cells control 3D cursor
Technology Research News
Researchers at Arizona State University
have developed a feedback system that lets monkeys use brain signals to
move a virtual ball within a computer-generated box, an advance that increases
the chances that scientists will be able to give disabled people neural
control of prosthetic limbs.
The research also suggests that surgeons will eventually be able to rewire
bodies to give people control over paralyzed body parts.
The feat is the latest in a string of advances that allow brains to directly
control electronics. Scientists have been planting electrodes in the brains
of animals to record electrical activity for decades, but they have only
recently been able to use these neural signals to control robots and computers.
In 2000 researchers from Duke University, MIT and the State University
of New York Health Science Center tapped a monkey's brain signals to make
a remote robotic arm mimic the movements of the monkey's own arm. Earlier
this year, researchers at Brown University showed that the method could
allow monkeys to consciously control a computer cursor, and found that
one monkey learned to move the cursor without physically moving its arm.
The Arizona State experiment goes beyond two-dimensional cursor control
to give a pair of rhesus macaque monkeys direct cognitive control of a
virtual ball in a three-dimensional space. The Arizona monkeys also showed
greater control over the ball than the Brown monkeys had over their cursors,
said Andrew Schwartz, a research professor of bioengineering at Arizona
State University. The ball control resembles "real biological movement,"
The key to the researchers' success is a feedback system between the monkeys'
neurons and the software algorithm used to translate brain signals into
Each of the many billion neurons in a primate brain is connected to as
many as 10,000 other neurons. Learning occurs when the brain adapts to
different conditions by changing the patterns of signals neurons transmit
The researchers added their software to the learning loop, allowing it
to adapt along with the changing neural signals within a small portion
of a monkey's brain. "We are using a more sophisticated approach that
allows two-way learning to take place," said Schwartz. "The animal learns
to move the cursor using biofeedback to change the discharge patterns
of its neurons. Our decoding algorithm tracks these changes as they occur"
in order to make better predictions about the new neural patterns, he
The monkeys were able to move the ball using brain signals alone almost
as well as they were able to control it with arm movements, said Schwartz.
To teach the monkeys this cognitive control, the researchers implanted
electrodes in the motor cortex region of their brains. Motor neurons coordinate
muscle activity. The researchers rewarded the monkeys for using arm movements
to move the ball to a particular spot, and recorded the neural activity.
They used this recording to calibrate the software that translates the
neural activity into the control signal for the computer.
The researchers then restrained the monkeys' arms so that the monkeys
could not physically move their arms as they attempted to move the virtual
ball. At first the monkeys pushed against the restraints in the direction
they wanted the ball to move, but stopped straining as their performance
improved. Measurements of the monkeys' brain activity showed that eventually
the monkeys could control the ball without using the normal brain signal
patterns associated with muscle movements, indicating that the neurons
had adapted to a new circumstance, according to Schwartz.
Both the Brown and Arizona State research teams were able to use a surprisingly
small number of neurons to generate a control signal. Monkeys have millions
of neurons in the motor cortex, but the Arizona State researchers used
the signals from less than two dozen motor neurons to generate the ball
control signal, according to Schwartz.
Using so few neurons would not be practical for controlling a prosthetic
device, said Miguel Nicolelis, a professor of neurobiology and biomedical
engineering at Duke University, and the lead researcher on the robotic
arm project. "If you [were to] lose a couple of neurons your entire implant
[would] become useless," he said. "Much larger neuronal samples are needed"
to make an implant practical over time.
Controlling a neuroprosthetic arm is also considerably more complicated
than moving a cursor in a three-dimensional space, said Nicolelis. "To
reproduce complex 3D hand and arm trajectories and to mimic the force
required to move objects with a prosthetic arm, hundreds of neurons would
be needed," he said.
The Arizona State researchers' next step is to replace the virtual cursor
with a robot arm that will be used by a monkey to retrieve food while
its arms are restrained, said Schwartz.
And over the next two or three years, "we would like to try these implants...
in human patients," said Schwartz. Such a neural bypass could be used
to give disabled people control over computer-driven prosthetic limbs.
The technique could also be combined with electrical signals that stimulate
muscle movement in order to let paralyzed people regain control of their
own limbs, said Schwartz.
Schwartz's research colleagues were Dawn Taylor and Stephen Helms Tillery
of Arizona State University. They published the research in the June 7,
2002 issue of the journal Science. The research was funded by the National
Institutes of Health, the Whitaker Foundation, the Philanthropic Education
Organization and the U.S. Public Health Service.
Timeline: 2-3 years
Funding: Government, Private
TRN Categories: Biotechnology; Human-Computer Interaction
Story Type: News
Related Elements: Technical paper, "Direct Cortical Control
of 3-D Neuroprosthetic Devices," Science, June 7, 2002
Software guides museum-goers
thread from nanotubes
Brain cells control
heightens Web class divide
One-way heat valve possible
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link