Rotating through a downward arc, the propulsion device collides inelastically with an orb approaching at 95 miles per hourse, reversing its angular momentum and direction. Two image acqusition units monitor the electromagnetic waves reflected by the orb, the measurements of which a processing module uses to calculate the orb's new velocity and direction in real-time. This information is seamlessly integrated with motor control of an orb-collection-device, otherwise known as a major league shortstop, allowing him to anticipate the bad hop, and prevent the ball from escaping into the outfield.
Among mammals, primates and people, in particular, have the really premium quality vision due to 1-2 million rod and cone cells, photoreceptors of our eyes, which allow for the incredible feats performed by our athletes. For the rest of us, as well as being a source of information, our rich visual experience accounts for a lot of our aesthetic appreciation of the world. Sadly, as we grow older, our sight dims noticeably. . We need more illumination to see clearly and the fine print gets fuzzier. An unfortunate minority descend into total darkness, due to trauma, glaucoma, congenital disease or just plain age.
Restoring sight to the blind has traditionally been the province of saints and holy men. Richard Normann, on the other hand, is a bioengineer, and his more pragmatic goal is to restore a "useful visual sense" to the blind, sufficient for reading, perhaps, or walking down a crowded sidewalk. These tasks do not really require the full range of vision that most of us enjoy. "Miracles of real-time visual behavior are performed by the common housefly whose brain is the size of a grain of rice," observes Christof Koch of the California Institute of Technology. A fly gets by on about 10,000 picture elements (pixels) in its visual field (a pixel is equivalent to one of the dots in your TV screen). Experiments utilizing pixelized goggles suggest that as few as 625 pixels is sufficient for trained subjects to read text or navigate mazes. With that in mind, Normann and his colleagues at the University of Utah are trying to develop a neural implant they hope will deliver an adequate number of pixels to blind patients. The system consists of a tiny camera (video encoder) mounted on a pair of spectacles which sends signals through a signal processor. The processed stream of signals pass through a "transcranial interconnect" to an array of electrodes mounted inside the skull along the visual cortex at the back of the head. The electrode array would be 32 rows by 32 rows in a square, enough for 1024 pixels. The "transcranial interconnect" , cyberpunks, is just what it sounds like, a jack or a plug mounted in the skull. Though a jack is the most efficient, fail safe method now available, Normann hopes eventually to replace it with some sort of telemetry device, operating on radio waves.
Stimulation of the visual cortex via a single electrode in a blind man yields the perception of a spot of light, a phosphene. (Phosphenes are what we appear to "see" when we press on our eyeballs with the lids closed. Continued pressure leads to quite a light show.) The phosphene maps to a particular spot in perceived visual space which is stable over time, according to work done by Terry Hambrecht and co-workers at the National Institutes of Health. They kept a blind woman volunteer wired for four months while they conducted experiments. Using a larger array of electrodes, Normann hopes to extend these results, so that simple patterns can be recognized. Certainly this is not vision as we know it, but perhaps evoked phosphenes are perhaps sufficient to alleviate some of the isolation felt by the blind, and give them a little bit more autonomy. Normann expects that the feasibility of the more refined visual prosthesis he has designed will be demonstrated within the decade. A proof of concept demonstration would show that "patterned stimulation results in patterned percepts," for example, if a volunteer is shown a vertical line, then he or she should perceive a pattern of phosphenes arrayed in a vertical line. Recognition of letters would be more difficult, Normann concedes, but with training of the volunteer and tuning of the device, he believes that success is possible.
Joseph Rizzo has another sort of visual prosthesis in mind, mounted not in the back of the head but within the eye, itself. Trained both in neurology and opthmalogy, Rizzo says that he was in the perfect position to recognize the importance of a singular insight--in degenerative diseases of the eye which lead to blindness (the most common of which, macular degeneration, is essentially caused by aging of the retina), it is only the photoreceptor cells, the rods and cones that are damaged. The rest of the electrical circuitry leading from the eye to the brain still works. It just lacks stimulation. Therefore, Rizzo reasoned that the rods and cones could be replaced with electronic versions thereof. In this way, most of the visual processing done by the brain could be retained.
