Brain–Computer Interfaces - Index of

210 D.M. Taylor and M.E. Stetner
designed analog circuits that transformed the voltage spikes detected by the electrodes
into the movement of a needle on a voltmeter. The circuits were designed so
that the needle would move when the monkey increased the firing rate of individually
recorded neurons sometimes while decreasing activity in associated muscles.
Most importantly, the animal was allowed to watch the needle’s movements and
was rewarded when he moved the needle past a certain mark. The animal quickly
learned to control the firing activity of the neurons as assigned in order to move the
voltmeter needle and get a reward. These early studies show the power of visual
feedback, and they demonstrated that monkeys (and presumably humans) have the
ability to quickly learn to willfully modulate the firing patterns of individual neurons
for the purpose of controlling external devices.
Along side this initial BCI work, other researchers were implanting intracortical
microelectrodes in brains of monkeys simply to try to understand how our
nervous system controls movement. These early studies have laid the foundation for
many BCI systems that decode one’s intended arm and hand movement in real time
and use that desired movement command to control the movements of a computer
cursor, a robotic arm, or even one’s own paralyzed arm activated via implanted stimulators.
One seminal study was done in the 1980s by Dr. Apostolos Georgopoulos
and his colleagues who showed that the different neurons in the arm areas of the
brain are “directionally tuned” [3]. Each neuron has what’s called a “preferred direction”.
Each neuron will fire at its maximum rate as you are about to move your arm
in that neuron’s preferred direction. The neuron’s firing rate will decrease as the arm
movements start deviating from that cell’s preferred direction. Finally, the neuron
will fire at its minimum rate as you move your arm opposite the neuron’s preferred
direction. Different neurons have different preferred directions. By looking at the
firing rates of many different directionally-tuned neurons, we can deduce how the
arm is moving fairly reliably [4].
Unfortunately, each neuron’s firing patterns are not perfectly linked to one’s
movement intent. Neurons have a substantial amount of variability in their firing patterns
and each one only provides a noisy estimate of intended movement. In 1988,
Georgopoulos and Massey conducted a study to determine how accurately movement
direction could be predicted using different numbers of motor cortex neurons
[5]. They recorded individual neurons in monkeys from the area of the brain that
controls arm movements. They recorded these neurons while the monkeys made a
sequence of “center-out” movements with a manipulandum, which is a device that
tracks hand position as one moves a handle along a tabletop. In these experiments,
the table top had a ring of equally-spaced target lights and one light in the center
of this ring that indicated the starting hand position. The monkey would start by
holding the manipulandum over the center light. When one of the radial target lights
would come on, the animal would move the manipulandum outward as fast as it
could to the radial target to get a reward 1 . The researchers recorded many different
1 Because the researchers imposed time limits on hitting the targets, the subjects tried to move
quickly to the targets and did not always hit the target accurately.