2023-08-31 08:10:30
The technique decodes neural signals and translates them into words with an AI
Two American women suffering from paralysis – one due to ALS and the other due to a stroke – that prevents them from speaking have been able to communicate with their environment again thanks to brain implants that decode neural activity and translate it into words. These are two technologies presented this Wednesday in separate articles in the journal Nature, which facilitate communication with a speed, precision and richness of language unprecedented up to now.
A team from the University of California at San Francisco (UCSF), in the United States, demonstrated in 2021 that it was possible to decode the brain signals that a person produces when trying to speak and transform them into text. His first attempt allowed a severely paralyzed man to communicate with a vocabulary of 50 words. The system showed that translation was possible, but it was limited: it was wrong one out of four times and transcribed the signals at a rate of 18 words per minute, much slower than in normal conversation, where the speed is about 160.
The implants presented this Wednesday by UCSF itself and Stanford University, both in the United States, multiply the speed and richness of communication. The former have achieved a rate of 78 words per minute formulating sentences with a vocabulary of more than 1,000 different terms, while the latter have reached 62 words per minute, but with a vast vocabulary of 125,000 words. The results bring closer the possibility that people who have lost their voice can maintain fluent conversations with their environment and “are a true milestone in the field,” says Edward Chang, a UCSF neurosurgeon and leader of one of the investigations.
Both technologies collect neural activity that should activate the patients’ tongue, pharynx, jaw and facial muscles, allowing them to speak if they were not paralyzed, and use artificial intelligence to transform the signals into words. However, the groups differ in the way they collect the data and train the AI.
While scientists at the University of California have read the neural activity of all the cells on the surface of the brain, the Stanford team has inserted electrodes inside the patient’s cerebral cortex to read neuron-by-neuron activity.
The fact that both approaches have given similar results fills researchers with optimism. “The most important message is that there is hope, that this will continue to improve and that it will provide a solution in the coming years,” concludes the UCSF neurosurgeon. For now, both technologies are purely experimental.
To translate the signal into words, research teams and patients trained individual artificial intelligences for hundreds of hours. Stanford University asked its volunteer to repeat over 10,000 different phrases randomly taken from telephone conversations over 25 days. The algorithm was capable of translating neural impulses into words from a vast vocabulary of more than 125,000 terms, making mistakes in 24% of them. Despite the fact that the failure rate is high, it is the same as two years ago with a much poorer language, of only 50 words. The new technology was wrong only one in ten times with this very small vocabulary.
Instead, the University of California opted to train the AI by repeating phrases from a vocabulary of about 1,000 words over and over again. With this they managed to get the system to err only 5% of the terms when trying to verbalize phrases from a repertoire of 50 statements. For new formulations, the error occurred again in one out of every four words. L.D.B. (SyM)
#women #lost #speech #manage #communicate #brain #implant #Health #Medicine