A BCI decodes neural signals acquired from electrodes placed on the surface of brain areas responsible for speech and upper limb function. Participant then navigates options on a communication board using a set of six navigational commands to control devices like room lights. Photo courtesy of Crone Lab.
A brain-computer interface (BCI) surgically implanted on the brain of an ALS (amyotrophic lateral sclerosis) patient has shown success in translating brain signals into computer commands, according to a new study led by Johns Hopkins Medicine researchers in collaboration with the Johns Hopkins University Applied Physics Laboratory.
A brain-computer interface is a device that can translate brain signals into computer commands.
The results of the study, published Oct. 25 in Advanced Science, showed that computer commands were accurately translated from brain activity over a three-month period without requiring the BCI algorithm to be retrained or recalibrated.
For the study, the BCI (called CortiCom) was surgically implanted on the surface of brain areas responsible for speech and upper limb function in a patient with ALS, a progressive nervous system disease that causes muscle weakness and loss of motor and speech functions. Tim Evans, 62, was diagnosed with ALS in 2014 has developed severe speech and swallowing problems. He can talk slowly, but it’s hard for most people to understand him.
By using the BCI with a special computer algorithm trained to translate his brain signals into computer commands, Evans was able to freely and reliably use a set of six basic commands (up, down, left, right, enter and back) to navigate options on a communication board to control smart devices like room lights and streaming TV applications.
“While past studies on speech BCI have focused on communication, our study addressed the need to control smart devices directly,” says Nathan Crone, M.D., professor of neurology at Johns Hopkins Medicine, senior author of the study. “The BCI accurately recognized a set of 6 commands from neural signals alone, allowing Tim to navigate a communication board and control household devices without needing a language model to fix errors.”
HOW IT WORKS
In the summer of 2022, William Anderson, M.D., Ph.D., M.A., professor of neurosurgery, and Chad Gordon, D.O., professor of plastic and reconstructive surgery, both at the Johns Hopkins University School of Medicine, placed two soft plastic sheets of flat electrodes on the surface of Evans’ brain. The electrode sheets were the size of large postage stamps and were each used to record the electrical signals produced by tens of thousands of brain cells (neurons).
Evans worked with the research team for a few weeks to train the BCI to recognize his unique brain signals, repeating each of the six commands aloud as they appeared on a screen. Once the BCI’s deep-learning algorithm was trained, Evans was asked to issue the same verbal commands to control a communication board in real time, usually for about five minutes every day for three months.
“While Tim’s speech was difficult for most human listeners to understand, the BCI was able to accurately translate his brain activity into computer commands, allowing him to navigate and select items on a communication board at his own pace,” says Shiyu Luo, graduate student in biomedical engineering at The Johns Hopkins University and first author of the paper. “In addition, Tim was able to express how he was feeling or what he wanted.”
Throughout testing, researchers discovered that using signals from both motor and sensory areas of the brain produced the best results. Brain areas related to the movement of the lips, tongue and jaw had the most influence on the BCI’s performance, and stayed consistent over three months of the study, playing a crucial role in making the BCI work well and reliably, says Luo.
Crone says that, unlike many other BCI studies, this approach used electrodes that do not penetrate the brain, allowing the team to record large populations of neurons from the surface of the brain instead of individual neurons.
“What’s amazing about our study is that the accuracy didn’t change over time, it worked just as well on Day 1 as it did on Day 90,” Crone says. “Our results may be the first steps in realizing the potential for independent home use of speech BCIs by people living with severe paralysis.”
Not having to retrain the BCI algorithm means this approach potentially allows participants the freedom to use the BCI whenever and wherever they want without ongoing researcher intervention says Crone.
“In the future, hopefully this means a participant with severe paralysis can start their day by turning on the lights and catching up with the news on TV using only their brain signals,” Crone adds.
One limitation of the study was the limited vocabulary used for decoding speech. Crone says that although the six commands they adopted were both intuitive and sufficient for controlling grid-based applications, a more comprehensive vocabulary may reduce the time needed to perform a broader range of tasks.
Currently, Crone and his team are working with Evans on a series of other studies that expand the vocabulary the BCI can pick up from his brain signals. The team is also working towards translating neural signals directly into acoustic speech. They are actively recruiting patients for clinical trials investigating BCI systems for those with movement and communication impairments.
“It’s a very exciting time in the field of brain-computer interfaces,” says Crone. “For those who have lost their ability to communicate due to a variety of neurological conditions, there’s a lot of hope to preserve or regain their ability to communicate with family and friends. But there’s still a lot more work to be done to bring this to all patients who could use this technology.”
For more information on the CortiCom study, click here.
Both Crone and Luo are available for interviews. Evans comes in for testing throughout the week and can be filmed or photographed by the media.
Additional study authors include Miguel Angrick, Christopher Coogan, Daniel Candrea, Kimberly Wyse-Sookoo, Samyak Shah, Qinwan Rabbani, Alexander Weiss, William Anderson, Donna Tippett, Nicholas Maragakis, Lora Clawson and Hynek Hermansky from the Johns Hopkins University School of Medicine; Griffin Milsap, Brock Wester, Francesco Tenore and Matthew Fifer from the Johns Hopkins University Applied Physics Laboratory; and Nick Ramsey and Mariska Vansteensel from University Medical Center Utrecht Brain Center in the Netherlands.
The study was funded by the National Institutes of Health NINDS.
All authors report no conflict of interest.