Brain Implants to Restore Vision

A bionic implant that bypasses the eye might eventually treat many forms of blindness.

By Duncan Graham-Rowe

Bionic vision: Researchers at Harvard Medical School aim to build a small digital camera that will feed images to an external signal processor worn by the patient. The processor will translate the image from the camera into neural impulses, then transmit them wirelessly to an implanted stimulator. The stimulator will drive a set of electrodes placed in the lateral geniculate nucleus of the brain to elicit images in the patient’s brain.
Credit: J. S. Pezaris, adapted with permission from D. H. Hubel



View an animation of the animal's eye movements.

One day it may be possible to restore sight in people who are congenitally blind by placing an implant in a part of the vision system hitherto ignored. Unlike most visual prostheses in advanced development, this new approach could make it possible to treat blindness even when the entire eye is damaged.

While the work is still in the early stages, researchers ultimately envision a device that translates images from a digital camera into neural impulses and then feeds that information into the visual system, allowing the wearer to see.

Previous research has shown that visual sensations, known as percepts, can be elicited in blind subjects by electrically stimulating nerve cells within the vision system. Researchers at Harvard Medical School, in Boston, are designing a visual prosthesis that builds on that observation.

Several types of sight-boosting prostheses are currently under development, with some already being tested in humans. But while these largely target the retina, the Harvard researchers chose to focus on part of the visual system called the lateral geniculate nucleus (LGN), a relay station along the route from the optic nerve to the visual cortex, where visual information is processed. Because it's upstream of the eye, this area could be targeted in people with extensive eye damage.

And unlike locations in the visual cortex, the LGN is one of the first "stops" in the visual system, meaning that the neural signals encoding visual information have not yet been extensively processed and spread throughout the brain. "[In the LGN] there is a straightforward mapping of the visual scene on the tissue," says John Pezaris, a neural-systems engineer at Harvard Medical School, who co-authored the research along with neuroscientist Clay Reid, also at Harvard Medical School. This means that specific parts of the LGN are linked to specific parts of the visual scene. When a light flashes in one location, for example, the corresponding area in the LGN will become active.

To determine if activity in the LGN can mimic visual stimuli, the researchers implanted electrodes in the LGNs of two monkeys that had been trained to move their eyes rapidly toward points of light when they appeared on a screen. When a part of the LGN corresponding to a specific part of the visual field was electrically stimulated, the monkeys would shift their gaze to that point on the screen. The results, published today in the Proceedings of the National Academy of Sciences, suggest that the monkeys were "seeing" the pulses in their field of view even though nothing was appearing on the screen.

"It's an absolutely stunning piece of work," says James Morrison, a physiologist and principal investigator for the Retinal Prosthesis Group at the University of Glasgow's Institute of Biomedical and Life Sciences, in Scotland. However, Morrison says, the position of the LGN is a major disadvantage of the approach. It's located in the middle of the head, making it difficult to access.

Recent advances in neurosurgical techniques, such as deep brain stimulators for treating Parkinson's disease, may help solve this issue: the LGN is just a few centimeters away from where these stimulators are placed, says Pezaris.

Still, it's too soon to say whether the findings will lead to better brain implants. "While I think the paper holds scientific merit, I think it will be extremely difficult to restore blindness from there," says Thomas Serre, a neuroscientist at the Center for Biological and Computational Learning at MIT's McGovern Institute for Brain Research. He believes that neurons in the LGN may be spaced too closely together to be stimulated individually, which would be important in trying to reproduce natural vision. "I don't think we will ever be able to go beyond generating very simple percepts like points of light," he says.

Pezaris accepts that a tremendous amount of work is needed before the LGN can be used to treat blindness, but he says this work does at least open the door to that possibility. "This was just the first very small step," he says.


Reconstructed video from the LGN of a cat. Right video shows what the cat sees, Left video shows the cat's LGN neurons firing pattern.

In a 1999 study, Garett B. Stanley, Fei F. Li and Yang Dan have literally jacked into the LGN of a cat. Multiple cells in the LGN of anesthetized cats were recorded simultaneously with multielectrodes. The spike trains of the neurons were binned according to the frame rate of the stimulus (32 Hz for movies, 128 Hz for white noise) and converted to firing rate signals. They recorded the responses of the cells to multiple repeats of eight short movies, and these data were used for subsequent analyses. The geniculate cells were well driven by the movie stimuli, as indicated by their mean firing rates, which were higher during movie presentation than in the absence of visual stimuli.


Read more from the original paper (Reconstruction of Natural Scenes from Ensemble Responses in the Lateral Geniculate Nucleus-PDF)

An interesting observation from the cat's LGN study is the apparent pre-processing of information prior to the visual cortex. A frame from the cat's LGN movie shows that the LGN or retina enhances cat like features of human faces. - Biotele

More in depth explanation of the LGN vision processing by Stanley Garret