science
The hardware isn’t new, but a team of researchers at the University of California, Davis has developed a machine learning-based method that translates brain activity in ALS patients into sentences with 92% accuracy.
Imagine being severely paralyzed, not only unable to move your arms or legs, but also unable to speak. For years, brain-computer interfaces have offered the enticing promise of reading brain waves well enough to allow humans to communicate and access PCs.
Now, a new groundbreaking discovery has revealed how people afflicted with movement-depriving illnesses are able to speak and even work.
A team of scientists at the University of California, Davis published a paper on Monday detailing years of research into a brain-computer interface (BCI) system implanted in patients with amyotrophic lateral sclerosis (ALS, also known as Lou Gehrig’s disease), which destroys motor neurons and causes loss of motor control and eventual paralysis. The patient, Casey Harrell, has been living with a BCI implant since 2023 and is still functioning, allowing him to control a computer cursor with his thoughts as well as gaining the ability to speak, according to the research team.
Davis’ team is part of a broader university alliance with the U.S. Department of Veterans Affairs known as BrainGate. They work on a variety of neuroscience projects to restore language, use computers, and in some cases, restore movement. In Harrell’s case, Davis’ team was trying to find a way to turn experimental techniques into something shelf-stable and practical that could be used outside the lab.
University of Davis neurosurgeon David Brandman, co-principal investigator and co-senior author of the paper published Monday and the surgeon who placed Harrell’s implant, said the results his team presented exceed the limits of BCI technology, and said Harrell’s implant has not only performed well in daily use since 2023, but is also incredibly accurate.
In controlled tests, the system was able to synthesize sentences from Harel’s brain activity with 99% accuracy. Even when used routinely outside the lab, Harrell rated it as accurate 92% of the time.
“What’s important to me is that it allows everyday communication for men who want to talk but can’t,” Brandman said. register In an interview. “Despite being paralyzed, [Harrell] He’s back to work full time and having meaningful conversations with his daughter, who has never heard his voice. ”
Previous research in the field of BCI, Brandman said, has always required researchers to be in the patient’s home or the patient to come to the researcher to use the technology. That wasn’t the case in this case, as the system allowed Harrell’s home care team to connect Harrell himself to the system, allowing Harrell to use the device for more than 3,800 hours over the past few years. Based on the time the study was submitted (it was published on Monday but entered peer review in July 2025), Harrell would have been using the device for an average of more than five hours a day.
“This is a more dynamic, action-filled life with friends, family, and colleagues, and one that allows us to communicate more in a natural way of communicating than any technology I’ve ever experienced,” Harrell told UC Davis through the BCI system.
Practical use of AI
Brandman is no stranger to BCI technology. As well as being a key figure in the BrainGate consortium, he has also worked as a principal investigator investigating the safety of commercial BCI technology for Paradromics, one of the leading companies in the field along with Synchron and Neuralink.
As Brandman explained, Davis’ research did not involve specialized hardware, but instead utilized an existing BCI design created by Blackrock Neurotech. The Davis neurosurgeon said his team’s use of machine learning techniques was a major advance.
The lab has built a unique software platform for operating BCI devices known as the Brain-Computer Interface for Rapidly Adaptive Neural Decoding (BRAND). Brandman said the name was serendipitous, and UCD postdoctoral fellow Nick Card built the machine learning algorithm. The brand is now used across the BrainGate consortium and is the secret to the project’s success.
According to the paper, BRAND’s AI algorithm can translate activity in Harel’s ventral precentral gyrus, the part of the brain that controls motor functions of the face, mouth, and jaw, into English phonemes. Additional algorithms in the software map these phonemes to words and words to sentences. The result is highly accurate speech synthesis, allowing Harrell to work full-time as an environmental advocate.
As for when the technology the UCD team is developing will reach the commercial market, Brandman says other technologies in the BCI space, such as Neuralink, are all approaching the technology with the same kind of goals. His team’s goal is simply to prove that BCI systems are more than just dead-end laboratory experiments.
“My job is to avoid danger,” Brandman said.
He likened the current state of BCI technology to early pacemakers. Pacemakers began in the 1950s and required wiring to hardware outside the body, often connected to a large battery or tethered directly to a wall. Seventy years later, pacemaker implantation is so simple that it is often performed as an outpatient surgery.
“We’re in the early stages of this kind of technology,” Brandman said. “Casey has proven that this type of technology is practical.”
Harrell may be currently connected to a slew of large external computers, but when the Davis UCD team’s AI advances are combined with the hardware work being done by other companies, the future looks bright for many people whose lives are limited by paralysis or other impairments.
“I really don’t want to be special or special because that would mean I don’t have the disease anymore or it would mean that everyone who has the disease like me could get the disease.” [BCI] It was prescribed to them,” Harrell said.
BrainGate is currently accepting applications for future research participants. ®
