Monkeys See Things That Aren’t There With Artificial Vision Brain Implant

Monkeys See Things That Aren’t There With Artificial Vision Brain Implant
Graphic showing the pedestal, or brain cap, and chip array. Each of the 16 chips contains 64 electrodes, for a total of 1,024 electrodes. (Image: Pieter Roelfsema)

Using brain implants, scientists have triggered the perception of shapes in the visual field of monkeys. Sounds spooky, but the technology could enable artificial eyesight in people with severe visual impairments.

Researchers have used a high-definition neuroprosthesis to trigger the perception of discernible shapes, and even the movement of these shapes, in monkeys. As described in Science, the device is implanted in the visual cortex of monkeys and causes them to see dots of light called phosphenes. These artificial dots can be displayed in meaningful patterns and eventually be made to represent objects in the real world, in what would be a remarkable advancement for treating blindness.

“Electrical stimulation of the visual cortex has long been proposed as an approach to restore vision in blind people, bypassing severe retinal degeneration or damage to the eye or the optic nerve,” write the authors in the new paper.

Indeed, this idea actually dates back to the 1970s, and scientists have made many attempts to stimulate the brain such that it produces artificial images. The problem, however, is that these prior solutions were only able to generate a small amount of data, that is, pixels, at a time, greatly limiting their practical use. The new approach, in which scientists created a brain implant with 1,024 electrodes, was made possible thanks to cutting-edge implant technology, namely new materials and better microelectronics. What’s more, the new implant is more stable and durable than previous versions.

Conceptual image showing how the technology could enable artificial vision for humans.  (Image: Pieter Roelfsema) Conceptual image showing how the technology could enable artificial vision for humans. (Image: Pieter Roelfsema)

The electrodes work by zapping the visual cortex with tiny spurts of electrical stimulation. This elicits the perception of the phosphenes, which can be made to appear at specified regions of a person’s visual field. Or, in the case of this experiment, in the visual space of two male rhesus macaque monkeys.

With their 1,024-channel neuroprosthesis, the scientists, led by Pieter Roelfsema from the Netherlands Institute for Neuroscience, were able to evoke “interpretable artificial percepts,” which were composed of multiple phosphenes appearing simultaneously, according to the paper. The number of artificial pixels made possible by the implant is unprecedented, as previous implants didn’t get any higher than 200 electrodes.

I asked Roelfsema to describe what the monkeys were seeing.

“The best analogy is a matrix board,” he explained in an email. “When you light up one bulb, the viewer sees a dot of light. This is like a single phosphene. But you can convey meaningful information by lighting up multiple bulbs as a pattern. That would be a pattern of phosphenes that conveys shape information.”

During the first phase of the experiment, Roelfsema and his colleagues trained the two monkeys to track and identify patterns of dots conveyed to them in real life, as these monkeys had normal vision. These training exercises were then replicated with the phosphenes. Roelfsema said the training process was a breeze, as the team takes “small steps” to “make sure that they always can find out what they are supposed to do.”

The monkeys, equipped with their brain implants, were first asked to perform basic tasks, such as moving their eyes to indicate the location of the phosphenes. The monkeys were then tested on more complex tasks, such as indicating the motion of the phosphenes, which was done by triggering a linear sequence of blinking phosphenes. Incredibly, the monkeys were also able to identify letters, which were produced by the simultaneous firing of eight to 15 electrodes.

“We trained them extensively before we implanted the electrodes, for a task in which they could use their eyes,” said Roelfsema. “But when we switched on the prosthesis after the brain surgery, we were thrilled that they immediately recognised the patterns imposed on their brain — the same patterns they had first learned to recognise visually.”

(A) Graphic showing the steps for generating the letter 'A' with phosphenes. (B) Impressions of phosphenes forming the letters 'A' and 'S.' (Image: X. Chen et al., 2020/Science) (A) Graphic showing the steps for generating the letter ‘A’ with phosphenes. (B) Impressions of phosphenes forming the letters ‘A’ and ‘S.’ (Image: X. Chen et al., 2020/Science)

In all, the monkeys were able to recognise shapes, including lines, moving dots, and letters, in a promising demonstration of artificial vision. Eventually, a similar technology could be used to treat people with severe eye injuries or degenerative disorders of the eye and optic nerve, as the implants bypass visual processing in the eye and work directly on the visual cortex in the brain.

In terms of limitations, the electrodes used in the study degrade and stop working after a couple of years. Roelfsema said his team is currently researching other electrode material to increase the longevity of the neuroprosthetic.

The experiment abided by the NIH Guide for Care and Use of Laboratory Animals. Animal welfare “is crucial” in this type of work, said Roelfsema, and his team made sure that the monkeys were held under “excellent conditions.” If the monkeys were uncomfortable, “they would not cooperate and take part in the tasks,” he said.

Brain implants that trigger the perception of phosphenes have already been used in humans, including a fascinating experiment done in 2014 that allowed a rudimentary form of brain-to-brain communication. What’s more, a co-author of the new study, Eduardo Fernandez from the Miguel Hernández University of Elche in Spain, has already tested the same type of electrode in a blind person, but with far fewer electrodes (so no patterns could be discerned).

A possible way in which a street scene could be conveyed with phosphenes.  (Image: Pieter Roelfsema) A possible way in which a street scene could be conveyed with phosphenes. (Image: Pieter Roelfsema)

The potential for artificial vision is quite exciting, and it’s got me imagining versions in which visually impaired people can recognise objects in their environment or even text from a book. The authors are imagining the same, as the graphic above, which shows a possible phosphene representation of a street scene, illustrates.

Basically, this solution can convey anything that can be represented by blinking pixels, which is actually quite a lot. Creating this visual “language” sounds like something for future scientists, linguists, and semioticians to figure out. It could be like Braille on steroids. And as this technology improves, meaning more pixels, these patterns could actually be made to look more like the objects they’re meant to represent. It will be fascinating to see where this technology goes from here.