"Speaking" again after 18 years of paralysis and aphasia!

Thousands of people lose the ability to speak due to stroke, accident or disease every year, which can be a huge disaster for patients

The study subject was a right-handed man, 36 years old when the study started

The researchers implanted electrodes in the sensorimotor cortex of the patient through craniotomy (the electrode array is placed on the surface of the pia mater in the subdural space)

▲The screen will present a paragraph to the subject, and the subject will try to answer using the words in the vocabulary (containing 50 words)

In multiple tests over an 81-week period, the researchers recorded the neural activity of the subjects when they wanted to say a word or sentence

In order to test the effect, the researcher feeds the results back on the computer screen (pictured above)

It was found that the researchers decoded sentences in real time based on the subjects’ cortical activity, with a median decoding speed of 15.

In the paper, the researchers pointed out that the language decoding method usually works after the word error rate drops below 30%, and the signal recorded by the implanted device is stable during the research period, and it can be successfully decoded without frequent repetition and correctness.

The author concluded that for this stroke-induced paralyzed patient with dysarthria, through machine learning and natural language modeling, words and sentences can be decoded in real time based on the cerebral cortex activity collected by the nervous system implant device when the patient tries to speak.

Note: The original text has been deleted

references:

[1] Moses, DA, Metzger, SL, Liu, JR, Anumanchipalli, GK, Makin, JG, Sun, PF, .

[2] A Paralyzed Man's Brain Waves Converted to Speech in a World-First Breakthrough.