Researchers Demonstrate Direct, Real-Time Brain Control Of Computer Cursor

The research was conducted using a primate model. Three Rhesus monkeys received brain implants similar to those used in treating certain human Parkinson's patients.

"We substituted thought control for hand control," said John Donoghue, chair of the Department of Neuroscience and the project's senior researcher. "A monkey's brain - not its hand - moved the cursor. Use of a reconstructed signal to allow the brain to accomplish immediate, complex goal-directed behavior has not been done before. We showed we could build a signal that works right away, in real time. And we can do it recording from as few as six neurons."

This work is a step toward enabling paralyzed humans to use thoughts to control a cursor that would allow them to read e-mail, surf the World Wide Web, or perform other functions through a computer interface.

Eventually, the technology may help individuals who have a spinal cord injury, Lou Gehrig's disease or muscular dystrophy, the researchers said. The researchers hope to apply the technology to restore some movement control in paralyzed patients. That step would entail seeking approval from the Food and Drug Administration. The FDA has not approved this "instant-control brain cursor" technique for human use.

The findings are described in the current issue of Nature. The lead author is Mijail D. Serruya, a graduate student enrolled in the M.D./Ph.D. program at Brown. Serruya performed the work as part of his Ph.D. research. As a medical student, he assists paralyzed patients. Serruya and Donoghue conducted the research with colleagues Nicholas Hatsopolous, a former Brown professor now at the University of Chicago; former Brown undergraduate Liam Paninski, now at New York University; and current Brown graduate student Matthew Fellows.

"This implant is potentially one that is very suitable for humans," Serruya said. "It shows enough promise that we think it could ultimately be hooked up via a computer to a paralyzed patient to restore that individual's interaction with the environment. Our goal is to make sense of how the brain plans to move a hand through space and to use that information as a control signal for someone who is paralyzed. We want to provide some freedom to this individual."

The device "would work for anything you can do or you can imagine doing by pointing and clicking," Donoghue said. 'This includes reading e-mail. Or imagine an onscreen keyboard that someone can use to type sentences or issue commands by pointing and clicking. We would be extraordinarily pleased if this system could allow a patient to become somewhat autonomous. It would restore some independence to paralyzed patients who are cognitively normal people unable to carry out their movement intentions."

The research involves use of thin electrodes to record the activity of a few neurons in the brain's motor cortex. This area contains the cells that fire when a hand moves. Activity of the neurons is first recorded while a cursor on a computer screen is moved to hit a target using a mouse-like handle.

The scientists built a series of mathematical formulas, called linear filters, to create a model that related the firing of the neurons to a cursor's target position. These linear filters then allowed the researchers to reconstruct hand trajectory from any new neural signals.

The electrode array was connected to a computer by thin cables. While the subject played a simple pinball-like video game, the researchers turned off the hand control and substituted the reconstructed signal. While the primate continued to move its hand as if playing the game, cursor motion actually was controlled solely by brain signals associated with moving the hand.

The subject then used its thoughts to move the cursor to different targets for periods averaging two minutes in length. While this instant-control brain cursor was active, the real-time signals allowed the animal to correct wayward cursor movements "on the fly" in order to strike the target, the researchers said. This entire processing took place nearly as fast as the hand responds to the brain's movement commands.

The research suggests that subjects can use visual and other feedback to compensate for inaccuracies in the mathematical model - in effect, to learn how to improve the brain's control of cursor movement, researchers said. "Our results demonstrate that a simple mathematical approach, coupled with a biological system, can provide effective decoding for brain-machine interfacing, which may eventually help restore function to neurologically impaired humans."

The work was funded in part by the National Institute of Neurological Diseases and Stroke, the Defense Advanced Research Projects Agency and the Burroughs Welcome Foundation.

Last year, Donoghue, Hatsopolous and others formed a company to transfer the technology to help patients who suffer from injuries and neurological disorders that result in paralysis. That firm is called Cyberkinetics.