In normal vision, images are captured by the retina, transferred through the optic nerve, and decoded in the visual cortex. In most cases of blindness, defects occur in the eye or optic nerve, but the visual cortex is intact. This suggests that one might electronically send images directly to the cortex, bypassing the eye and optic nerve, as a way of artificially inducing vision in the blind. Dr. Philip Troyk and I, working with a team of neurosurgeons, engineers, histologists, physiologists and veterinareans from around the country, carried out a large-scale project to develop and test an artificial vision system of this type. We refer to the system as a cortical visual prosthesis.

The basic idea is to position an electrode in the thin layer on the surface of the brain, called the cortex, and pass electrical current to activate the surrounding neurons. When this happens, subjects perceive a tiny flash, called a phosphene, and the location of the phosphene depends on where the cortex is stimulated. It is usually assumed--though so far untested--that stimulating simultaneously at multiple sites will produce a kind of pixellated image.

But this is easier said than done. Figure 3, left shows a face sampled with approximately 1000 points (i.e. pixels) and it is recognizable. Unfortunately, it is not yet possible to stimulate at 1000 cortical locations; in fact up to 2003 the largest-ever implant was a 39-electrode array. Figure 3, right sows the same face sampled with 72 pixels. Clearly, we need to find a way to implant far more electrodes

Figure 3. A face sampled at 1000 points, simulating what a 1000-electrode stimulating array might achieve. Roll over to see the same face sampled at 72 points.

In 2003 we implanted 114 electrodes in the primary visual cortex of a rhesus monkey. We then trained the monkey to report the location of stimulated phosphenes, which he indicated by moving his eyes. After establishing that the animal did in fact perceive spatially distributed phosphenes, we implanted a second monkey with a 256-electrode array, by far the largest stimulation array ever implanted. The secret to implanting such a large array was in fact a combination of many things: every wire, connector, electrode and amplifier was reduced to its smallest possible size; electrodes were specially designed and rigorously tested to deliver current without damaging surrounding tissue; a variety of new surgical techniques were worked out; and a novel animal training paradigm was developed to test the array, not just in terms of physical function, but in terms of the perceptual effects it actually produced. With further research (and more funding) a 1000-point array will likely be possible before long, and using trained animals for rigorous testing, a functional human implant might be possible within a decade or so.