 |
| In a first, a brain implant converts a paralyzed man's thoughts into "speech." |
While Elon Musk's brain-chip company messes around with gaming monkeys, another group of researchers has achieved a major milestone in neuroprosthetics: allowing a man who can't talk to form sentences with his mind.
The accomplishment was accomplished by researchers at the University of California, San Francisco (UCSF), who used a brain implant that decoded complete phrases from brain activity.
They put the technology to the test on a man in his late 30s who had a brainstem stroke in his teens that rendered him speechless. He now communicates by poking letters on a screen with a pointer affixed to his baseball cap.
The individual, who requested to be called BRAVO1, was the first person to take part in a clinical trial of the new technology.
BRAVO1's speech motor cortex was initially surgically implanted with a high-density electrode array. A connector in his brain was used to link the implant to a computer.
The scientists monitored BRAVO1's brain activity for 22 hours over several months while he attempted to utter a list of 50 basic phrases including "water," "family," and "good." This vocabulary may be used to create almost 1,000 sentences.
Custom neural network models identified which phrases he was trying to speak when he spoke by distinguishing between the neurological signals. They could generate over 1,000 sentences if they worked together.
After that, BRAVO1 was prompted to recite a series of brief sentences made up of 50 words. On a screen, the words were deciphered from his brain activity.
The team then attempted a question-and-answer exercise. When they questioned him, "How are you today?" his response was displayed on the screen: "I am really good."
This is the first time the brain activity of a disabled person who cannot speak has been decoded into entire words, according to the researchers.
UCSF neurosurgeon Edward Chang remarked, "It shows strong potential to restore communication by tapping into the brain's normal speech system."
Many prior studies in this subject relied on spelling-based methods for typing out letters one by one. Chang claims that his team's method uses a more natural feature of speech:
We often exchange information at a high velocity using voice, up to 150 or 200 words per minute. Going straight to words, like we are here, provides a lot of advantages because it is more like how we generally speak.
However, the system is still prone to errors. At 15 words per minute, words were deciphered with a median accuracy of 74 percent, with a peak performance of 93 percent at 18 words per minute.
The team is now working to improve the vocabulary and speaking pace of their system.
They anticipate that with time, their approach will be able to restore complete communication to the thousands of people who lose their ability to speak each year.