While armies of robots may still be well ahead into the future, immense progress has been made in research that could have a tremendous impact on the lives of people who suffer from paralysis. Often caused by diseases such as spinal cord injury after an accident, or by stroke, paralysis can be, at its most extreme, virtually complete, leading to locked-in syndrome. One of the most famous cases of paralysis after spinal cord injury is perhaps the late actor Christopher Reeve, who became paralysed from the neck down after a fall from a horse. Over the past couple of decades, research into paralysis has focused on the development of a brain-machine interface that allows people to control muscles and even prosthetic limbs using thought alone.
|Penfield's motor homunculus|
Further exploration into motor functioning led Eberhard Fetz and colleagues to discover that monkeys were able to voluntarily control neural activity through operant training, as long as they were given biofeedback. The monkeys, who had electrodes implanted in their brain, were looking at a meter that showed their own neuronal activity. Whenever the needle of the meter was pointing to the right side, the monkey received a reward. As soon as monkeys had learned that there was a link between the meter and the reward, they were able to adjust their neural firing patterns for more rewards (Fetz, 1969; Fetz and Baker, 1973). In the beginning the neuronal activity still resulted in active movement of the monkeys’ limbs, but over time, while the monkey still voluntarily controlled the neuronal activity, the overt movements extinguished (Fetz and Baker, 1973).
In the early 1980s, researchers began to slowly adopt the idea that ensembles of neurons spread across the brain, and not just single neurons, create physiological activity in the central nervous system of mammals (Gerstein and Aertsen, 1985). Donald Hebb had already suggested this in 1949, but research in this area was delayed by the persistent idea that single unit recordings would provide us with the answer to the neural code, because it was believed that only a few specific neurons controlled a particular movement. Since there are over a 100 billion neurons in our brain, finding those specific neurons would be like looking for a needle in a haystack. Thus, the discovery that ensembles of neurons were responsible for movements made looking for the target areas a lot easier.
By using different imaging techniques to look at this widespread brain activity, Jeannerod and colleagues made an interesting discovery. They compared activity in the brains of participants as they were imagining making a specific movement and compared these activations with the activity observed when the participants were actually making the movement. They found a striking resemblance between the observed activations, leading to the conclusion that imagining a movement may not be so neurally different from actually performing the movement. This result provided a significant step in brain-machine interface research (Jeannerod, 1995; Jeannerod and Frak, 1999).
In parallel, Nicholelis and Chapin managed to train rats to receive rewards for activating lever press related activity in their brain. First, they trained thirsty rats to press a lever to get a drink of water. As they were pressing the lever, the researchers recorded the patterns of activity from 46 neurons in the rats’ brains, which they used in the next stage of the experiment. The lever was disconnected from the water supply and the rat no longer received water from a lever press. However, the rat went on pressing and the scientists gave the rat water whenever its brain produced the command for pressing the lever. After a while, the rat stopped pressing the lever altogether, but kept producing the press lever command in its brain. The external machine that delivers the reward is now directly operated through the command ‘press lever’ in the rat’s brain (Nicholelis et al., 1998).
The next step was to attempt something similar in primates and make the movements more intricate. Using a similar method as in the rats, Nicolelis and his colleagues managed to let a monkey swing an artificial arm from left to right, just by using thought. Then it was time to allow the monkey to make more complex movements. They first learned to use a joystick to drag a cursor onto a target on a computer screen while the researchers recorded the patterns of activity like before. Soon, the monkeys learned that the command for ‘drag cursor’ resulted in reward, and they stopped actively dragging the cursor onto the target. This was swiftly followed by an ability to reach and grab, trained in similar ways (Nicholelis et al., 2003).
Thus, scientists are now able to create communication pathways between the brain and an external machine (e.g. a computer cursor or an artificial limb). The first human trials have already successfully shown that paralysed patients were able to directly control computer cursors (Hochberg et al., 2006). However, they have also raised issues about safety and reliability of the intracranial electrodes that need to be addressed before clinical trials can extent to larger patient populations.
|Sensory feedback is essential|
Research is also currently focusing on the striking finding that the brains of the monkeys that participate in these studies become structurally adapted to the external devices. Thus, different areas within the motor cortex of these monkeys now seem to be representing the robot, as if the robot was a part of their own body. If further research proves this to be right, the implications can be unprecedented. It would allow the patient to perceive the prosthetic device as an actual part of their body (Nicholelis and Lebedev, 2009).
A global team of neurophysiologists, computer scientists, engineers, roboticists, neurologists and neurosurgeons are now working together in the “Walk Again Project” to do no less than develop a generation of neuroprosthetic devices that can restore full-body mobility in patients with severe paralysis. Thus, while DARPA is dreaming of creating armies of robots, their money has been well spent on research that could mobilise the immobile in the not too distant future.
Fetz EE. (1969). Science 163, 955-958.
Fetz EE, Baker MA. (1973). Journal of Neurophysiology 36, 179-204.
Gerstein GL, Aertsen AM. (1985). Journal of Neurophysiology 54, 1513-1528.
Jeannerod M. (1995). Neuopsychologia 33, 1419-1432.
Jeannerod M, Frak V. (1999). Current Opinion in Neurobiology 9, 735-739.
Nicolelis MAL et al. (1998). Nature Neuroscience 1, 621–630.
Nicolelis MAL et al. (2003). Proceedings of the National Academy of Sciences 100, 11041-11046.
Hochberg LR et al. (2006). Nature 442, 164-171.
Suminski AJ et al. (2010). Journal of Neuroscience 30, 16777-16787.
Nicolelis MAL, Lebedev, MA. (2009). Nature Reviews Neuroscience 10, 530-540.
This article appeared in the Trinity Term issue of 'Phenotype' in a slightly modified form. Phenotyope is the termly science magazine published by the Oxford University Biochemical Society. Here you can read this issue of the magazine in its entirety.