For the first time ever a team of researchers says it has deployed a brain-computer interface (or BCI) to decode the neural signals that represent the writing of letters and turn them into text. A 65-year-old man paralyzed from the neck down showcased the BCI, which allowed him to write 90 characters per minute. Or approximately double the previous record for “writing” with a brain-computer interface.
Science Alert reported on the brain-computer interface breakthrough, which the researchers outlined in a paper published in the journal Nature. The researchers, working at BrainGate—a nonprofit consisting of neuroscientists, engineers, and the like—have the ultimate aim of helping paralyzed people. Specifically by restoring their ability to communicate. As well as by enhancing their independence in the absence of physical movement.
Earlier BrainGate research related to turning thoughts of writing into text focused on using neural patterns related to arm movements. Meaning that, previously, machine learning algorithms would decode neural signals of arm movements—gleaned from implanted sensors in the brain—in order to turn them into the movements of a mouse cursor. In turn allowing a BCI user to click on letters on a computer screen in order to write out words. This new system, however, focuses on the actual writing of individual letters.
“We want to find new ways of letting people communicate faster,” Frank Willett, a research scientist at Stanford University and the Howard Hughes Medical Institute (HHMI), and lead author of the study, said in a press release. “This new system uses both the rich neural activity recorded by intracortical electrodes and the power of language models that, when applied to the neurally decoded letters, can create rapid and accurate text,” Willett added.
In essence, the BCI works by having a user think about writing letters. Because each letter of the alphabet has its own distinct pattern, thinking about any one results in distinct neurological activity. The BrianGate researchers then use machine learning algorithms to decode which letter a BCI user is thinking about writing. (Machine learning algorithms are algorithms trained on “big data” that can subsequently identify patterns.)
The result of the new “brain-to-text” BCI is a tool that allows paralyzed users to “write” 18 words per minute. And with 99% accuracy when researchers enable autocorrect. In the video above we see a 65-year-old paralyzed BCI user write out his responses to questions. And while the process seems slow, the user’s able to write as quickly as somebody his age texting with a smartphone.
BrainGate researchers say new iterations of their BCI will incorporate other ways of communicating as well. New versions will combine brain-to-text with point-and-click navigation, similar to what’s available on smartphones. The researchers even want to use their BCI to take the sounds of words in people’s heads and turn them into text.
Feature image: HHMI