A 65-year-old man who suffered a severe spinal cord injury a decade ago is able to write with his own thoughts
A new study published in the journal Nature shows that for the first time, neuroscientists have converted cognitive signals associated with writing into text.
The new technique surprised researchers by doubling the speed of writing, with the paralyzed man able to send text messages at a rate of 90 characters per minute.
Designed by Brain Gate, the system is designed to “enable people facing severe speech and movement disorders to communicate through text messaging, email or other forms of writing,” said Jimmy Henderson, co-director of Stanford’s Translational Neuroprosthetics Laboratory and co-author of the research paper.
The system uses brain implants and machine learning algorithms to decode brain signals associated with handwriting and bring paralyzed people into the digital world of modern communication.
Brian Gate is an alliance that has made significant advances in the field of brain-computer interfaces (BCIs) in recent years.
The company’s inventions include a precision robotic arm controlled by the brain (originally introduced in 2012) and the latest high-bandwidth wireless brain-computer interface for human use.
The new project, led by Stanford scientist Frank Willett and overseen by Howard Hughes Medical Institute neuroscientist Christine Sinoy and Stanford neurosurgeon Henderson, aims to develop brain-computer interfaces that can be used for writing.
Sinoy and his colleagues developed a text-thinking system that improves on technology used in 2017, when it allowed monkeys to write text at a rate of 12 words per minute.
That success set the stage for a brain-computer interface that emerged later that year, which allowed paralyzed people to write 40 letters per minute.
In an email to the Gizmodo website, Henderson wrote, “The current speed is more than twice the speed at which paralyzed people were writing in 2017, and it seems right that a new and different approach is being used today, because until then, no one had tried to capture and repeat the mental action of handwriting, and it was from the design of this new experiment to convert ideas into text This new experiment began with the design of a system to convert ideas into text.”
The test subject was a 65-year-old man who had suffered a severe spinal cord injury a decade earlier.
The two sensors were 4 x 4 millimeters long and wide, the size of an aspirin, and had 100 microelectrodes that were placed in the outer layer of the motor cortex, the area that controls movement on the other side of the body,” Henderson said.
He added: “These electrodes can record signals from about 100 neurons, which are then processed by a computer to decode the brain activity associated with writing individual letters.”
Brain-computer interface turns mental activity into words (social networking site)
Halfway through the experiment, the man tried to move his paralyzed hand as he would when writing words physically. He imagined in his mind the action of “writing letter after letter with a pen on a yellow board,” and the decoder transcribed each letter he wrote in his head and “recognized it by the neural network. The “greater than” sign is used to indicate spaces between words, which becomes the only way to identify the author’s desire or intention to leave spaces between letters.
The system can distinguish individual letters with an accuracy rate of about 95%; however, it does have some definite limitations.
Brain surgery is very dangerous. In addition, machine learning systems are not universally accepted and must adapt to the cognitive differences of different users, as well as “very intensive computational processes that require high-performance dedicated computers or computing clusters.
We are still a long way from hearing doctors recommend brain-computer interfaces for paralyzed people. But the system could one day change the way we write forever.