Replaceable Eyeball -

Replaceable Eyeball


'Bionic eye' builds on prostheses milestones with an implanted artificial sight organ as the goal

Eye-implant technology is fighting a tough battle in the human body. Germs, infections, caustic body fluids, nerve degeneration, toxic reactions to chip-generated heat and the conflicting needs for high speed and ultralow power conspire against it. Nevertheless, scientists worldwide are hard at work on various projects to cure blindness, making it likely that many of these problems will be solved within the next 20 years.

Pure medical science also promises solutions within 20 years-specifically, the ability to regress the DNA in dead retinal cells so that they regenerate themselves. Within 10 years, medical science also promises to perfect transplanting living retinal cells harvested from organ bank donors.

In case that doesn't happen, the electronics industry is promising prostheses that outperform “original equipment” with every imaginable enhancement-from zooming to infrared to textual annotation and memory augmentation.

But before these enhancements can be realized, two huge engineering problems face retinal implants, said Gene Frantz, principal fellow at Texas Instruments Inc. One is their interface to neurons; the other is the heat they generate. “How do you create a wetware interface-how do you connect electronics through the skin and inside the eye without introducing germs, paths for infection and other similar problems?” Frantz asked.

“This is not a straightforward microelectronics problem, because you have to mimic biology in order to give the brain a better chance at 'seeing' with an artificial retina,” said Mark Humayun, director of the National Science Foundation-sponsored Biomimetic MicroElectronic Systems Engineering Research Center at the University of Southern California. “Everything must be customized so it can mate with neurons-the packaging, the electronics, the software and everything else.”

Density milestone
Humayun is working on an artificial retina project that shows great promise. “We have just developed the highest channel-electrode density per unit area in the entire microelectronics industry,” said Humayun. “Within an area smaller than 5 millimeters [square], we have been able to package the microelectronics to drive up to 60 electrodes [pixels]-nearly four times as dense as our previous 16-electrode retina.

Humayun is also a USC professor of ophthalmology, biomedical engineering and cell neurobiology, as well as the associate director of research of the Doheny Retina Institute of USC's Keck School of Medicine-affiliated Doheny Eye Institute.

The retinas Humayun described were created as a part of the Department of Energy's Artificial Retina Project (

In phase one, six recipients were implanted with a 4 x 4 16-electrode retina, allowing patients who formerly had virtually no light perception to differentiate among such objects as a cup, knife and plate.

Biocompatible coatings, or “passivation” materials, were contributed by U.S. National Laboratories, which also participated in the Artificial Retina Project.

Passivation is a medical term that mainly means the material avoids attracting the attention of the body's immune system, which will reject implants that produce too much heat or otherwise disturb normal metabolic functions. With passivation being tested at the Doheny Eye Institute, the 16-electrode retina has now been successfully implanted for over two years.

Optobionics slips its solar-powered silicon artificial retina underneath the normal retina but above the optic nerve. The device's solar cells stimulate dormant photoreceptors in the patient's eye to begin working again.

Model 1 of the retinal chip and the new, 60-electrode Model 2 were developed in conjunction with retinal-implant company Second Sight LLC, Texas Instruments and USC as part of the DOE Artificial Retina Program.

The 8 x 8 electrode array in the new 60-electrode retina really has 64 electrodes, but the pitch was so slight that the engineers had to drop four electrodes. As a result, they were able to squeeze all the necessary vias between the electrodes and the underlying electronics into a 5-mm-square area. The metal electrodes are in the corrosive salt water of the eye, whereas the underlying semiconductor electronics must be isolated from that liquid. Potentially, each of the 60 vias between the electrodes and the electronics could leak destructive salt water into the electronics.

“We believe that our new platform, which now has 60 electrodes, is a big breakthrough in density because it also keeps the saline solution in the eye from leaking into the electronics. We believe that in the future we can use this platform to go to 300 electrodes or more with no leakage problems,” said Humayun.

TI's Frantz is helping pioneer the replacement of dead photoreceptors on the retina with neural software running on TI's TMS320C6000 DSP. In DOE's artificial-retina project, TI's DSP stimulates the underlying ganglion cells of the eye with a neural-like spike train designed to match that of a normal retinal cell.

The Dobelle Institute is the only artificial-vision provider that directly installs an implant into the visual cortex of the brain, bypassing the eyes altogether.

According to Frantz, eye implants are following the path of cochlear implants, which have a 20-year development lead over eye prostheses.

A successful cochlear implant company, Advanced Bionics, was founded by Al Mann, the same individual who founded Second Sight. “Second Sight is using, as its first approximation, the connection system originally designed for the cochlear implants,” said Frantz.

Cochlear implants solve the wetware interface problem by using two parts-one inside the body and one outside-that communicate wirelessly through the skin. “The cochlear implants are so good that after the operation, you can't tell the difference between a normal child and one born profoundly deaf,” said Frantz.

