Researchers from the University of Utah have found a way for severely paralyzed humans to "speak with their thoughts" through the use of microelectrode implants on top of the brain.  

Brain signals of paralyzed people can be translated into words through the use of two grids of 16 microelectrodes which are implanted above the brain and beneath the skull. But unlike many other types of electrodes, these microelectrode implants are much smaller than traditional electrodes developed half a century ago and they don't penetrate the brain at all -- they simply sit on top of it. These microelectrodes are called microECoGs, and their ability to work without penetrating the brain makes them safer for speech areas of the brain, which cannot be said for penetrating electrodes up to this point. 

The microelectrode implants work by placing the microECoGs, each spaced 1 millimeter apart, on two different speech areas of the brain: the facial motor cortex and the Wernicke's area. The facial motor complex controls movement of the lips, mouth, tongue and face while Wernicke's area is a "little understood" area in the brain that is linked to language understanding and comprehension. 

Bradley Greger, an assistant professor of bioengineering and lead author of this study, along with his team, tested the use of the microelectrode implants on a volunteer with severe epileptic seizures. They were able to place the implants between the brain and the skull due to the volunteer's previous craniotomy, which is the temporary partial removal of the skull in order to locate and surgically stop the seizures. 

Greger and his team then placed the microelectrodes on the man's brain and had him repeatedly read 10 words that would be useful to him while his brain signals were recorded. The words were repeated 31 to 96 times, and were “yes”, “no”, “goodbye”, “hello”, “more”, “less”, “hot”, “cold”, “hungry”, and “thirsty”. After recording the volunteer's brain signals, the team then went to figure out which brain signals represented which of the 10 words. 

When comparing brain signals for words like "yes" and "no" or "hot" and "cold," Greger and his team were able to recognize brain signals for each word 76 to 90 percent of the time. When all 10 brain signals were compared at once, it was a bit more difficult. The team could only correctly match a signal with a word 28 to 48 percent of time, but Greger argues that these numbers are still better than chance, which is 10 percent. 

"This is proof of concept," said Greger. "We've proven these signals can tell you what the person is saying well above above chance. But we need to be able to more with words with more accuracy before it is something a patient really might find useful."

Even though this new method requires a lot of improvement before it can be used in clinical trials, Greger and his team did find some interesting results, such as the fact that Wernicke's area is more involved in high-level understanding of language and that the team was able to distinguish signals for one word from those for another 85 percent of the time when signals were recorded from the facial motor complex. When recorded from Wernicke's area, they were only able to distinguish signals 76 percent of the time.

"The obvious next step - and this is what we are doing right now - is to do it with bigger microelectrode grids," said Greger. "We can make the grid bigger, have more electrodes and get a tremendous amount of data out of the brain, which probably means more words and better accuracy."

The study was published in the Journal of Neural Engineering this month.