Bridges to the brain

FOXBOROUGH, MA.—Imagine being a prisoner of paralysis, unable to move your arms or legs and totally dependent on others for everything from taking nourishment to turning on a TV set. Yet, at the same time, your mind remains as sound as ever.

Hundreds of thousands of people worldwide—a majority under age 30—suffer that fate as a result of spinal-cord injuries or progressive neurological diseases, such as Amyotrophic Lateral Sclerosis (ALS), commonly known as Lou Gehrig’s disease. It’s a population that has been largely neglected by a booming neurotechnology industry, which is using implanted electronic stimulators to address conditions ranging from chronic back pain to deafness to Parkinson’s disease.

Now, however, Cyberkinetics Neurotechnology Systems, a Massachusetts startup, has earmarked those with severe paralysis as its target population for electronic devices designed to accomplish what some might say is impossible. Its implanted Andara Oscillating Field Stimulator (OFS) System promotes tissue growth in the vicinity of a spinal injury, thereby restoring some sensation and movement (see “The quest for regeneration”).

Perhaps even more incredible, the company’s BrainGate Neural Interface System harnesses signals from the brain to allow severely motor-impaired patients to control a computer, appliances, or a prosthetic arm.

“Cyberkinetics is doing great things and deserves a lot of credit for focusing on a market that other companies have bypassed as being too small,” noted James Cavuoto, a biomedical engineer who publishes Neurotech Reports, an industry newsletter. “This company has looked at a segment of the disabilities community that is most in need and is making significant progress.”

Those gains, however, have come only after millions of dollars in investment and years of research for the publicly held firm, including a long litany of development tests, many of them homegrown, to validate the systems.

Experts both within and outside Cyberkinetics see the BrainGate system as by far the company’s most challenging from a design and test standpoint. It begins with the design and test of a reliable three-dimensional electrode array, implanted in the region of the cortex associated with motor control function.

BrainGate could establish a new line of communications for severely paralyzed individuals who have been cut off from the world, said John Donoghue, the Brown University neuroscientist who founded Cyberkinetics. Courtesy of Cyberkinetics.

The current array, now being implanted in patients in clinical trials, consists of 100 platinum-tipped silicon electrodes, insulated with Parylene C. Individual electrodes, 96 of which are available for recording neuronal activity, are 1 mm long and spaced just 400 μm apart. A tiny wire bundle from the electrodes connects through the skin to a titanium pedestal anchored to the exterior of the skull.

The pedestal, which contains amplifier and signal-conditioning hardware, connects to a cable that carries the amplified neural signals to computers resting on a cart near the patient. Future designs, now under development, will feature infrared wireless technology linked to a more compact computer, mounted on the patient’s wheelchair.

The challenge in all this is twofold. First, the engineers must obtain good quality signals from the electrode array. And second, they need to devise computer algorithms or “filters” to decode the neural spiking from the electrodes, which appear as waveforms on the technician’s computer screen. The goal is to screen out noise and convert those brain signals in real time into control signals that will allow a patient to move a computer cursor, click on an icon that will turn on an appliance, or even maneuver a robotic arm.

Patients have demonstrated all these functions in clinical trials.

“We are asking the system to deliver 30 kHz of signal for 96 channels, so there is a lot of processing,” noted Donoghue, the company’s chief scientific officer and director of the Brain Science Program at Brown University. “One of our biggest challenges is decoding the signals to convert them from a stream of impulses into a command signal. We have to process 96 channels of data, signal condition it, discriminate the signals, and then interpret the impulse—all in less than 150 ms.”

Critical to this operation is knowing how each of the electrodes is functioning. To address that need, Andras Pungor, the lead electrical engineer at Cyberkinetics’ development and manufacturing facility in Salt Lake City, designed what Donoghue describes as a “very elegant” automatic impedance tester.

Lead electrical engineer Andras Pungor designed an automated impedance tester that played a key role in speeding development of the BrainGate system. Within seconds, the tester can identify leaks, shorts, or connection problems in the 96-channel electrode array.

By contrast, his new patented automatic impedance tester can measure current in the range of 10 pA and voltages as low as 5 mV. Moreover, results of this impedance test of all 96 channels are graphically displayed on a monitor in just 10 s.

