Science Communications - Publicity for Technologya


monkey see. robot do.
How University of Pittsburgh Researcher Andrew Schwartz Is Discovering the Bases of Thought and Action by Implanting Electrodes in the Brains of Living Monkeys

Wouldn't it be nice to control a machine just by thinking about it?


In the laboratory of Dr. Andrew Schwartz at the University of Pittsburgh, a group of monkeys is doing just that. Each primate is equipped with an electrode array implanted in the motor cortex of its brain that senses the electrical "firings" of nerve cells as the animal thinks. With their arms restrained during experimentation, the monkeys learn how to actuate a robotic arm to grasp food morsels and feed themselves by thinking as though the arm were their own.


Since monkeys cannot understand verbal directions, experiments in Schwartz's lab begin with the monkeys doing simplified tasks under human guidance, then increasingly complex ones on their own, in response to rewards of food or drink.


During the experiments the electrodes detect the firing rates of neurons, which indicate the direction in which the animal intends to move. Each neuron has a preferred direction for which it fires fastest. For instance, a cell with a preferred direction pointed upward will fire fastest for movements in that direction. The neuron will fire at its lowest rate for downward movements. Although movements rarely occur in a strictly up or down way, this neuron will fire for all other movements, with a rate proportional to the angle away from its preferred direction. Since each neuron has a different preferred direction and all neurons fire in all directions, many neurons fire at a combination of rates for each movement. Schwartz decodes those changes in firing rates across a population of recorded neurons to calculate the direction of intended movement. He then sends the decoded signals to a brain-machine interface, which uses them to control the robotic arm in real time.


"We implant arrays of microelectrodes with tips that are actually smaller than individual neurons," Dr. Schwartz explained. "When a neuron communicates, it fires a pulse of electricity that travels from a cell body down a long neuronal cable called an axon. If enough of those signals arrive at a cell at the same time, it triggers an explosion of electricity that we can pick up. That explosion is called an action potential.


"The action potential is caused by sodium ions flowing down a concentration gradient. The sodium ions depolarize the cell membrane, allowing them to pass from one cellular gate to the next. When the next cell gets to a certain threshold, it opens up the next gate in the axon."


By correlating all the electrical activity of all the electrode tips in an array with the motions of the monkey's robotic arm and weighting for general tendencies of the neuron population, Schwartz and his colleagues can discern the monkey's intention to move as an instruction for the robotic arm.


To the casual observer the question arises, is the monkey teaching the robot or is the robot teaching the monkey? The answer is, both, Schwartz explained. "The monkey changes the way his neurons fire to make the device work better. At the same time we are altering our algorithm to make the device easier for the monkey to use. We call it co-adaptation. It's a two way process."


But as exciting as controlling a robotic arm by means of a monkey's brain activity may be, Schwartz has his eye on even more exciting things. "What we're doing is a lot bigger than just making prosthetic arms," he said. "The exciting thing is that we can see the cause and effect of how neurons fire." he said. "I'm interested in looking at the massive activity that goes on in the brain simultaneously and by trying to understand that in terms of behavior output."


This article first appeared in Tom Imerito’s TEQ column, Innovation Chronicles

© Copyright 2008, Thomas P. Imerito / dba Science Communications


Bookmark and Share

 
©2009 Science Communications
thomas@science-communications.com