Paralyzed Man Thinks Robotic Devices into Motion

PROVIDENCE, R.I., July 12 -- A man paralyzed from the neck down has been trained to control a computer cursor and manipulate robotic prostheses through a sensor implanted into his motor cortex.

The 25-year-old man, a quadriplegic since a knife wound that transected his spinal cord, can open e-mail and operate a television -- even while having a conversation -- and exert rudimentary control over a multi-jointed robotic arm, reported John P. Donoghue, Ph.D., and colleagues, of Brown University here, and other centers, in the July 13 issue of Nature.

"The broad question we are addressing is whether it's possible for someone with paralysis to use the activity of the motor cortex to control an external device," said neurologist Leigh Hochberg, M.D., Ph.D., of Massachusetts General Hospital and Spaulding Rehabilitation Hospital in Boston.

"There has been a question of how the function of the cortex might change after it was disconnected from the rest of the body by damage to the spinal cord," added Dr. Hochberg, a co-author. "We're finding that, even years after spinal cord injury, the same signals that originally controlled a limb are available and can be utilized."

Three years before the study began, the man, identified in the Nature paper as MN, had suffered the transsection between the third and fourth cervical vertebrae. (A second participant, a 55-year-old man paralyzed after an injury at cervical vertebra four in 1999, has since been enrolled in the clinical trial of the system).

In June of 2004, surgeons implanted a minuscule sensor attached to a wire into the primary motor cortex of MN's brain, in the region responsible for arm control.

The sensor, which contains 96 microelectrodes measuring 1 mm in length and spaced 400 microns apart, records the patients motor intentions and relays the signals through a percutaneous wire to a pedestal on MN's scalp. This, in turn, is connected to signal amplification and conditioning hardware and a cable that relays the amplified signal to a computer, which then translates the signals into meaningful action.

In videos available on the Nature web site, MN can be seen opening a simulated e-mail program and controlling other computer functions, drawing a circle on the computer screen using a paint program, and controlling, although somewhat crudely, the motions of a robotic arm -- all tasks that since his injury had been out of his grasp.

"This is not the first neuromotor prosthetic that has been implanted into a person -- a previous experiment used a couple of implanted electrodes to generate limited horizontal control of a cursor -- but this study reports several significant advances," wrote Stephen H. Scott, Ph.D., of Queen's University in Kingston, Ontario, in an accompanying commentary.

One important finding of this study, Dr. Scott said, is that neural activity in the primary motor cortex can be preserved several years after a paralyzing injury, even though axonal connections from that region of the brain to the spinal cord had been severed, and even though the brain had not been called upon to control limb function for quite some time.

Furthermore, "the calibration of the device was achieved by simply asking the subject to imagine moving his hand to track a moving cursor on the computer screen. This process took only minutes, much less than the weeks or months of training required for current non-invasive electro-encephalography systems," he wrote.

In addition, although implantable sensors are invasive, similar attempts to noninvasively monitor neuronal activity have been hampered by the background noise of millions of neurons firing at will, making it hard to detect and sort out the desired signals, he noted.

The fact that MN was immediately able to exert some degree of control over a computer, television, and robotic devices suggests that it may be possible, using similar "neuroprosthetic" devices, to control motorized wheelchairs and other devices, Dr. Scott pointed out.

The training and calibration took place during 57 sessions spaced over the course of nine months, during which the implanted sensor (BrainGate) recorded motor cortical activity while MN imagined moving his limbs through space, and then imagined using those motions to complete computer-based tasks.

"What's truly exciting is this: the cortical activity of a person with spinal cord injury, controlling a device simply by intending to move his own hand, is similar to the brain activity seen during preclinical studies of monkeys actually using their hands," Dr. Hochberg said. "Whether it is real or attempted movement, neurons seem to respond with similar firing patterns."

"What is also encouraging is the immediate response from the brain," Dr. Donoghue added. "When asked to 'think right' or 'think left,' patients were able to change their neural activity immediately. And their use of the device is seemingly easy. Patients can control the computer cursor and carry on a conversation at the same time, just as we can simultaneously talk and use our computers."

Dr. Donoghue is chief scientific officer at Cyberkinetics Neurotechnology Systems, Inc., maker of the BrainGate system.

Dr. Scott pointed out in his commentary some of the many considerable problems that must be addressed before this technology can be put to regular clinical use.

• It is not clear for how long the very fine microelectrodes can be used to record neural activity. Patients with spinal-cord injuries are often young and would need to use these technologies for many decades.
• The prototype device involved passing a large bundle of wires directly through the skin to a connector attached to the skull, and all signal processing was done by an external computer. Wires passing through the skin promote infection, so for a clinical device, the power into and neural signals out of the implant must be transmitted using telemetry across the skin.
• Unlike the prototype device, a clinical implant would therefore have to minimize data transfer by carrying out a considerable amount of signal processing within the implanted component of the neuroprosthetic.

However, he also pointed out that these types of problems are not insurmountable; indeed, they were overcome during the development of cochlear implants -- the most successful neuroprosthetic so far.