Carried out in the United States, this scientific feat marks the first stage in the development of a technology that could make it possible to speak by thought.
- The man, who had been mute for fifteen years following a stroke, was able to express himself thanks to a neuroprosthesis which transcribed the words he imagined he was saying aloud.
- The accuracy of the device is 76%.
Transcribing into words the thoughts that a mute man imagines he is saying aloud: this is the feat that a scientific team from the University of San Francisco (United States) has managed to achieve.
In an article published in the New England Journal of Medicineand relayed by the wall street journal, they explain that they have developed a speech neuroprosthesis that converts the tiny electrical signals emitted by the brain into actions. And, more specifically, in words, typed on a computer.
76% accuracy thanks to algorithms
This neuroprosthesis was tested on a man who lost his speech following a serious stroke that occurred more than fifteen years ago. Today, the latter communicates by typing letters on a screen using a pointer worn on a cap. He agreed to have electrodes grafted onto the outer surface of his brain.
For 50 sessions over 81 weeks, the researchers hooked up a computer to the electrode array to record his brain activity as he observed words displayed on a screen and imagined saying them out loud. In 47% of cases, they were able to identify this word precisely. They then incorporated word prediction algorithms similar to those found, for example, in automatic email suggestions and word processors. This accuracy then climbed to 76%.
A technology still in the experimental stage
“To our knowledge, this is the first successful demonstration of the direct decoding of complete words from the brain activity of a person who is paralyzed and unable to speak, says Dr. Eddie Chang, a neurosurgeon at the university and lead author of the paper. This method holds great promise for restoring communication by harnessing the brain’s natural speech mechanisms.”
For the time being, this neuroprosthesis remains at the experimental stage. Indeed, the high error rate of the system, its limited vocabulary and the considerable time required to train it to recognize the imagined words are all obstacles to overcome before this device is available on the market. But this first step offers a glimmer of hope for the thousands of people who each year lose the ability to speak due to injury or illness.
.