Researchers from Stanford University (USA) have coupled artificial intelligence software to a brain-computer interface (ICO) implanted in the brain of a man with tetraplegia.
The system has been able to decode the information of the ICO to quickly convert the thought or attempt to write by hand of the participant into visible text on the computer screen.
The researchers asked the participant to try to ‘write’ sentences as if his hand was not paralyzed, imagining that he was holding a pen on a piece of paper
The results of the study, published in Nature, show that this person achieved a write speed 90 characters per minute, with 94.1% accuracy, when you used this method.
The man was able to write more than twice as fast than with a previous method, developed by the same team and where letter by letter was selected on a virtual keyboard, and whose results were published in the eLife magazine in 2017.
The speed achieved now by the participant is comparable to that of the typical typing on a smartphone by people of the same age group (115 characters per minute), highlight the authors.
The researchers asked the participant to try to ‘write’ sentences as if his hand was not paralyzed, imagining that he was holding a pen on a piece of paper. During this exercise, the ICO used a neural network, a type of machine learning, to translate the intent of the movements from writing to text, from brain activity and in real time.
As noted Jaimie henderson, Professor of neurosurgery from the American university and one of the leaders of the work, “the findings could drive new advances that benefit millions of people in the world, who have lost the use of their upper extremities or their ability to speak due to spinal cord injuries, strokes or amyotrophic lateral sclerosis, also known as Lou Gehrig’s disease ”.
During the exercise, a neural network was used to translate the intent of writing movements, based on brain activity, into text in real time.
The volunteer, referred to as T5 in the experiments, lost virtually all movement below the neck due to a spinal cord injury in 2007.
Nine years later, Henderson placed two chips brain-computer interface, each the size of a baby aspirin, on his left side of his brain. Each chip has 100 electrodes that pick up signals from neurons that fire in the part of the motor cortex – a region on the outermost surface of the brain – that governs hand movement.
Artificial intelligence algorithms
These neural signals are sent through wires to a computer, where the algorithms artificial intelligence decodes the signals and deduces the intended movement of the hand and fingers from T5.
The algorithms were designed at the Stanford Neural Prosthesis Translational Laboratory, co-directed by Henderson and Krishna Shenoy, professor of engineering at this university.
The volunteer has two chips implanted in his left side of his brain. Each chip has 100 electrodes that pick up the signals from neurons that fire in the part of the motor cortex.
Shenoy and Henderson, who co-lead the work, have been collaborating in the field of brain-computer interface since 2005. For their part, the first author is Frank Willett, a Howard Hughes Medical Institute and Laboratory Investigator.
Willet highlights that in this research they have learned that the brain “retains its ability to predict fine movements a whole decade after the body has lost its ability to execute these movements.
Also, he concludes, “we have discovered that the artificial intelligence algorithms we use can more easily and quickly interpret complicated movements that involve speed changes and curved trajectories, such as handwriting.”
Rights: Creative Commons.