Similarly, the 16- and 60-electrode artificial retinas also use two parts, with most of the outside electronics contained in a belt pack connected to a video camera mounted on a pair of eyeglasses. The camera sends a megapixel scene to the belt pack, where the DSP encodes the image into a pulse train similar to that of the natural retina. A high-radio-frequency (RF) transmitter then transfers the coded signal to a receiver on the implanted artificial retina, where platinum electrodes stimulate the ganglion cells below the dead retina. (Simultaneously, a low-RF transmitter sends a constant ac signal to the chip, which it rectifies into the dc to power the chip.)

Retinal prosthesis implant shows the imaging camera at bottom (on glasses frame), transmitting power and information via a loop antenna to modules in the eyeball.

“Ultimately we'd like everything integrated, possibly as flexible, foldable electronics with a MEMS [microelectromechanical system] lens so that the whole assembly could be inserted through a sub-1-mm incision that would expand in the eye and give you enough resolution to be able to read,” said Humayun.

“The hope is that all the electronics, including the DSP, will eventually be inside the eye. It needs to be inside because of the constant motion of the eye as it 'paints' the scene for the brain. If all the electronics is inside the eye, it will naturally move around to paint that visual scene for the brain,” said Frantz.

The countdown
The goal-a total replacement for dead retinas-is quite distant. The heat generated by a DSP alone is about 1,000x more than the eye can tolerate. The eye can only handle microwatts, according to Frantz, but today even the lowest-power DSP dissipates milliwatts.

“In today's low-power designs, engineers minimize power after achieving their performance goals. What we need here is what I call ultralow power, which means we try to achieve the absolute lowest power and then measure how much performance we get from it,” said Frantz.

To predict when DSPs will be at the microamp level, Frantz refers to a chart named after himself: “Gene's Law” states that power goes down by half every 18 months. That would mean that reaching the critical threshold for direct implantation would take 10 years-from milliwatts per million instructions/second (Mips) today down to microwatts per Mips.

Until then, DOE's Artificial Retina Project will probably stick with the two-piece scheme to keep the electronics that generate the most heat outside of the eye.

All electronics, including the digital signal processor, will eventually be inside the eye, but only when DSPs are at the microamp level. 'Gene's Law' of power goes down by half every 18 months, meaning that reaching the critical threshold for direct implantation would take 10 years—from milliwatts per million instructions/second today down to microwatts per Mips.

“Today our software has to carefully manage the power budget of the implant. We know exactly how much heat we can put into the eye, and we know exactly how much heat each part of our circuitry will produce. Our basic strategy is just not to power everything on the chip at the same time,” said Humayun.

The Doheny Eye Institute is currently doing preclinical lab testing on the new retina with saline soaks and implantations into blind animals. Humayun plans to implant the first 60-electrode artificial retina in a human subject next spring.

Model 1 recipients, the oldest of which has had the implant for over two years, continue to use cameras mounted on a pair of eyeglasses, but because of the seriousness of the operation, none will receive Model 2 retinas until the Doheny Eye Institute has evaluated how much better the new Model 2 works.

“If you had a 16-electrode device, you'd want to know how much better the 60-electrode device was before you considered trading up. And we just won't know until we implant a few people,” said Humayun.

“What we don't know is how much is filled in by the brain. Our device is crude, considering there are over 100 million photoreceptors in a normal eye, but we have found that the brain has an incredible ability to fill in missing information,” Humayun said.

For instance, there is a huge blind spot in normal eyes where the optic nerve exits the back of the eye-about 1.5 mm in diameter. That should produce a hole right in the middle of everybody's field of view, but the brain fills in that hole so that it is virtually undetectable. Humayun is hoping that this same mechanism can be harnessed to fill in the missing information between the electrodes.

“We hope to get enough information from the 60-electrode model to gauge how many pixels we would need to really make a difference in people's lives. We don't just want to double or triple the number haphazardly. This is not a technology problem; we are trying to help people. We want to make a judgment on how many pixels people need to have acceptable vision. Will it be 256 or 512 or 1,024 or 10,000? We just don't know yet.

“Our first goal is to give people enough resolution for mobility, which we think we can do with the current design in about five years. At the same time, we want to develop a new device with enough resolution to enable reading in about five to seven years. And we are doing all this in parallel with developing less-invasive surgery techniques and better passivation. For instance, the current designs are designed to last at least a decade, but the reading-resolution retina will probably have a lifetime of 20 years,” said Humayun.

There are many other parameters to this search as well. “When you have 1 million neurons firing, how do they fire in concert? Is it a frequency encoding, an amplitude encoding or a combination-or are some cells 'on' and some cells 'off,' and are some cells giving you edge information? How do you compile all that into a code that lets the brain say, 'Now I can see it?' We hope the brain will meet us 80 percent of the way, so that we only have to do 20 percent of the encoding,” said Humayun. “There will probably always be some personalization required, just like when you go to get your glasses. But instead of switching lenses, we are changing the algorithm our software uses to encode the signal going to your electrodes.”

