ISSCC: Papers outline biochips to restore eyesight, movement

ISSCC: Papers outline biochips to restore eyesight, movement
SAN FRANCISCO — The possibility of using RF-powered electronic implants to stimulate the nerves of paraplegics and the retinas of those with certain eye diseases arose this week in a session on emerging technologies at the International Solid-State Circuits Conference, here.

Jeffrey A. Von Arx presented what was his doctoral thesis at the University of Michigan-a fully integrated neuromuscular electrical stimulation system, called Finess, used to stimulate nerves in paralysis patients. Use of a wireless system served to eliminate the leads that tend to break off or damage nerves when implanted.

Separately, Mark Clements of North Carolina State University described a retinal stimulator being developed in a project with the Wilmer Eye Institute of Johns Hopkins University. The idea is to restore eyesight lost to macular degeneration or retina pigmentosa, two common conditions that cause failure in the rods and cones that act as the eye's photoreceptors.

The two projects had several things in common. Both used biphasic waveforms, for example, but they had different parameters. Von Arx stuck to particular figures for the amplitude and period of the waves, although both values were left programmable. Clements, on the other hand, found different parameters worked better for certain patients and therefore hasn't yet settled on particular values.

Finess is powered by a transmitter the patient wears. The receiver, implanted in the body, uses a silicon-rubber cuff that wraps around the nerve. Inside is a system-on-a-chip that includes an integrated receiver coil, a change from the discrete coils previously used. Mimicking the long, stretched shape of nerve cells, the 3-micron BiCMOS chip measures 2 x 8.7 mm; researchers felt a traditional square chip would anchor the nerve cell down, damaging it.

The receiver coil spirals around the chip in 17 loops, making maximum use of die area. The coil had to be designed with a high coupling coefficient but with low resistance, Von Arx said, leading to the use of a nickel-iron layer over the silicon substrate, with electroplated copper used for the coil itself.

Dealing with human tissue became a design consideration for Finess. The transmitter's effective range is only 3 cm under the skin, for example. And the telemetry between transmitter and receiver works best at a frequency of around 4 MHz, Von Arx said, as higher frequencies begin to get muffled by surrounding tissue.

The receiving chip requires a minimum 12 mW of power. Of that, 8 mW goes to the nerve output, 2 mW powers the circuitry and an additional 2 mW is lost in RF-to-dc conversion.

In the next paper, Clements presented a neurostimulator, again controlled with RF signals, to be embedded in the retina. Work at Johns Hopkins has succeeded in partially restoring vision using 5 x 5 arrays of retinal electrodes; Clements' paper discussed a chip that can drive a 10 x 10 array, and he's hoping to take the technique up to 40 x 40 arrays.

In this case, the transmitter is a single-chip, quarter-VGA camera mounted on a special pair of eyeglasses. It transmits information to the implanted neural stimulator, which in prototype form measures 4.6 x 4.7 mm in a 1.2-micron CMOS process.

The receiver was left programmable, because research is still sorting out requirements for certain types of vision. "If a patient wants to read a newspaper, it might turn out that kind of vision is different from what you need walking down the street," Clements said.

A data rate of 30 kbits/second is the minimum to produce useful vision, with each visual frame taking 400 clock cycles to construct, Clements said. The carrier frequency varies from 1 to 10 MHz, a range left purposely wide to accommodate future design decisions, as the device specifications are still the subject of experimentation at Johns Hopkins.

The stimulus circuit ends up using a wide range of output voltages, a fact reflected in the other specifications: less than 600 mA for each electrode, but an impedance on the order of 10 kohms. One handicap of using CMOS is that voltages are likely to climb as the design advances; the integration of high-voltage output devices will probably be necessary in the future, Clements said.