A tetraplegic capable of writing with his mind with brain chips
Posted on May 12, 2021
The typing speed of a spinal cord injured was close to that of other people typing on their mobile.
Sensors in the brain of a tetraplegic have captured neural activity while imagining that he was writing and have converted it, letter by letter, into text. This brain-computer interface (BCI) has achieved a writing speed close to that of anybody while typing on a mobile phone. In addition, with the help of a simple text predictor, it transcribes almost without errors all the signs of words and complete sentences.
Writing or speaking just thinking about it can be a dream, but it is a necessity for those who due to illness or injury cannot communicate. It depends on the pathology, but systems have been devised that track eye movements that activate a virtual keyboard. Others are mechanical, like the sophisticated interface Stephen Hawking used to turn his right cheek movements into a robotic voice. There are also those who seek to convert brain activity into spoken or written words. This is what has been investigating a multidisciplinary group of neuroscientists, bioengineers and experts in artificial intelligence with T5 (as researchers have called it to respect their privacy), a 65-year-old person who lost mobility from the neck down after a spinal cord injury in 2007.
Ten years after he completely lost the mobility of his hands, T5 was able to use a virtual keyboard just by thinking that he was moving his hand, as if he were operating a cursor on the screen. Now, scientists from the American universities of Stanford, Brown and Harvard have gone further, going to the part of the motor cortex where the brain stores when and how we learned to write. After opening the skull, they placed two chips with 100 electrodes each in the so-called precentral gyrus. And they told T5, literally, "come on, think you're writing". This BCI observed that electrical activity that followed the same pattern occurred when he imagined that he was writing the same letter. After training the algorithm for several days, the system was able to write the strokes of each sign, even commas, accents or question marks with a very low error rate.
"You can write a complex shape like a letter by hand and that provides a distinctive mark of unique neural activity"
KRISHNA SHENOY, NEUROSCIENTIST AT STANFORD UNIVERSITY, UNITED STATES.
The creators of it put their BCI to the test and thoroughly. T5 began by thinking that he was writing series of words that were coming out on the screen. But after 28 days of training, this man's brain was already writing his own sentences to open-ended questions. Even then, with an algorithm he had never seen such letter combinations before, the speed reached 73.8 characters, or about 18 words, per minute. And he was little wrong. On average, this artificial intelligence system converted T5's thoughts into text correctly 94% of the time. And with the help of a text predictor, the percentage rose to 99.1%.
The design, conveniently patented, still needs a lot of progress before it leaves the laboratory. Dealing with one's particular writing style shouldn't be difficult for an algorithm designed to learn. Another thing will be to see how it works with other alphabets that are neither English nor the others based on Latin. More complicated is the mechanical part of the system. Placing electrodes in the brain itself is not an easy task, although it is already done routinely with techniques such as deep brain stimulation in Parkinson's cases, who spend years with chips under their skull. But these electrodes only emit electrical pulses and those of a brain-computer interface like the one now shown need to read brain activity and send the data wirelessly (T5 had a pair of wires coming out of its head). So there is still a lot of engineering to do.
The Spanish neuroscientist José Carmena, from the University of California at Berkeley, has been researching this field for years. Regarding this research, he says that, however preliminary his results may be, "it is a great advance in this field", referring to brain-computer interface technologies that seek to help with different disabilities. "This research is the perfect example: the interface decodes the thought of writing and produces action".
Source: El País
Leave a Comment:
Daily Dose of Positivity