A brain-computer interface (BCI) surgically implanted in a 45-year-old man with amyotrophic lateral sclerosis (ALS) and severe dysarthria has demonstrated remarkable success in restoring conversational communication, according to findings from the BrainGate2 trial.

On the first day of use, just 25 days after surgery, the BCI achieved an impressive 99.6% accuracy with a 50-word vocabulary. By the second day, it reached 90.2% accuracy using a vocabulary based on a 125,000-word dictionary, reported David Brandman, MD, PhD, from the University of California Davis, and his colleagues in the New England Journal of Medicine.

With continued data collection and training, the BCI maintained a 97.5% accuracy rate over 8.4 months following implantation.

“This paper presents the most accurate speech neuroprosthesis ever reported,” Brandman said in a statement. “Previous speech BCI systems frequently made word errors, hindering consistent communication. Our goal was to develop a system that enables users to be understood whenever they wish to speak.”

The BrainGate2 study involved the implantation of four microelectrode arrays into the patient’s left ventral precentral gyrus. These arrays recorded cortical neural activity from 256 electrodes, detecting the patient’s attempts to move his muscles to speak. The decoded words were displayed on a screen and read aloud in a voice designed to resemble the patient’s pre-ALS speech.

The patient was able to communicate at a rate of approximately 32 words per minute, using the system for over 248 cumulative hours, Brandman and his co-authors reported.

This study was one of two BCI trials published in the New England Journal of Medicine this week. The second study, led by Mariska Vansteensel, PhD, from the University Medical Center Utrecht in the Netherlands, documented the 7-year outcomes of a 58-year-old woman with advanced ALS who participated in the Utrecht NeuroProsthesis (UNP) trial.

The UNP study aimed to help locked-in patients like her communicate at home. In 2015, the researchers implanted a BCI with subdural electrodes placed over the patient’s motor cortex and a transmitter positioned subcutaneously in her thorax. The system, combined with decoding software, enabled her to interact with others.

For more than three years, the BCI served as the patient’s sole means of communication and her only way to attract her caregiver’s attention. As her control over an eye-gaze tracking device diminished, her use of the BCI increased, although it gradually decreased again starting six years after surgery.

Toward the end of the study, the BCI’s performance declined, and the patient ceased using it at home in 2023. Despite this, no technical malfunctions were observed. CT scans indicated progressive atrophy, which may have impaired her ability to produce reliable neural signals, suggesting that “ALS-related neurodegeneration ultimately rendered the brain-computer interface ineffective after years of successful use,” Vansteensel and her co-authors wrote.

“This study raises important questions about the long-term performance of brain-computer interfaces in patients with progressive neurodegenerative conditions like ALS,” commented Edward Chang, MD, from the University of California San Francisco, in a corresponding editorial.

Chang noted that future research may need to focus on interfacing with different brain regions that are less susceptible to degeneration, though decoding activity from these regions may prove more challenging.

Both studies published in this edition of the New England Journal of Medicine “offer compelling new evidence of rapid progress in the development of clinically viable, practical brain-computer interfaces for restoring communication in people living with paralysis,” Chang added.

He emphasized that while these studies used older-generation neural interfaces with limitations in electrode counts and long-term stability, they clearly demonstrate the meaningful benefits that communication BCIs can provide. The development of newer devices and neural interfaces is critically needed to advance this technology further.

By Impact Lab