Tech News

Patients with Lou Gehrig’s disease can communicate in 60 words per minute… The Miracle of the Brain Transplant ‘Chip’

A patient who has been unable to speak due to Lou Gehrig’s disease, which gradually paralyzes muscles, can communicate at a speed similar to that of ordinary people thanks to a brain implanted chip.

Researchers at Stanford University found that an American woman (67) who lost her language skills due to Lou Gehrig’s disease eight years ago succeeded in communicating through text or mechanical voice at a speed of 62 words per minute using an EEG measuring chip implanted in her brain. Recently, it was published in ‘Bio Archive’, a repository for pre-published papers.

This is less than the speed of 160 words per minute for people without language impairment, but it is more than three times faster than the previous record of brain implanted chips (18 words per minute) set by researchers at Stanford University and the University of California, San Francisco.

In addition, the error rate (based on 50-word vocabulary) was 9.1%, which was 2.7 times lower than before, the researchers said. However, when broadened to a large vocabulary of 125,000 words, the error rate was still high at 23.8%.

The research team gave meaning to this study as “experimental results showing that it is feasible to communicate at the speed of conversation with a paralyzed patient.”

Phillip Saveth, a researcher at the University of California, San Francisco (UCSF), who was not involved in the study, told the technology media ‘MIT Technology Review’, “This is the level many people want.” It’s a huge achievement that could be a product.”

 

Read movement of articulation organs other than handwriting

The brain implant chip made by the research team is a chip with tiny electrodes that are inserted into the motor cortex of the brain.

Connected to a computer, the chip records the activity of dozens of neurons at a time. When a paralyzed person imagines an action, the chip transmits the nerve signals contained in the action, and the computer deciphers it. The research team has focused on developing technology that allows the user to imagine hand movements necessary for a specific action and reads them with a brain implanted chip. For example, if you imagine the action of writing a certain letter, you select that letter on the virtual keyboard. The researchers confirmed that it is possible to control video games and robotic arms in this way.

On the other hand, the research team this time focused on technology that deciphers the movements of the muscles of articulatory organs such as the mouth, tongue, and vocal cords when speaking, rather than hand movements. This is a much smaller and more complex movement than hand movements. The research team was able to achieve this research by finding the locations of four neurons that allow a computer program to almost accurately predict the words a paralyzed patient is about to say, and connecting them to a brain implanted chip.

 

Brain implantation chip 'Utah Array' developed by Professor Richard Norman at the University of Utah.  Provided by the University of Utah
Brain implantation chip ‘Utah Array’ developed by Professor Richard Norman at the University of Utah. Provided by the University of Utah

Great expectations for integration into the large-scale language model GPT 3

This technology, which deciphers the brain nerve signals related to the movement of the articulator when speaking, was previously developed by Dr. Edward Chang’s research team at the University of California, San Francisco. Using this technology, the research team published the results of an experiment that speaks 18 words per minute in the ‘New England Journal of Medicine (NEJM)’ in 2021. Researchers at Stanford University said they developed a system that is more accurate and three to four times faster than Dr. Chang’s team.

“The technology we applied is a promising approach, but it is still in the proof-of-concept stage and not fully clinically viable,” the researchers said.

In the future, the researchers plan to connect this system with artificial intelligence to further increase its accuracy and speed. He says he is testing several machine-learning programs, including software that predicts what comes next.

The research team expects that accuracy and speed will increase significantly if a large-scale language model (LLM) such as GT3, which has recently been gaining sensational popularity, is connected to the brain implanted chip.

This achievement by Stanford University researchers is one of the projects of a brain implantation chip research consortium called ‘Braingate’, in which researchers from various institutions participated. Currently, more than 10 volunteers have participated in BrainGate and have electrodes implanted in their brains.

The brain transplant chip they use is a rectangular metal chip with 100 fine needles called ‘Utah Array’.

Dissertation information

https://doi.org/10.1101/2023.01.21.524489

A high-performance speech neuroprosthesis

bioRxiv

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button