Rizzo, who is from Harvard Medical School and John Wyatt, an electrical engineer from the Massachusetts Institute of Technology cofounded the Retinal Implant Project to develop a practical prototype from Rizzo's idea. In its current implementation, the retinal implant contains a photodiode array mounted inside of the eye on the front of the retina, opposite from where the normal visual image is focused. The photodiodes are stimulated not from the outside world directly but by an amplitude modulated laser beam coming from a miniature camera mounted on a pair of spectacles. The laser beam also supplies the power to the electrical components in the eye (less than a quarter of a milliwatt is required for the current model). Output from the photodiodes would be circuited through a "stimulator chip" which would direct current into electrodes connecting directly to the ganglion cells, a million of which are the source of all visual input to the brain. Compared to Normann's cortical implant, the retinal implant has the additional advantage of leaving no wires or jacks hanging out of the skull, with the consequent risk of infection.
So far the work of the Retinal Implant Project has been directed toward developing electronics, figuring out the physiological parameters of electrical stimulation, and solving the not trivial problems of operating within the eye. All work so far has been done on rabbits or other animals. The results have been impressive enough to inspire competition; at least three other groups are involved now in similar projects including a four-year, $14 million dollar effort sponsored by the German government. Did Rizzo get any of that money? "Not a penny." he says. The Retinal Implant Project is funded solely by private foundations and donors, some of them Rizzo's patients. In the midst of our discussion of money, Rizzo demonstrates some nervousness:
"It would do incalculable damage to the project," Rizzo warns, "if your story sensationalized or misrepresented the work done here. In particular, I would not want to falsely raise the hopes of the blind that a solution was near at hand."
The quality of vision supplied by the retinal implant is hard to gauge prior to human experimentation. Initially, it might be little greater than simple pattern recognition. In time, however, the simple photodiode array in the retinal implant might be supplanted by a "neuromorphic silicon retina" being developed at the California Institute of Technology, among other places. The first beneficiaries of this "vision chip", however, are likely to be machines. Machine vision has been an intractable problem-- despite myriad advances in digital signal processing, it still is difficult to design a robot that "sees" well enough to navigate through a cluttered room. It will be a long time before a mechanical "orb collection device" tries out for the major leagues. Although electronic systems can be designed with virtually unlimited resolution, how does the robot's digital brain know how to connect the dots, to make meaningful shapes and contours out of the pixelated input?
Biological vision utilizes massively parallel analog processors, usually called nerve cells, which combine output to extract from the visual field such details as edges, local contrast, and movement. Odd as it may sound, we do not see the world as it really is. Using only a fragment of the total electromagnetic spectrum available, from reflected wavelengths we impute to objects completely subjective properties like color and texture. Unlike a digital camera, our pixels "talk to each other." Long before the visual input gets to our brain, the data have been processed to make edges edgier, and contrast more visible. Movement is given special attention, especially in our peripheral vision. Neuromorphic chips try to implement this biological signal processing in silicon.
The advantage of designing neuromorphic chips according to Tobi Delbruck, "is that it forces you to adopt an efficient solution, in terms of processing power and silicon. The brain doesn't want to use a lot of wire, either." In addition to his exceptional repertoire of card-tricks, Delbruck is known for designing state-of-the-art "adaptive photoreceptors". Unlike a camera which whites out the image in the presence of too much light, the eye is capable adjusting to logarithmic changes in light intensity. So too with Delbruck's neuromorphic photoreceptors which uses silicon transistors to accomplish the same thing.
In the silicon retina, logarithmic photoadaptors are connected to a two dimensional analog grid, using design principles of the natural retina. Like biological systems, the silicon retina can sharpen edges and enhance features that are in shadow. Though the resolution of silicon retinas are still relatively low, they are perhaps sufficient for security applications such as recognizing faces or fingerprints, according to Christof Koch. Ultimately, he believes, "neuromorphic systems can provide a "natural" substitute for damaged parts of the human nervous system, such as the retina..."
Because he maintains several web pages devoted to artificial vision, Tobi Delbruck is regularly contacted by people who want to know when the silicon retina will be available for implantation in humans. These are early days yet. Educated observers do not expect even a prototype visual prosthesis until about the year 2010. Even so, it appears that there is an ample supply of volunteers waiting to be the first humans with bionic eyes.
Return to the home page of Steve Alan Edwards