Wireless is getting under our skin
Wireless microsystems are the next volley in the health-care revolution. These wearable devices improve patient monitoring in and out of hospitals, enhance medication delivery, and serve as lifelines for the elderly.
Most microsystem components are custom made in tightly controlled manufacturing facilities, not in research labs. The Johns Hopkins Univ., Univ. of Southern California, Univ. of Utah, and Univ. of Michigan are doing basic physiology work and developing hardware. Some of the companies in the forefront are Medtronic Inc., Minneapolis; Guidant Corp., St. Paul; Cochlear Corp., Australia; Advanced Bionics Corp., Sylmar, Calif.; and MedEl Corp., Austria.
Wireless microsystems that measure arterial pressure and flow can be built into stents. Researchers are exploring pressure sensors that run at nanowatt levels and are small enough to fit inside the eye. And chips that analyze DNA are in the works. Experts predict the next 50 years will bring progress in biotechnology resembling that of microelectronics in the last half century.
Though most commercial systems use wire electrodes, virtually all emerging electrode designs use some form of silicon chip. Silicon technology means high-resolution lithography and the ability to record from many sites simultaneously. This is essential for most next-generation prosthetics and neuroscience applications. Embedding electronics with the electrode will also reduce lead count, a major problem in more advanced systems.
A wireless link, usually inductively coupled RF telemetry, transmits data out and receives commands. In many neural devices, outside sources supply power to the implant via the same RF carrier. Wireless is essential in virtually all implanted biosystems.
A million or more pacemakers and about 125,000 cochlear implants have been implanted worldwide, plus some 30,000 deep-brain-stimulation (DBS) systems for Parkinson's tremor. Research on retinal implants for the blind and cortical implants that treat paralysis is ongoing.
CONNECTING WITH THE NERVOUS SYSTEM
A typical neural implant consists of an array of electrodes that works with the nervous system, either by recording neuronal activity (recording) or by electrically stimulating them. Electrodes connect the electrochemical functions within the tissue and the electronic system. A circuit chip with site selection, amplifiers, and multiplexers works with some form of signal processing/embedded computing. Finally, a wireless link usually handles bidirectional data and power input.
Implanting neural implants in the brain itself generally requires electrode sites every 200 µm or so for recording, and perhaps every 400 µm for stimulation. In cochlear electrodes, sites are on 250-µm centers, consistent with about 128 sites in the human cochlea. Generally, neural implants either record brain signals or stimulate the brain, but scientists are developing implants that could do both.
A wide variety of high-density microelectrode structures are used in the central-nervous system. Most have a silicon substrate. Others use metal foils and polymers.
The nervous system has similarities to microelectronics. Neurons, which are specialized nervous-tissue cells, form complex networks that perform sensory and other physiological functions. Neuronal-cell bodies take inputs from other cells, launching spike discharges to stimulate other cells. Closely placed artificial electrodes sense voltages generated near the cell body during depolarization (a decrease in potential absolute value) of the cell's membrane. To explore changes over time, as in the case of neural prostheses, it is important to record simultaneously from dozens or possibly hundreds of cells over a period of years or even decades.
Until recently, neurophysiological studies used hand-assembled arrays of microwires insulated by polymers, except for 50 to 100- m2 areas at the tips, which record biopotentials. These arrays are relatively coarse and must be moved around to explore neural circuitry in depth. This creates lead problems and considerable tethering and tissue damage, a serious problem for longer-term (chronic) implants and most prostheses.
High-density electrode arrays are of great interest for neurophysiology and neural prosthetics. High-density mapping of the brain is yielding information on how the brain processes information and, in turn, is shedding light on the nature of various neurological disorders. For example, scientists are working to understand short and long-term memory formation in the hippocampus. They hope neural prostheses will help overcome other conditions such as deafness, blindness, epilepsy, Parkinson's disease, and paralysis.
Cochlear prostheses for the deaf are among the most successful neural prostheses to date. They let many profoundly deaf persons function normally in a hearing world. But cochlear implants are limited by the number of electrodes that fit in the cochlea.
Using 16 to 22 wires inserted in the cochlea of the inner ear, cochlear prostheses analyze the frequency content of sound and generate 100 to 1,000- A currents to stimulate receptors in the auditory nerve, replacing defective hair cells that would normally transduce sound. A device like a hearing aid, worn behind the ear, handles speech processing, and an inductively coupled RF link relays power and frequency information to the implant.
Neural prostheses for DBS have been nothing short of miraculous. A device resembling a pacemaker drives the prosthesis, and a four-electrode probe masks tremor. Of the tens of thousands of DBS patients worldwide, few, if any, experience side effects. The electrodes must be positioned closer than 1 mm, at a depth of several centimeters. Missing the target area can result in speech, balance, or gait difficulties. Present systems have been in use more than five years with no loss of efficacy.
Other devices in the pipeline include retinal implants for the blind, cortical implants for paralysis, and implants for managing epilepsy. Driven by solid-state imagers, retinal implants stimulate the optic nerve. Human volunteers have identified simple objects using only 4 4-site arrays. Larger arrays are in development.
Implants that record signals from the motor cortex could provide a front end for functional neuromuscular stimulation, offering hope to the paralyzed. Someday, electrode arrays could detect developing seizures and suppress them even before the patient senses them. All these wireless, implantable microsystems could come in the next decade.
Electrodes placed in living subjects are typically recognized as foreign bodies and encapsulated by astrocytes (star-shaped cells). Usually, the encapsulation is only a few microns thick and current passes through with little effect on stimulation. But it can degrade recording quality over a few weeks. Without encapsulation, sites can "see" cells at a distance of 40 to 100 µm from 150-µm-wide electrode shanks. The reason for this degradation may relate to the tethering of the probes to the skull and the movement of the brain within the skull.
MICROPOWER SIGNAL PROCESSING
Neural signals usually run from tens to hundreds of microvolts in amplitude, with frequencies extending to about 10 kHz. There is no way of knowing in advance where to position electrode sites near neurons of interest. Probes with on-chip electronics interface with sites through selectors that let the user choose a subset of sites to monitor or stimulate. Concentrations of eight or so are common. This compensates for any probe movement in tissue over time.
Selected channels are fed to amplifiers that are usually capacitively coupled. They boost signal levels by 60 dB, operate from 60 to 80 µm in <0.1 mm2, and have significantly less noise than the thermal noise from the site itself. In some cases, the lower cutoff frequency is programmable to record of low-frequency waves in addition to recording neural spikes.
Output multiplexers are sometimes used to time-multiplex the signals from several channels onto a single output lead, reducing the number of leads from dense multi-channel arrays. Lead count is one of the biggest problems in such systems. Output buffers are also important in making signals immune to leakage and noise externally coupled onto the output leads.
The use of dozens or hundreds of sites can quickly exhaust the available bandwidth in inductively coupled stimulation/recording systems. This, and the development of totally implantable microsystems, will require in-vivo interpretation of neural events and proper responses. And in-vivo neural processing chips for spike recognition are already here.
Wireless microsystems may be getting under our skin but millions of disabled persons aren't complaining.