Brain–Machine Interface Holds Promise for Prosthetics

Jose Carmena of UC Berkeley’s department of electrical engineering and computer sciences. PEG SKORPINSKI PHOTOJose Carmena of UC Berkeley’s department of electrical engineering and computer sciences. (Photo by Peg Skorpinski.)
“Practice makes perfect” is the maxim drummed into anyone struggling to learn a new motor skill, be it riding a bike or developing a killer backhand in tennis. New research by UC Berkeley assistant professor of electrical engineering and computer sciences Jose Carmena and colleagues now reveals that the brain can also achieve this motor memory with a disembodied device.

Published July 21 in the open-access journal PLoS Biology, the study addresses a fundamental question: whether the brain can establish a stable, neural map of a motor task, making control of an artificial limb more intuitive. It provides hope that physically disabled people could one day master control of artificial limbs with greater ease, the researchers say.

In the study, macaque monkeys using brain signals learned how to move a computer cursor to various targets. The researchers found that the brain could develop a mental map of a solution to achieve the task with high proficiency and that it adhered to that neural pattern without deviation, much like a driver sticks to a given route commuting to work.

 “When your own body performs motor tasks repeatedly, the movements become almost automatic,” Carmena says. “The profound part of our study is that this is all happening with something that is not part of one’s own body. We have demonstrated that the brain is able to form a motor memory to control a disembodied device in a way that mirrors how it controls its own body. That has never been shown before.” Carmena, principal investigator on the study, also has UC Berkeley appointments in cognitive science and at the Helen Wills Neuroscience Institute.

In experiments involving moving a cursor from a starting point to a nearby target, Carmena and colleagues found that the brain can create a stable mental representation of a disembodied device. (© John Blanchard illustration)In experiments involving moving a cursor from a starting point to a nearby target, Carmena and colleagues found that the brain can create a stable mental representation of a disembodied device. (illustration by © John Blanchard.)Researchers in the field of brain–machine interfaces have made significant strides in recent years in the effort to support people with physical disabilities. An April 2009 survey by the Christopher and Dana Reeve Foundation found that, in the United States alone, nearly 1.3 million people suffer from some form of paralysis caused by spinal cord injury. When stroke, multiple sclerosis, cerebral palsy and other paralyzing diseases are included, the number jumps to 5.6 million. Another 1.7 million are living with limb loss, according to the Amputee Coalition of America.

Already, researchers have demonstrated that rodents, nonhuman primates and humans are able to control robotic devices or computer cursors in real time using only brain signals. But what had not been clear before was whether such a skill had been consolidated as a motor memory. The new study suggests that the brain is capable of creating a stable, mental representation of a disembodied device so that it can be controlled with little effort.

To demonstrate this, Carmena and Karunesh Ganguly, a post-doctoral fellow in his lab, used a mathematical model or “decoder” that remained static during the length of the study, and they paired it with a stable group of neurons in the brain. The decoder, analogous to a simplified spinal cord, translated the signals from the brain’s motor cortex into movement of the cursor.

It took about four to five days of practice for the monkeys to master precise control of the cursor. Once they did, they completed the task easily and quickly for the next two weeks.

As the tasks were being completed, the researchers were able to monitor the changes in activity of individual neurons involved in controlling the cursor. They could tell which cells were firing when the cursor moved in specific directions. The researchers noticed that, when the animals became proficient at the task, the neural patterns involved in the “solution” stabilized.

“The solution adopted is what the brain returned to repeatedly,” Carmena said.

That stability is one of three major features scientists associate with motor memory, and it is all too familiar to music teachers and athletic coaches who try to help their students “unlearn” improper form or technique. Once a motor memory has been consolidated it can be difficult to change.

Other characteristics of motor memory include rapid recall upon demand and resistance to interference when new skills are learned. All three elements were demonstrated in the UC Berkeley study.

“This is a study that says that maybe one day, we can really think of the ultimate neuroprosthetic device that humans can use to perform many different tasks in a more natural way,” Carmena said. The researchers acknowledged, however, that prosthetic devices could never match what millions of years of evolution have accomplished to enable animal brains to control body movement.

The study was supported by the Christopher and Dana Reeve Foundation, the American Heart Association and the American Stroke Association.