Contact: Claire Bowles
A device that reads brain waves through the skull has enabled paralysed people to write sentences on a computer screen.
Several teams worldwide are trying to develop systems to allow "locked in" patients to communicate. Last year, for instance, two patients managed to write messages on a computer screen via electrodes implanted in their brains by researchers at Emory University in Atlanta, Georgia (This Week, 17 October 1998, p 5).
Because brain surgery involves serious risk of infection or haemorrhage, other groups are trying to develop systems that do not require implants. An Australian team has already used signals from an electroencephalograph (EEG) to control a simple switch (This Week, 4 May 1996, p 6). Now researchers in Germany and the US claim to have developed a much more sophisticated system. "We've got patients writing messages who couldn't communicate at all," says Edward Taub of the University of Alabama at Birmingham, a member of the team.
The researchers, led by Niels Birbaumer of the University of Tybingen, placed two contact-lens-sized electrodes near the top of their patients' heads, over the motor cortex, to record their slow cortical potentials. These signals can be recorded more reliably than other components, such as alpha waves, which don't always show up in an EEG readout.
The researchers worked with three patients with amyotrophic lateral sclerosis, a neurodegenerative disease that often leads to total paralysis. The patients learnt to make their cortical potentials more negative or positive, which moved a cursor up or down on a computer screen towards a goal box. Each session, the researchers increased the slow cortical potential their patients had to generate to get the cursor to reach the box-the cerebral equivalent of adding weights to a powerlifter's bar.
Once the patients could control the cursor reliably, they began to write messages. First, they moved the cursor either up or down to choose one half of the alphabet. That half was then split again on the screen, and the patient chose the half with the target letter, and so on. Each choice took four seconds, the first two to obtain a "baseline" cortical potential, followed by two seconds to move the cursor. The start and end of this "active" phase were marked by low and high-pitched tones.
To select a desired character from a panel of 32 letters and punctuation
marks, the patient had to choose up or down five times. Occasional failures
to move the cursor caused delays, but in the latest issue of Experimental
Brain Research (vol 124, p 223) the researchers report that it took an
average of 80 seconds to choose each character. The patients could write
sentence in about half an hour.
Researchers working with implants welcome Birbaumer's results. "If you
can do it without surgery, that's great," says Phil Kennedy, a member of
the Emory team. But he says that a system based on simple binary choices
has limited scope. His team hopes one day to use brain implants to exert
fine control over devices that
simulate muscles, allowing paralysed patients to regain control over their bodies.
However, Taub believes it will be possible to train paralysed patients to make choices between more than two options by setting their cortical potentials to several positive or negative levels. "In concept, there is no advantage to either approach," he claims.
Birbaumer's team is now working to speed up the "thought translation
device." One approach is a computer program that guesses the intended word
after two or three letters have been typed, on the basis of context and
the frequency of words in the language.
US CONTACT -- Barbara Thurlow, New Scientist Washington Office: Tel:
202-452-1178 or email firstname.lastname@example.org