Two neuroscientists have created a prosthesis that can partially restore the sight to blind mice. The device could eventually be developed for use in humans.

More than 20 million people worldwide become blind owing to the degeneration of their retina, the thin tissue at the back of the eye that turns light into a neural signal. Only one prosthesis has been approved for treatment of the condition — it consists of an array of surgically implanted electrodes that directly stimulate the optic nerve and allow patients to discern edges and letters. Patients cannot, however, recognize faces or perform many everyday tasks.

Sheila Nirenberg, a physiologist at the Weill Medical College at Cornell University in New York thinks that the problem is at least partially down to coding. Even though the retina is as thin as tissue paper, it contains several layers of nerves that seem to encode light into neural signals. "The thing is, nobody knew the code," she says. Without it, Nirenberg believes that visual prostheses will never be able to create images that the brain can easily recognize.

Now, she and her student, Chethan Pandarinath, have come up with a code and developed a device that uses it to restore some sight in blind mice.

The duo began by injecting nerve cells in the retinas of their mice with a genetically engineered virus. The virus had been designed to insert a gene that causes the cells to produce a light-sensitive protein normally found in algae. When a beam of light was then shown into the eye, the protein triggered the nerve cells to send a signal to the brain, performing a similar function to healthy rod and cone cells. 

Coded vision

A prosthetic retina that can translate an image into neural signals was tested using a picture of a baby's face. A is the original image. B is the image after it passes through the coding software. C is after it has been processed by the retinal cells. D is the processed image without coding. Credit: S. Nirenberg/PNAS

Earlier efforts had managed to get this far, but Nirenberg and Pandarinath went a step further. Rather than feeding visual signals directly into the eye, they processed them using a code that the pair had developed by watching how a healthy retina responds to stimuli. After receiving the encoded input, the mice were able to track moving stripes1, something that they hadn't been able to do before. The pair then looked at the neural signals that the mice were producing and used a different, 'untranslate', code to figure out what the brain would have been seeing. The encoded image was clearer and more recognizable than the non-encoded one (see image). Nirenberg believes that the combination of computer code with the genetically-modified nerve cells constitutes a proof-of-principle prosthetic.

The importance of encoding has been debated among scientists working on visual prostheses, says James Weiland, an ophthalmologist at the University of Southern California in Los Angeles. Some think that it will be crucial, but others think the brain can adapt to an unprocessed signal. Nirenberg and Pandarinath have shown that encoding provides an advantage, he says, but how effective it is won't be known until the technique is tried out in people. "You can't say for sure until you have the patient telling you 'yes I see it. It's better when you do that'," he says.

Nirenberg hopes to test the system in human trials soon. The encoding is simple enough to be done by a microchip, which, together with a small video camera could fit onto a pair of glasses. The camera would record a signal and the encoder would then flash it directly onto the genetically treated nerve cells in the eye. If it works, the technique is simple enough that it could be done in a doctor's office. "We would like to [try it] in patients in the next one or two years," she says.