With Humayun's design, augmentation becomes trivial. For instance, to be able to see light in the infrared, as the creature does in the movie Predator, all you would do is substitute an infrared camera for the standard video camera. Because the camera is external to the current design, almost any type of input device can be substituted, enabling you to “see” smells, for instance. “Black and white takes you a long way, but we could feed any spectrum into our device-infrared, ultraviolet, anything,” said Humayun.

Others advocate such augmentation as an integral feature of the system. Why, after all, make an eye just an eye when it could also just as easily see through walls or sense heat or even lies?

“We are building a thermal-imaging array that is only 5 millimeters diagonal [2.2 mm square], which could allow people to see thermally, or to be used in combination with a visual imager to provide more information than a normal eye,” said Gary Havey, president and co-founder of Advanced Medical Electronics Corp. (AME;

According to Havey, the smallness of AME's infrared sensors-called microbolometers-could enable an implant chip to augment low-vision patients with new information that compensates for vision loss. Thermal imaging, for instance, could substitute for visual information, because the temperature of an object is related to its color, with smoothly changing contours and shadings that match the object's shape. Also, some objects that are hard to discern with visual imaging are easier to see with thermal imaging, such as hot food on plates and cold liquids in glasses.

“There are a lot of interesting things that are identifiable by heat- the location of people in a room, or the different items on a table, since they are usually either hotter or colder than the rest of the room,” said Havey. “Everything is different colors, everything is shaded, everything is lighted differently. When you look up at the ceiling, the lighting shows you the size, shape and location of walls in a room. Exterior doors are easy to see because they are typically cooler in winter and hotter in summer than the rest of a wall. All this could be very valuable information for people with little or no normal vision,” he said.

Thermal isolation
Microbolometers are MEMS devices, each one of which is thermally isolated from its neighbors. They are packed in a wafer-scale package that works at room temperature (unlike conventional thermal imagers, which require cooling to a specific temperature). Many companies are working on thermal imaging today, but AME, which is not developing a full-fledged retina, believes it has patent coverage for the application of thermal imaging as an avenue for artificially generated sensations.

In place of camera-like pixel-imaging arrays-visible or infrared-one company is pitching a solar-powered retina that requires no external power source and can squeeze up to 5,000 microelectrode-tipped microphotodiodes into a 2-mm-diameter silicon chip. Incident light produces electricity that stimulates retinal cells that were once thought to be dead, bringing them out of dormancy and restoring normal vision. The solar-powered retina is slipped underneath the normal retina, where it can give an energy boost to existing retinal cells.

“We have implanted our artificial silicon retina into 10 patients for four years so far, and in every case the patients have had moderate to very substantial improvement in their vision,” said Alan Chow, co-founder of Optobionics Corp.

The least helped were patients who were almost totally blind, though even that group has now begun to see the world as shades and shadows. The most helped, so far, are patients who were already perceiving the world as shades and shadows; they are now able to read an eye chart. One patient who has not been able to do so in 20 years can now read an alarm clock.

So far, the FDA has only permitted the solar-powered retina to be implanted in patients suffering from the early stages of retinitis pigmentosa-a general term for a number of diseases that affect the photoreceptor layer of the retina. These include Usher's syndrome, Leber's congenital amaurosis, Stargardt's disease, cone-rod dystrophy, Best's disease, choroideremia and gyrate atrophy. Each of those diseases was present in at least one of Optobionics' test subjects.

“With our retina there are no wires, no batteries, no heat generation, no external computers, none of the problems associated with other implants,” said Chow. “The most exciting finding is that in half the patients, our chip has helped to regenerate nearby retinal cells that were thought to be dead but were only dormant. It's like turning back the aging clock by 10 to 20 years. These patients are experiencing a return of color vision, of resolution, of depth perception, of everything that we associate with normal vision.”

Patients whose cells are not regenerating see flashes of light only where the chip was installed beneath their retina. But, to explain the cases where normal vision seems to be returning, Chow believes that before retinal cells die off completely from one of the aforementioned diseases, they become dormant.

Thus, if installed early enough, the Optobionics solar-powered chip could not only extend the lifetime of normal cells but also reinvigorate the cells gone dormant, thereby restoring normal vision.

“We hope that we have discovered a common pathway of regeneration that will also be useful for the treatment of the much more common problem of age-related macular degeneration,” said Chow.

The company expects it will receive FDA approval to implant the chip in sufferers of macular degeneration, a widespread condition that affects more than 40 million people worldwide.

Today, the industry dedicated to curing blindness only nets a few million dollars annually. As the industry's technology advances to cure more kinds of blindness, however, it will eventually be measured in billions of dollars-perhaps before the end of the decade.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.