Real biological eyes—diagrammed at top—use arrays of retinal cells that are lined up in rows, but their interconnections (middle) use a fractal pattern common in nature, according to University of Oregon professor Richard Taylor. (Source: University of Oregon)
By designing electronic-eye (eEye) implants using fractal interconnects, researchers aim to overcome the mismatch between using conventional image chips in bionic eyes. Today, several efforts are under way worldwide to create silicon retinas that can be implanted in the eyes of the blind, thereby enabling them to see again, albeit at vastly reduced resolution. Now researchers are aiming to remedy that by replicating the fractal-like interconnection topology of real eyes.
Real biological eyes contain the equivalent of 127 million pixels, whereas conventional eEyes are currently using sensors with less than 64 pixels, and even next-generation designs are only aiming for about 1,024. What is even worse, these researchers say, is that the interconnection topology of the an image chip is a square array, whereas the interconnection matrix for the "pixels" in a real biological eye is a branching structure called a fractal.
Fractals are common among all living things as a result of growing techniques that repeat a basic set of instructions—a fractal algorithm. For instance, the trunk of a tree divides into branches using the same fractal algorithm that is used for the veins in a leaf. In nature, trees, clouds, rivers, galaxies, lungs and neurons use the same fractal pattern of interconnections.
Now researchers are aiming to replicate this technique for interconnecting imaging elements in eEyes. Today's eEyes just sink metallic electrodes—one for each pixel—into the ganglia behind the eye, which then depends on the plasticity of the visual cortex in the brain to decipher the output from these new pixels—even though they do not match the normal topology of the biological retina. However, new research efforts are developing a technique that starts with a metallic seed that then grows all the repeated branching structures that in turn mate to the optic nerve behind the eye, thereby delivering to the brain the same kind of signals as retinal neurons.
The specific algorithm harnessed by the technique is called "diffusion limited aggregation," which researchers are using to grow image sensor interconnections that mimic a natural neural topology before being surgically implanted and interfaced to the optic nerve.
This summer Professor Richard Taylor and doctoral candidate Rick Montgomery will begin a yearlong quest with Professor Simon Brown at the University of Canterbury in New Zealand to grow these metallic fractal interconnection topologies for the backside of silicon image chips.
Instead of just providing a single output for each pixel, as with conventional eEyes, image sensors with the fractal interconnects will connect to the optic nerve with the same overlapping topology used by real biological retinal neurons. As a result, the researchers hope that the brain's visual cortex can perform the same sort of functions for eEyes that it does for real eyes, enabling the blind to recover not just some vision, but a visual experience that rivals that of normal people.
One challenge cited by the researchers is finding metals that can be coaxed with diffusion limited aggregation to form the Brownian trees typical of retinal cells and yet can be safely implanted into humans without side effects. Funding is being provided by the Office of Naval Research (ONR), the U.S. Air Force and the National Science Foundation.
Further Reading
