesis for about 1 year could recognize objects on a table and read printed or on-screen letters. The artificial vision is good enough to make out the title of a book, Palanker says, though not the words on its pages. His team is now working to shrink the photodiodes—creating finer pixels and sharper vision—without losing too much signal strength.
To push to higher precision than electrical stimulation of the eye can achieve, other teams are turning to optogenetics, a technique for activating cells with light. In a clinical trial by Paris-based GenSight Biologics, researchers have injected a harmless virus carrying the gene for a light-sensitive protein into the eyes of five people with retinitis pigmentosa. Retinal ganglion cells that take up the gene can then respond to red light projected into the eye. Whether the trial participants will gain useful vision should become clear next year, says José-Alain Sahel, an ophthalmologist and neuroscientist testing the technology at the University of Pittsburgh School of Medicine in Pennsylvania and the Vision Institute in Paris.
But therapies targeting retinal cells won’t help people who have lost much of their eye to injury or have severe damage to the optic nerve from conditions such as glaucoma.
Second Sight aims to treat these patients with Orion, an implant of 60 electrodes that sits directly on the visual cortex and feeds the brain signals from a glasses-mounted video camera. Four of five blind patients who have had the implant for about 1 year could better locate a roughly fist-size white square on a black screen. All five were better able to detect the direction in which a white bar moved across the screen. “We’re encouraged,” says Jessy Dorn, director of scientific research at the company in Sylmar, California.
Electrodes that sit on the brain’s surface have drawbacks. Because it takes a relatively strong current to activate the target neurons in the tissue below, activating multiple electrodes at once risks triggering a seizure. And activating neighboring electrodes can stimulate the tissue between them, fusing two discrete visual points into a blob. But at the meeting, Second Sight’s collaborators at Baylor College of Medicine in Houston, Texas, presented evidence that the 60 electrodes could create phosphenes at more than 60 locations. The scientists employed a technique known as current steering, which is already used to enhance pitch perception with cochlear implants.
Electrodes that penetrate deeper into the visual cortex could get closer to target neurons and use a lower current to activate smaller, more precise points in the tissue. Last week, Xing Chen, a neuroscientist in Pieter Roelfsema’s lab at the Netherlands Institute for Neuroscience in Amsterdam, presented tests of implants containing 1000 of these penetrating electrodes in two sighted monkeys. The animals could distinguish between different letters the researchers flashed into their visual field by activating 10 to 15 electrodes at a time. Roelfsema hopes to start human trials by 2023.
Stephen Macknik, a neuroscientist at the State University of New York’s Downstate Health Sciences University in New York City, cautions that the brain will eventually form a scar around implanted wires, walling them off from their target neurons. Such implants are “ruining the cortex for all other implants in the future,” he says, “and at best, [the user is] not going to see much.” He contends that optogenetics promises sharper vision, making penetrating electrodes unethical. At the meeting, Macknik presented plans for a technology called OBServ, which would add a light-sensitive opsin gene to neurons that reach into the visual cortex from a signal waystation at the base of the brain. Those cells, he explained, could be activated with light shined from the brain’s surface.
Cortical optogenetic systems such as OBServ won’t reach the clinic anytime soon. Researchers still need to demonstrate that a virus can safely and reliably endow particular neurons with an opsin gene that sticks around for years. They’ll also need to implant a highly precise, yet compact, device under the skull that flashes light into the brain while reading out neural activity to calibrate stimulation on the fly.
But one of the biggest barriers to beaming ultraprecise vision into the brain, many researchers say, is much more fundamental: discovering which stimulation patterns the brain will be able to interpret. “We don’t think that just because you had, say, a million electrodes or perfect spatial optogenetic activation, everything is just solved,” says Baylor neuroscientist William Bosking. “We need to learn how to talk to the cortex.”