'Neuroprosthesis' gives paralyzed man ability to communicate in sentences

Dr. Edward Chang performing brain surgery. (Photo by Barbara Ries)

TORONTO — Researchers have successfully given a paralyzed man the ability to speak in full sentences again through the use of a “speech prosthesis.”

The University of California San Francisco (UCSF) researchers developed a technology with the help of a clinic trial patient that builds upon years of prior research conducted by UCSF neurosurgeon Dr. Edward Chang. Their findings were published on July 15 in the New England Journal of Medicine.

“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, the chair of neurological surgery at UCSF and senior author on the study, said in a press release.

The first trial participant, BRAVO1, was a male in his late 30s who had lost his ability to speak after suffering a brainstem stroke over 15 years ago.

He helped researchers build a 50-word vocabulary that Chang’s team could recognize based on his brain activity patterns. The vocabulary included words such as “water,” “family,” and “good” and allowed BRAVO1 to communicate hundreds of relevant sentences.

At first, BRAVO1 was given constructed sentences to attempt repeatedly, with the words appearing on a screen in response. Soon, he was able to answer questions such as: “How are you today?” and “would you like some water?” BRAVO1’s attempted speech responses appeared on a screen: “I am very good,” and “No, I am not thirsty.”

The UCSF team's findings and success suggest that communication can be restored in the thousands of people who have lost their ability to speak due to illness, injury or stroke.

“It shows strong promise to restore communication by tapping into the brain's natural speech machinery,” said Chang.

Chang joined forces with Dr. Karunesh Ganguly, an associate professor of neurology at UCSF, to conduct the Brain-Computer Interface Restoration of Arm and Voice (BRAVO) study.

For the study, Chang surgically implanted high-density electrode array over BRAVO1’s speech motor cortex from which the researchers recorded 22 hours of activity over nearly 50 sessions and several months. During the sessions, BRAVO1 would attempt to say each word in the 50-word vocabulary several times while the implanted electrodes recorded the brain activity.

To translate the patterns, bioengineers and authors on the study Jesse Liu and Sean Metzger used a form of artificial intelligence that could distinguish between subtle changes in brain activity to determine speech attempts and identify the words BRAVO1 was trying to say.

The researchers system was able to decode words quickly, 18 per minute, and with a 93 per cent accuracy thanks to an auto-correct system similar to what is used in texting.

“We were thrilled to see the accurate decoding of a variety of meaningful sentences,” said Moses. “We’ve shown that it is actually possible to facilitate communication in this way and that it has potential for use in conversational settings.”

Previous work on restoring speech to people who lost the ability focused on spelling out words, typing letters one-by-one, but Chang’s approach builds whole sentences by translating the signals that control the body’s vocal system. Previous work used brain signals sent to the arm or hand to simulate typing.

“With speech, we normally communicate information at a very high rate, up to 150 or 200 words per minute,” he said.

Spelling- and typing-based approaches are typically much slower than the typical speech process.

“Going straight to words, as we’re doing here, has great advantages because it’s closer to how we normally speak,” he added.

Over the span of a decade, Chang was able to work toward developing this technology with the help of his patients at UCSF Epilepsy Center. He asked patients who were undergoing neurosurgery to search for the origins of their seizures if he could use their brain recordings to analyze for speech-related patterns. All of the patients had normal speech, according to the press release.

Chang and his team had successes identifying brain activity patterns correlating with vocal system movements that formed vowels and consonants. David Moses, an engineer on the team and one of the lead authors on the new study, was able to develop a method for decoding these brain patterns and used statistical models to increase accuracy.

Despite this development, the researchers weren’t sure it would work in patients who had lost their ability to speak, particularly those who had been unable to speak for years, until they worked with BRAVO1.

“The best way to find out whether this could work was to try it,” Moses said in a press release.

Chang and team hope to be able to broaden the clinical trial to include more patients, and also to increase the vocabulary.

“This is an important technological milestone for a person who cannot communicate naturally,” said Moses, “and it demonstrates the potential for this approach to give a voice to people with severe paralysis and speech loss.”