'Brain-listening' Method May Lead to Better Machine Interfaces

by Elizabeth Norton Lasley

June 9, 2011

A technique used to trace the origins of seizures in the brain is proving a powerful new tool for exploring the brain’s everyday activity. Known as electrocorticography, the approach is also laying the groundwork for a “brain-computer interface” that could allow paralyzed people to control cursors on a screen, for example, using the brain’s power of intention.

Neurosurgeon Eric Leuthardt at Washington University, St. Louis, and colleagues were using electrocorticography to pinpoint seizures in patients with severe epilepsy. Because the sensors would need to remain in place for several weeks before the surgery, Leuthardt, who is also a bioengineer, saw a unique opportunity to learn more about how the brain controls specific actions.

Electrocorticography records changes in “high-gamma” frequencies, which indicate activity in the cortex, the brain’s outer layer. Measuring brain activity in milliseconds, the technique offers far greater specificity than other imaging methods, Leuthardt says. For example, functional magnetic resonance imaging only detects changes occurring one to two seconds apart. “The brain can do a lot in two seconds,” Leuthardt says.

As an added advantage, changes in high-gamma frequencies reveal activity in large groups of neurons that are not necessarily all firing at the same time. Thus, researchers can tweak apart individual components of a specific action.

Leuthardt’s team tested the possibility with a simple verbal task. Working with six epilepsy patients who’d had the grid of electrodes in place for a few weeks, the investigators positioned the electrodes over several brain areas involved in speech. The participants were asked to repeat words that they either saw or heard. For this apparently simple task, the subjects had to do six things:  read, prepare to say, and speak a written word, then go through the same three steps with a word they heard.

Each of the six steps had a different frequency, Leuthardt and colleagues reported in the Feb. 9 Journal of Neuroscience. “The results add a third pillar of understanding the brain’s activity, adding frequency to the time and location of cell firing,” says Leuthardt.

By separating volition from action, the finding also provides the basis for a brain-computer interface, harnessing the power of intent for those who are paralyzed or even left without speech.

In the April 7 Journal of Neural Engineering, Leuthardt and several members of the same team coupled their electrocorticography approach to a software interface. Study participants (four of the epilepsy patients from the previous research) were asked to either think or say four different vowel sounds. When the scientists programmed the distinct frequencies to the interface, participants could move a cursor on a computer screen by saying, or even just thinking, the sound.

“These findings show that electrocorticography is a powerful and practical way to collect signals from the brain,” says neurosurgeon Nader Pouratian at the University of California, Los Angeles. “Using the signals to help patients has been a dream in many movies. This technology is getting us closer to making the dream a reality.” 

He adds that although positioning electrodes on the surface of the brain may sound drastic, the procedure is comparatively noninvasive. Though surgery is required, nothing is inserted into the brain itself--unlike deep-brain stimulation, for example, which is widely used to treat Parkinson’s disease and other conditions. “Patients with epilepsy already have these grids implanted for days to weeks, just to record the signals,” he notes. “The next step to a practical brain-computer interface is simply to leave the grids in place for life.”

Electrocorticography is a young field, says Pouratian, and its advantage is also a potential limitation. “By listening to millions of neurons at once, can we get the specificity we need to control complex behaviors with a brain-computer interface? Decades ago, scientists would have said absolutely not. It will take a lot of analysis, a lot of patience, but we are learning how to harness the information from these signals. There’s huge potential to help a lot of people.”

More information:

Journal of Neuroscience study: DOI:10.1523/JNEUROSCI.4722-10.2011

Journal of Neural Engineering Study: http://iopscience.iop.org/1741-2552/8/3/036004