Machine Reads Abstract Thought

Signals from brain cells involved in abstract thinking have been decoded, an important step towards creating better prosthetic devices for paralyzed people.

Working with monkeys, researcher Richard Andersen and colleagues at the California Institute of Technology in Pasadena recorded signals from neurons related to reaching movements and then decoded the signals using a computer.

The decoded signals enabled the researchers to predict the monkeys' arm movements in tasks in which they thought about reaching for an item without actually doing so.

"These signals could also be rapidly manipulated to expedite the learning that patients must undergo to use an external device," say the researchers. "Moreover, this research suggests that all kinds of cognitive signals can be decoded from patients."

Signal detection

Past studies on monkeys have shown that information from neurons coding movement instructions can be used to control prosthetic devices .

Last year, for example, it was reported that Rhesus monkeys were taught to control and assimilate a robot arm using signals from their brain.

To achieve this, researchers implanted an array of microelectrodes into the frontal and parietal lobes—areas of the brain involved in producing multiple output commands to control complex muscle movements.

The faint signals from the electrodes can be detected and analyzed by a computer system to recognize patterns of signals that represent particular movements by an animal's arm. These can be translated into similar movements of a robotic arm.

Until now, however, nobody has succeeded in tapping the messages of higher-order neurons involved in planning and motivation.

Cracking the code

For their study, Andersen and colleagues implanted in monkeys arrays of electrodes into areas of the brain that encode the goals of reaching movements rather than controlling movement itself.

While the monkeys waited for a cue that told them to reach for an icon flashing on a screen, a computer program interpreted the brain signals recorded by the electrodes.

Once the "neuronal code" was cracked, the researchers used the program to decipher the direction that the monkeys were planning to reach for during trials in which they thought about reaching but didn't actually do so.

When monkeys remained still while having thoughts that were consistent with requested movements, they received a reward.

Practice makes perfect

At first, the program had trouble matching the monkeys' intentions to the icon's position much more often than chance.

As the monkeys practiced thinking about reaching, however, their neural signals became stronger, enabling the program to decode the correct direction more frequently.

Eventually, the program could predict the intended direction of the monkeys' reach as much as 67% of the time.

When the monkeys knew that accurately thinking about the requested movement would yield a preferred reward, the computer's ability to predict direction improved by as much as 21%.

Paralysis aid

Similar goal and preference information from people with paralysis could be useful, says Andersen.

Future technologies based on this research could allow a paralyzed person's goal-directed thoughts to be interpreted with smart machines.

The research also suggests that other kinds of cognitive signals can be decoded.

"For instance, recording thoughts from speech areas could alleviate the use of more cumbersome letter boards and time-consuming spelling programs, or recordings from emotion centers could provide an online indication of a patient's emotional state," say the researchers.

<< HOME