Some physical injuries and neurological conditions can temporarily or permanently impair movement, leaving some people unable to speak, type on keyboards or use electronic devices. Brain-computer interfaces (BCIs), systems that can decode brain activity patterns and convert them into computer commands or written text, could be of great value for paralyzed patients.
Despite their potential, most of the best-performing BCIs developed to date require patients to undergo invasive surgical procedures. These systems typically rely on small sensors that need to be implanted on or within the brain and can detect electrical signals associated with neural activity.
Researchers at Meta artificial intelligence (AI), Université PSL and Hospital Foundation Adolphe de Rothschild recently introduced a noninvasive brain activity-to-text approach that does not require surgical procedures. Their proposed approach, presented in Nature Neuroscience, combines a new deep learning algorithm with electroencephalography (EEG) or magnetoencephalography (MEG) recordings.
“Modern neuro-prostheses can now restore communication in patients who have lost the ability to speak or move,” wrote Jarod Lévy, Mingfang Zhang and their colleagues in their paper. “However, implanting these invasive devices comes with risks inherent to neurosurgery. We introduce a non-invasive method to decode the production of sentences from brain activity and demonstrate its efficacy in a cohort of healthy volunteers.”
A noninvasive system for converting brain activity into text
Lévy and his colleagues first recruited 35 participants with no known medical or neuropsychiatric conditions. They asked these individuals to memorize specific sentences and recorded their brain activity as they typed the sentences on a standard keyboard.
As the participants typed the sentences, the team recorded their brain activity using EEG and MEG, two noninvasive techniques. EEG measures the brain’s electrical activity using sensors placed on the scalp, while MEG can detect tiny magnetic fields generated by active neurons using a helmet filled with ultrasensitive sensors.
The researchers then developed a deep learning algorithm and trained it separately on the EEG and MEG recordings they collected. During training, the algorithm learned to predict the characters that participants were typing based solely on their brain activity.
“We present Brain2Qwerty, a new deep learning architecture trained to decode sentences from either EEG or MEG, while participants typed briefly memorized sentences on a QWERTY keyboard,” explained the authors. “With MEG, Brain2Qwerty reaches, on average, a character error rate of 29% and substantially outperforms EEG (character error rate: 65%). For the best participants, the model achieves a character error rate of 18% and can perfectly decode a variety of sentences outside of the training set.”
In the team’s experiments, the model could predict the characters that participants were typing more accurately from MEG recordings than from EEG recordings. For some participants, the system could predict the characters they would type with an error rate of 18%, a promising early result.
Boosting the safety of brain-to-text communication technologies
This study highlights the potential of emerging AI-enhanced BCIs that do not require the surgical implantation of electrodes. In the future, the team’s deep learning algorithm could be improved further to translate brain activity into written text with even lower error rates.
“Overall, these results narrow the gap between invasive and non-invasive methods and thus open the path for developing safe brain–computer interfaces for noncommunicating patients,” wrote the authors.
Eventually, the recent work by Lévy and his colleagues could contribute to the creation of safer assistive technologies that allow people with motor impairments to communicate with others or use digital devices. Their efforts could also inspire the development of more AI systems that can translate brain activity recorded noninvasively into written text and, potentially, nto commands for prosthetic systems or other assistive technologies. https://techxplore.com/news/2026-08-noninvasive-ai-based-brain-written.html
Recent Comments