San Mateo, Calif. - Researchers at the crossroads of medicine and electronics are developing implantable silicon neurons that one day could carry out the functions of a part of the brain that has been damaged by stroke, epilepsy or Alzheimer's disease.
The use of neural chips to replace brain functions is "mostly in an animal-research phase now. Work in humans could be five or 10 years away," said Metin Akay, an associate professor of engineering, psychology and brain sciences at Dartmouth College and chairman of the first international conference on neural engineering, to be held this week in Capri, Italy. But some of the roughly 180 papers to be presented in Capri give tantalizing hints of the potential of this emerging technology.
"We are trying to figure out how to develop a prosthetic that allows one part of the brain to talk to another," said Theodore Berger, director of the center for neural engineering at the University of Southern California and one of the researchers at the forefront of the implantable-neuron effort. "We've done all the pieces of the problem and now we are trying to fit them together," said Berger, who has prepared a paper on his work for the meeting.
Akay called the conference the first "to really unite rehabilitation and computer scientists, electronics engineers and neural scientists." Researchers from 30 countries will attend. "I hope it will further stimulate research in the field," said Akay, who has helped establish and promote the discipline of neural engineering. He described it as "a new concept and a diverse field that spans medicine, physics, computer science, electronics and more."
Applications for this combination of electronics and neural science span neural computing, advanced robotics and improved drug sensors as well as better fundamental knowledge of the brain. "We still do not understand perhaps 80 percent of the functions of the brain," said Akay.
Because the field is so new, it's tough to say exactly where it is headed. "I can't tell what the real applications are yet, but it could be important both for computing and medicine," said Peter Fromherz, a professor of neurophysics at the Max Planck Institute for Biochemistry (Munich, Germany), who will present a plenary paper on his experiments growing neurons directly on a 2-D transistor array from Infineon Technologies (see www.biochem.mpg.de/mnphys/publications/02fro3/02fro3.pdf).
At USC, meanwhile, Berger's team first stimulated a slice of a rat's brain with a random-signal generator to determine its functional patterns and develop mathematical models representing them. The group then coded those models into an analog chip.
USC researchers have spent 10 years working largely with analog devices because they found the continuous rate of discharging capacitors was good for mimicking continuous neural functions. "But if you are going to need tens or hundreds of thousands of neuron models on a chip, analog capacitor models won't scale well," Berger said.
"We're now looking at an SoC [system-on-chip] approach that mixes digital and analog techniques," said John Granacki, director of the advanced-systems division of USC's Information Science Institute. "We are looking at digital blocks such as fixed-point arithmetic units that can generate recursive Laguerre polynomials to provide universal functions." These polynomials, he said, "are very compact and don't require much addition, subtraction or multiplication to generate functions. Meanwhile, simple analog circuits help indicate peaks for neural signals."
The team's next-generation chip, a hybrid digital/analog device, will use 130-nanometer process technology and could form the basis of what researchers envision as a host of chips that run the gamut from fixed-function to highly programmable parts. "We think we can create whole families of devices that can do [neural] pattern matching," Granacki said.
Wake Forest University (Winston-Salem, N.C.) is starting work with Berger's team in a three-year program to integrate all their separate research projects in an experiment on an intact rat brain. A future project on a monkey could take five years, Berger said.
At the same time, Berger is in the middle of reviews to compete for a National Research Foundation grant for a new center at USC devoted to neural engineering. Besides the work on neural prosthetics, that center would include two other testbeds.
One involves injectable neuromuscular stimulators that could operate a paralyzed limb and be managed wirelessly through controls sewn on a sleeve. A separate project, in conjunction with Johns Hopkins University (Baltimore), is working on a silicon-based array of photosensors made on a curved surface that could be fitted to the back of a damaged human retina. Both projects are in clinical trials and could see commercial results in five years, the scientists said.
One of the toughest problems in neural prosthetics is how to connect chips and real neurons. Today, many researchers are working on tiny electrode arrays that link the two. However, once a device is implanted the body develops so-called glial cells, defenses that surround the foreign object and prevent neurons and electrodes from making contact.
"We are working with chemists and materials scientists to figure out how to coat interface devices with a biological or biological-like material that will attract neurons," said Berger-perhaps something sticky to which neurons would adhere. "It's a hard problem."
The Max Planck Institute grew this 'snail' neuron atop an Infineon Technologies CMOS device that measures the neuron's electrical activity, linking chips and living cells.
In Munich, the Max Planck team is taking a revolutionary approach: interfacing the nerves and silicon directly. "I think we are the only group doing this," Fromherz said.
Fromherz is at work on a six-month project to grow three or four neurons on a 180 x 180-transistor array supplied by Infineon, after having successfully grown a single neuron on the device. In a past experiment, the researcher placed a brain slice from the hippocampus of a monkey on a specially coated CMOS device in a Plexiglas container with electrolyte at 37 degrees C. In a few days dead tissue fell away and live nerve endings made contact with the chip.
"Sometimes the [nerve-silicon] coupling is good, other times it is poor," Fromherz said. "We understand the physics, but in terms of the engineering we have very little control. We have to improve both the chips and the cells."
Solving the interface issue lies in letting the brain do much of the work, said James Hickman, a surface chemist and assistant professor in the Hybrid Neuronal Systems Lab at Clemson University (Clemson, S.C.).
Hickman, who has successfully grown systems of neurons using geometric queues, noted that in cochlear (inner-ear) implants, the auditory nerve actually reconfigures itself to interpret the 10-channel signal the implant emits. "Now we have to get more sophisticated in the types of connections we make," he said. "We would like to send out a signal that recruits the right nerves to the right contacts. How you control that reconfiguration is the next big step in neural prosthetics."
Developing a device that has both transistors for recording neural data and capacitors for stimulating neurons is a next step for Fromherz. "With brain slices we do not know yet how many [or how few] nerves we can stimulate and record. We need to learn how many nerves we need to precisely control [to get a desired result], and we are just at the beginning of this learning."
Help for Parkinson's
The U.S. Food and Drug Administration has approved implantable neurostimulators and drug pumps for the treatment of chronic pain, spasticity and diabetes, according to a spokesman for Medtronic Inc. (Minneapolis). A sponsor of the Capri conference, Medtronic says it is already delivering benefits in neural engineering through its Activa therapy, which uses an implantable neurostimulator, commonly called a brain pacemaker, to treat symptoms of Parkinson's disease.
Surgeons implant a thin, insulated, coiled wire with four electrodes at the tip, then thread an extension of that wire under the skin from the head, down the neck and into the upper chest. That wire is connected to the neurostimulator, a small, sealed patient-controlled device that produces electrical pulses to stimulate the brain.
The Activa device was approved for use in Europe in 1998 and in the United States last year. "I have interviewed some of these patients and it is amazing to see them do simple things like drink a cup of coffee or walk-things they cannot do without this device," said Akay of Dartmouth.
Computer chip model of neural function for implanted brain protheses. One problem is rejection at the interface between artificial and biological neurons.