TEL AVIV – For individuals facing the profound isolation of speech loss due to paralysis, a new innovation offers a beacon of hope. Researchers have developed a brain-computer interface (BCI) capable of translating brain activity into audible words, potentially enabling those with conditions like advanced Amyotrophic Lateral Sclerosis (ALS) or post-stroke aphasia to communicate naturally once more [[2]].

A window into the Brain

The groundbreaking work was conducted during epilepsy surgery at Ichilov Hospital, where patients undergo electrode implantation to pinpoint seizure origins. In this instance, the epileptic focus resided in the frontal lobe and hippocampus – areas intricately linked to speech production, perception, and even imagined speech.

The research team, led by Dr. Ariel tankus from Tel Aviv University’s Gray Faculty of Medical and Health Sciences and Sagol school of Neuroscience, in collaboration with Dr. Ido Strauss, Head of the Functional Neurosurgery Unit at Ichilov, published their findings in *Neurosurgery*, the journal of the World Federation of Neurosurgical Societies. The study received support from the ministry of Innovation, Science and Technology.

“Fortunately, the patient agreed to participate in the experiment, which in the future may help paralysed individuals express themselves again through artificial speech,” says Dr. Ariel Tankus.

“We have, for the first time in history, succeeded in linking elements of speech to the activity of individual cells.” – dr.Ariel Tankus

From Thought to Audible Sound

Initially, the patient was instructed to vocalize two phonemes: /a/ (“ah”) and /e/ (“eh”). Microelectrodes, finer than human hair, meticulously recorded the activity of individual brain cells as he spoke.

“The first stage of the experiment is essentially data collection to train the computer,” explains Dr. Ariel Tankus. “Each microelectrode records data from a single cell, or at most two or three neighboring cells.”

These recordings were then used to train machine learning models to discern the brain’s electrical patterns associated with these sounds. Subsequently, the patient was asked to merely imagine uttering the same sounds. The computer then decoded this activity and reproduced the corresponding sounds.

A Novel Approach to Speech Decoding

the TAU-Ichilov team distinguished itself by recording from multiple, dispersed brain regions, rather than concentrating on a single location.

“These are not the classic speech areas associated with activating the speech organs,” notes Dr. Ariel Tankus. “They are connected to higher levels of speech – speech production, speech perception, and imagined speech. We expect that in the future, systems will integrate signals from multiple regions, achieving broader coverage of the brain’s speech system.”

The Cutting Edge of Brain-Computer Interfaces

Brain-computer interface research is rapidly evolving on a global scale. Prior advancements include a 2021 Stanford University project that enabled a patient to “type” letters via thought, and a University of california, San Francisco (UCSF) team that successfully translated brain signals into writen words on a screen.

The TAU-Ichilov study represents a significant milestone, marking the first instance of converting imagined speech into audible sound with high precision.

The Foundation of language

For over a decade, Dr. Ariel Tankus has dedicated his research to understanding how speech is encoded and decoded at the single-cell level. This direct linking of phonemes to brain cell activity is a historic achievement.

“We have, for the first time in history, succeeded in linking elements of speech to the activity of individual cells,” says Dr.Ariel Tankus. “This allowed us to distinguish between the signals characteristic of /a/ and /e/.”

The team is now expanding their research to encompass additional sounds. “These are the building blocks of our language,” Dr. Ariel Tankus adds. “Once we can connect them together, we will ultimately enable a person to produce words, and hopefully even sentences.”

Dr. Ariel Tankus, Leader of the Study

Future directions

the next step involves decoding fluent, continuous speech – whether spontaneous or read aloud – directly from brain activity. The team is developing sophisticated artificial intelligence tools to predict phonemes and speech units in sequence.

“Our vision is that in the future, we will be able to train the computer on an ALS patient in the early stages, when speech is still intact,” explains Dr. Ariel Tankus. “Later, when the disease progresses and the patient can no longer use the speech muscles, the system will still know how to translate their brain signals into words.”

What began with two simple vowels now paves the way for a future where even those completely paralyzed may regain their ability to communicate freely,their voices restored through the synergy of neuroscience and artificial intelligence.

Dr. Ido Strauss, co-leader of the study.Photo: Tomer appelbaum