In a groundbreaking development, neuroscientists have demonstrated a brain-computer interface that can translate brain signals detected from outside the skull into both horizontal and vertical movement of a computer cursor. The method may be a safer alternative to interfaces that rely on surgically implanted electrodes and could lead to noninvasive prosthetics for paralyzed people.
Monkeys with implanted electrodes have used brain signals to move cursors or robotic arms in two dimensions (Science, 24 January 2003, p. 496). And this fall, surgeons implanted 100 electrodes into the brain of a 25-year-old quadriplegic man and connected them to a computer that enables him to check his e-mail and choose a television channel with his thoughts alone. But neurosurgery carries risks, including infection and brain damage.
A non-invasive interface would get around these risks, but most researchers have thought the usefulness of such interfaces would be limited. One early version, reported in 1991 by neuroscientists Jonathan Wolpaw and Dennis McFarland (now at the Wadsworth Center, part of the New York State Department of Health in Albany) and colleagues measured brain-wave changes using an electroencephalogram (EEG) and enabled a person to move a cursor up or down on a screen. Wolpaw and McFarland's new interface adds a second dimension of movement and includes a learning algorithm: The software program translating brain signals into cursor movement optimizes a user's performance based on the trials a user has completed so far.
Putting the new program to the test, four volunteers--two of them with spinal cord injuries--donned caps speckled with 64 recording electrodes and used mental imagery to push a cursor from the center of a computer screen to a target in any of eight possible locations. As the volunteers did the task, a computer translated their EEG rhythms into horizontal and vertical cursor movements. After dozens of short practice sessions spread out over weeks, the two volunteers with spinal cord injuries could hit the target about 90% of the time within the 10-second time limit, the authors report in a paper published online this week in the Proceedings of the National Academy of Sciences. (The others did so 70% to 80% of the time, perhaps because they were less motivated.)
Two-dimensional cursor control could be used to operate a wheelchair, a chess-playing robot, or a computer mouse, among other things, says computer scientist Melody Moore at Georgia State University in Atlanta. "It's earthshattering that we may be able to reconnect the brain to a paralyzed limb or a robotic arm without surgery," she says.