Minimum criteria set for useful artificial vision

November 15, 2006

Geneva-How much visual information should be transmitted to the brain of blind patients to restore reading ability? Visuomotor coordination? Mobility?

A European consortium has set minimum requirements for useful artificial vision, according to Marco Pelizzone, ScD, of the Ophthalmology Clinic, Department of Clinical Neurosciences, University Hospitals of Geneva.

Testing computer simulations of an experimental retinal prosthesis in normal volunteers suggests that about 500 phosphenes, arranged over a 10° by 7° retinal area, provide the minimal amount of visual information required to restore useful function, Dr. Pelizzone said.

He also noted that the volunteers' ability to learn to perform tasks with both central and eccentric vision was a promising development.

"The real question now is: 'Can we develop a device where we can effectively transmit this kind of information at the electric interface?' " said Dr. Pelizzone, whose European consortium is working on a retinal implant that will meet this requirement.

He cautioned that visual perception with any retinal prosthesis would be severely limited. Also, these devices will have a finite number of stimulation contacts that will limit image resolution, and the stimulation will be concentrated on a fixed, relatively small, and probably eccentric retinal area.

Dr. Pelizzone and his colleagues studied the effects of these limitations using simulations of artificial vision on young, healthy subjects. This was accomplished with a very fast eye-tracking system, cameras, screens, and computer technology, all of which were incorporated into a helmet-like device. He summarized several years of research.

Everyday tasks

Investigators assessed the performance of everyday tasks such as full-page reading, visuomotor coordination, and whole body mobility in various environments. Image resolution, image eccentricity, and the effective visual field were varied.

Reading was first tested using simple, four-letter words. Investigators switched the volunteers to the more realistic task of full-page reading, using excerpts from local newspapers. The eye tracker provided the exact point of the subjects' gaze on the text, while the computer system limited their central vision to a 10° by 7° window containing 572 pixels. In these experiments, subjects used their eye movements to control page navigation.

With central vision, volunteers could follow lines of text relatively easily, identify 100% of the words correctly, and read at speeds of 70 to 120 words per minute, Dr. Pelizzone said.

The subjects had more difficulty with eccentric vision. Initially, some volunteers couldn't even see the window of the screen clearly. Their ability improved with habituation and additional training, however, and they slowly learned to see and follow lines of text. After 45 hours of practice, one volunteer was able to read with 90% or greater accuracy and at a speed of up to 50 words per minute.

Visuomotor coordination was assessed with a manipulation task in which volunteers had to move and place chips according to a random model and a pointing task involving a flashing light.

These tasks were more complicated than the reading experiments, Dr. Pelizzone said. Designing an effective reading test was easy, because research has shown that subjects need to see a minimum of four to seven characters and about two lines of text. No comparable data existed for visuomotor coordination and mobility, however, so Dr. Pelizzone's group experimented by systematically varying the effective visual field projected in the viewing window as well as the pixel density. Subjects were tested under conditions ranging from a large field of view with an unrealistic resolution of about 20,000 pixels to a very narrow, tunnel vision field of view of only 120 pixels. The subjects generally preferred resolutions in a middle range.

Test results showed that with central vision, volunteers could achieve more than 95% accuracy in arranging the chips after a period of habituation. The subjects performed the task slowly, however, often taking three to four times as long to complete the task as someone with normal viewing.