Brain (minus machine) interface

Learning to use an implanted brain electrode to control a prosthetic or robotic arm might be easier than researchers thought, suggests a linkurl:study;http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.1000153 published online yesterday (21 July) in PLoS Biology. Ideally, the goal of a brain-machine interface is "to control the prosthetic naturally," said lead author linkurl:Jose Carmena;http://www.eecs.berkeley.edu/~carmena/ from the University of California, Berkeley. To date,

Monkeys asked to move a cursor
between the center dot and
radial dots were proficient by day 13
Image: Jose Carmena, Karunesh Ganguly
and PLoS Biology

Brain-implanted neuroprosthetics use a computer algorithm to interpret the brain's impulses. But the standard practice with these implants is to reset the algorithm every day -- sometimes twice a day -- to adjust for neurons that move away from the electrodes. "It's one feature of the neuron that makes them tricky," said Chet Moritz from the University of Washington, who was not involved in the research. "Sometime they come and go; sometimes they disappear for good." Carmena and coauthor Karunesh Ganguly, also at UC Berkeley, wanted to find a more consistent approach. They started with the protocol researchers generally use after implanting a neuroprosthetic -- "training" the computer algorithm to take note of the neurons that fired when they asked the monkeys to move their arm. Normally, researchers repeat this process each day to find a suite of neurons that demonstrate the ability to control movement. Instead, Ganguly and Carmena searched for a suite of stable neurons that they could record from consistently every day. They then let the animal play a computer game in which it must control those neurons to move the cursor left, right, up or down across a computer screen. Daily computer training translates into variable performance by the monkey (or patient) from day to day. But when Ganguly and Carmena instead trained the computer just once with the same neurons, and then used the same algorithm on subsequent days, not only were the monkeys able to reach 100% proficiency quickly, they were able maintain that level each day after. The results suggest that "the brain may be able to learn to control these devices better when the computer is not changing the algorithms along the way," said Moritz. Reducing the amount of computer intervention may also help monkeys remember the task -- a feat that has proved difficult in previous studies. After the monkeys learned the task, the researchers changed the rules of the game by switching how the cursor on the screen was controlled. The tweak was comparable to having the buttons on your mouse reversed so that when you move the mouse left, the cursor goes right, and visa versa, explained Moritz. Once the monkeys gained proficiency with the new controls, the researchers restored the old controls; the monkeys still performed at 100% accuracy. "The brain can consolidate the motor skill of the prosthetic device," said Carmena. That means it might be possible to add more complicated controls one by one, he said, consolidating each to memory before adding another -- an approach that could lead the way to more seamless and natural neuroprosthetics control.
**__Related stories:__***linkurl:Of Cells and Wires;http://www.the-scientist.com/2009/01/1/32/1/
[January 2009]*linkurl:Single neuron power;http://www.the-scientist.com/blog/display/55091/
[15th October 2008]*linkurl: Neuroprosthetics Today;http://www.the-scientist.com/article/display/55286/
[January 2009]