Hardware consists of a 1x1-in. circuit board embedded in the cable that connects to the pedestal mounted on the patient’s skull. To take a measurement, a technician simply flips a switch on the cable, sending a 1-kHz signal at 5 mV. In contrast to using a passive impedance tester, operators do not have to disconnect the patient from BrainGate to do the test.

During the test, the graphic display shows a map of the array system, with each channel appearing as a block with impedance values. “It is a snapshot, and you can set up desired parameters,” explained Pungor. “For example, you can ask the system which electrodes are within specific ranges, such as between 200 and 400 kΩ. The channels within that range will show up as black, those above the range in red, and those below the range in yellow. When you are looking at 96 channels at the same time, it is much easier with this color-code system.”

With the automated impedance tester, operators can quickly identify which electrodes are operating within specific ranges, such as between 200 and 400 kΩ. The channels within that range will show up as black, those above the range in red, and those below the range in yellow.

For human clinical use, the acceptable operating range for electrodes is between 100 and 800 kΩ, added Pungor.

Members of the Cyberkinetics team agree that the impedance tester has been a vital tool both in developing the system and in facilitating sessions with patients using the system in clinical trials. For example, the tester can quickly identify problems such as leaks, shorts, or poor connections in the array.

“The quality of the circuit may change because of all sorts of factors, including tissue response and material changes in the device itself,” noted Donoghue. “This impedance tester is an extremely sensitive and extraordinarily valuable tool to insure that the circuit is intact and performing as it should.”

The impedance tester is just one example of many “homegrown” test and measurement devices and methods created by the Cyberkinetics staff. Pungor recalled that early in the development of the electrode array, there were no off-the-shelf high-power amplifiers for 100 channels. “You had racks of amplifiers, and there were a lot of ground problems and noise problems. So we said, 'Why don’t we build an artificial electronic neuron source to simulate every channel?’”

What the engineers came up with was a signal generator with 128 channels and capable of putting out neural type signals in the range of 120 to 150 μV at a 200 kΩ impedance. Each channel in turn could generate four different kinds of signals each second.

“This was a very valuable tool in troubleshooting as we developed the array,” said Pungor, “versus having to do testing in living animals.”

In clinical trials, patients with the BrainGate implant have performed a number of functions, such as moving a computer cursor, clicking on an icon to activate appliances, and maneuvering a robotic arm.

As the technology evolved, the engineering team developed a PC-based system, called the motor cortex simulator, featuring 100 channels of firing neurons and a touch screen that let researchers alter the modulation of the neurons. One PC was used as a function generator, simulating the firing of the neurons, and two others served as signal processors.

“This tool allowed us to validate all parts of the system, both hardware and software,” explained Donoghue. “It evaluated both the throughput speed and the accuracy of capturing the analog waveforms and processing them into a digital pulse.”

On the software side, Cyberkinetics has relied heavily on tools such as The MathWorks’ Matlab to address many technical issues, and the team created diagnostic software that detects crosstalk between electrode channels. Even more significant, the engineers used software to devise a time-saving measurement tool for sorting the vast amount of neural data pouring into the system from the 96 channels.

Before Almut Branner and her colleagues developed an automatic spike sorter, it could take up to an hour for engineers to classify the neural "spikes" or action potential waveforms from brain activity picked up by the BrainGate electrode array.

Manual methods for detecting and classifying these neural “spikes” or action-potential waveforms had become too time consuming. Using software analysis of waveform shape and amplitude, researchers have to set up different classification rules for each of five waveforms coming from each channel. Classifying these waveforms is an essential preliminary step for devising the software-based rules that enable patients to perform control functions during their sessions with researchers and technicians. Yet this “spike sorting” was taking an hour or more in setup time.

“We needed to automate this process by coming up with an algorithm to create these rules,” explained Almut Branner, a Cyberkinetics biomedical engineer who headed the team that developed what is called “the automated spike sorter.” The new system, devised with the help of Matlab and its signal-processing toolbox, now takes the continuous signal that comes in from the patient and makes decisions about whether the signal is noise or one of the action-potential waveforms. The process now takes just a few minutes.

