In a groundbreaking new study, scientists were able to measure brain activity and brain waves that they examined in paralyzed people. The team then managed to convert the signals into written text. Anarthria, i.e. the loss of the ability to speak, can have various causes, such as a stroke or aamyotrophe Lateralsklerose, have. Affected individuals often retain their language skills, but paralysis can prevent them from using technological tools to communicate. So neuroscientists used machine learning to translate electrical activity in the part of the brain responsible for language into words and sentences.
How neuroscience can measure and implement brain waves
The research team implanted electrodes the size of a credit card into the subjects' sensorimotor cortex. This is the part of the brain that controls the articulation of speech. The scientists then trained computer models using deep learning algorithms to recognize and classify words based on patterns in participants' brain activity. The result was a milestone on the path to restoring speech for people who cannot speak because of paralysis. After 48 training sessions with the new mind-reading technology, the subjects were able to generate around 15 words per minute in 81 weeks with an error rate of 26%. According to the authors, this study represents a transformative breakthrough in the field of brain-computer interfaces. Although the procedure is invasive and requires brain surgery, the paradigm is groundbreaking for them. The most surprising thing is that the sensorimotor cortex does not have to play an important role in language comprehension or word formation.
The algorithm and language model made it possible to decode thoughts without errors in more than half of the sentence attempts. After training, the computer model managed to correctly recognize 98% of participants' attempts to produce single words. The method used decoded brain activity in real time and was able to measure brain waves in order to produce texts. The big stepthis studyis that the authors present their results to patients who can actually no longer speak. To date, all such experiments have been carried out with subjects who could speak and communicate.
