Researchers at UC Davis Health have unveiled a groundbreaking brain-computer interface (BCI) that translates brain signals into speech with an unprecedented accuracy rate of up to 97%. This innovative system represents a major advancement in assistive technology, offering new hope to individuals with speech impairments due to neurological conditions like amyotrophic lateral sclerosis (ALS).
The team tested the BCI on a man severely affected by ALS, a neurodegenerative disease that progressively impairs motor functions, including the ability to speak. Within minutes of activating the system, the patient was able to communicate his thoughts effectively, marking a significant milestone in restoring speech capabilities for those who have lost them.
The research findings were published in the New England Journal of Medicine, detailing the system’s potential to transform communication for people with paralysis or severe neurological conditions. ALS, commonly known as Lou Gehrig’s disease, deteriorates nerve cells responsible for movement, leading to loss of mobility and, eventually, the ability to speak clearly.
This new technology interprets brain activity during speech attempts and converts these signals into text, which a computer then vocalizes. “Our BCI technology has enabled a man with paralysis to reconnect with his loved ones through speech,” said David Brandman, a neurosurgeon at UC Davis and co-principal investigator of the study. “This is the most accurate speech neuroprosthesis reported to date.”
Brandman, an assistant professor in the Department of Neurological Surgery at UC Davis, also co-directs the university’s Neuroprosthetics Lab. He emphasized the device’s potential to break communication barriers for individuals with severe speech impairments.
The breakthrough system was developed as part of the BrainGate clinical trial, with Casey Harrell, a 45-year-old ALS patient, participating. Harrell, who had lost much of his ability to move and speak, was the first to benefit from this technology. In July 2023, four microelectrode arrays were implanted into Harrell’s left precentral gyrus, a brain area responsible for speech coordination. These arrays capture brain activity through 256 cortical electrodes.
“We’re detecting their effort to speak,” explained Sergey Stavisky, a neuroscientist and co-director of the UC Davis Neuroprosthetics Lab. “We record brain signals that would normally command speech muscles and translate these patterns into phonemes—the basic units of speech—and then into the intended words.”
Despite previous advances in BCI technology, decoding speech from brain signals has been a slow and error-prone process. This new system significantly reduces those errors, allowing for clearer and more reliable communication.
“In the past, frequent word errors made it hard for users to be consistently understood, which limited their ability to communicate,” said Brandman. “Our goal was to develop a system that allows users to be understood whenever they want to speak.”
The system was tested in both structured and spontaneous conversational scenarios, with the decoded speech appearing on a screen in real-time. Remarkably, the computer voiced the text in a tone resembling Harrell’s voice before ALS affected his speech. This voice was recreated using software trained on recordings of his pre-ALS voice.
During the first training session, the system achieved a word accuracy of 99.6% with a limited 50-word vocabulary in just 30 minutes. In subsequent sessions, the vocabulary was expanded to 125,000 words, with the system maintaining an impressive 90.2% accuracy after only 1.4 additional hours of training. Ultimately, the system consistently achieved 97.5% accuracy.
“At this stage, we can accurately decode what Casey intends to say 97% of the time, surpassing the performance of many voice-recognition apps available today,” Brandman noted. “This technology is transformative because it offers hope to those who have lost the ability to speak. I believe this BCI could enable future patients to communicate with their loved ones once again.”
The study included 84 data collection sessions over 32 weeks, during which Harrell used the BCI for over 248 hours in self-paced conversations, both in person and via video chat.
Reflecting on the impact of the technology, Harrell said, “Losing the ability to communicate is incredibly frustrating. This technology can help people reengage with life and society.”
Nicholas Card, the study’s lead author, expressed his satisfaction in seeing Harrell regain his ability to speak. “Casey and our other BrainGate participants are truly remarkable. They’ve joined these trials not for personal gain but to help us develop systems that will restore communication for others.”
Leigh Hochberg, a neurologist and neuroscientist involved in the BrainGate trial, emphasized the participants’ critical role in advancing this life-changing technology. “Their contributions are invaluable as we strive to develop solutions that will help others facing similar challenges.”
For more information about the ongoing BrainGate2 clinical trial, visit BrainGate.