The automated spike sorter, added Branner, is also an essential step toward the goal of a future BrainGate system that could be operated by patients without the assistance of technicians. “Eventually, the algorithm could perhaps be implemented on a DSP chip that is implanted in the body,” she said. “This would greatly reduce the bandwidth of the signal that we now send out of the body. We collect 30,000 data points per second on every channel, and right now it is almost impossible to do this wirelessly.”

The BrainGate system follows nearly two decades of research, both in New England where Donoghue and his team did 10 years of animal studies in his Brown University lab, as well as in Utah, dating back to Richard Normann’s first electrode designs.

Engineer Rick Van Wagenen does a microscopic inspection of BrainGate’s electrode leads, one of a long litany of procedures performed on the system.

At the Salt Lake City development and manufacturing facility, senior biomedical engineer Rick Van Wagenen cited a whole litany of essential tests that were instrumental both in developing the initial electrode array, as well as other components of neural interface systems that the company sells to researchers around the world. These products range from cables, connectors, and customized electrode arrays to amplifiers and neural signal processors. Among the tests and procedures that Van Wagenen said have been essential to the team’s research and development work on BrainGate:

• scanning electron microscopy for such tasks as studying the morphology of the platinum metal that is sputter-deposited on the electrode tips;
• impedance tests, using such tools as the homegrown automatic impedance tester, as well as electrochemical impedance spectroscopy with gamma ray instruments; such tests help determine a profile of the impedance performance as a function of frequency, a key step in making an electrical model of the neural interface;
• profilometry for measuring the surface roughness and other parameters of metal elements of the implanted array, and ellipsometry for measuring thickness of coatings and films used in the implanted assembly;
• accelerated life tests, ranging from several months to years, of operational arrays in a saline bath at temperatures in the range of 58ºC to 68ºC; a multichannel signal generator provides the pulsing energy to the electrodes; and
• electromagnetic interference (EMI), radio-frequency interference (RFI), electrostatic discharge (ESD), and other susceptibility tests on prototypes, typically performed by outside labs, such as Curtis Straus, in accordance with ISO 60601 standards for implantable electronic devices.

Manufacturing of the company’s neural interface components takes place in an FDA-registered clean-room environment, monitored regularly for humidity, temperature, and bacteria counts. Technicians routinely employ optical microscopes to inspect electrode tips, and a stereomicroscope with photographic attachment is used to take a picture of every electrode array for record keeping.

Packages of array components for implants are sent to an outside vendor for sterilization and are returned with test data certifying the effectiveness of the process. The company even validates its packaging methods for array assemblies by sending samples to vendors, such as UPS Professional Service, for vibration, compression, and drop tests.

With so much effort involved in this ground-breaking neural interface work, the logical question is how long before the company sees a commercial payoff from its showcase system, BrainGate?

Field clinical engineer Abe Caplan conducts a session with an ALS patient, who uses his thoughts to control a cursor on a computer monitor.

Even so, the R&D work is already paying off, not just in sales of research products, but in medical devices that are offshoots of BrainGate. For example, the company has received FDA clearance for its NeuroPort Neural Monitoring System, which uses the BrainGate array to help physicians analyze abnormal brain activity in patients, such as seizures suffered by patients with epilepsy.

“BrainGate is really the basic platform for a whole range of applications, because it is, in effect, a test and measurement device for the brain,” said Donoghue. In epilepsy patients, for example, Donoghue envisions future applications in which the array would first detect the onset of seizures and then activate an electronic stimulus to restore normal brain activity or, alternatively, an implanted microfluidic pump that would inject preventive medication.

Researchers outside the company also are working with Cyberkinetics to link BrainGate with other implanted devices, such as stimulators that enable paralyzed individuals to move their arms and hands. In addition, Cyberkinetics is already designing compact versions of the system that would use infrared technology to transmit the brain signals.

Behind all this effort is the desire to make a difference in the lives of severely paralyzed individuals who are cut off from the world. “This technology promises to give them an output to perform the functions that most of us take for granted,” said Donoghue. “It may ultimately allow them to communicate and connect, which could have a huge impact on people who have gotten a really raw deal from life.”