Ever since he was paralyzed by a severe stroke in 2003, a man known as Pancho has been unable to speak. Now, researchers at the University of California-San Francisco (UCSF) have developed a brain implant that translates his brain signals into words on a computer screen.

Resource library: Neuroscience clinical technology compendium

How Pancho was paralyzed

In 2003, Pancho—whom the New York Times identified only by his nickname to protect his privacy—was in a serious car crash. Although he suffered severe stomach injuries, he underwent a successful surgery and was discharged from the hospital able to walk and talk.

One day later, his condition turned sharply worse.

Pancho suffered a brainstem stroke apparently caused by a post-surgery blood clot. He fell into a coma, and when he awoke a week later, he could no longer speak.

"I wished I didn't ever come back from the coma I was in," he told the Times.

Webinar recording: 2021 neurosciences market trends

For many years, his only way of communicating was by using a pointer attached to a baseball cap to tap out individual letters on a screen. He had to bend his head down to poke letters one-by-one, a slow and error-prone process.

That all changed when Pancho agreed to work with UCSF neuroscience researchers on a first-of-its-kind experiment known as "BRAVO": the Brain-Computer Interface Restoration of Arm and Voice study.

His first recognizable sentence: 'My family is outside'

Researchers for the study, which was published in the New England Journal of Medicine, first worked with Pancho to develop a bare-bones, 50-word vocabulary. Edward Chang, chairman of neurological surgery at UCSF, then implanted a sheet of 128 electrodes to detect signals from Pancho's speech motor cortex.

Next, Pancho participated in dozens of sessions in which he repeatedly attempted to say the 50 words while researchers recorded signals from his speech cortex. The research team used that data to train artificial intelligence to distinguish patterns in Pancho's brain activity and identify his words.

The results: In 47% of the initial 9,000 times Pancho tried to say single words, the trained algorithm accurately understood them. The accuracy rose even higher when the researchers added a word-prediction algorithm to interpret entire sentences.

According to Chang, after the algorithm accurately identified Pancho's words, "you could see him visibly shaking, and it looked like he was kind of giggling."

Pancho's first recognizable sentence was, "My family is outside," the Times reports.

The first technology of its kind, but likely not the last

The results appear to be unprecedented. "To our knowledge, this is the first successful demonstration of direct decoding of full words from the brain activity of someone who is paralyzed and cannot speak," Chang said.

According to the Wall Street Journal, tech companies hope to commercialize the technology. Facebook sponsored the BRAVO study and has said it is interested in noninvasive, wearable devices that could allow people to type by thinking.

Researchers emphasized that the technology does not read someone's private thoughts but rather detects brain signals related to words they are actively attempting to say. "[I]In the future, we might be able to do what people are thinking," which could raise ethical questions—but according to Chang, the current study "is really just about restoring the individual's voice."

Some outside experts questioned whether the technology was yet practical. Amy Orsborn, an assistant professor of bioengineering at the University of Washington, who was not involved in the research, told the Journal that the system's high initial error rate, limited vocabulary, and significant training requirements mean there is still substantial work to be done.

But Leigh Hochberg, a neurologist with Massachusetts General Hospital, who was not involved in the BRAVO study, told Axios, "It's now only a matter of years before there will be a clinically useful system that will allow for the restoration of communication." (Belluck, New York Times, 7/14; Falconer, Axios, 7/15; Marks, UCSF News, 7/14; Winkler, Wall Street Journal, 7/14)