Scientists have developed new technology to help nonverbal patients communicate with others.
It is safe to say the future is here. On November 6th, a new technology that has the capability to translate a person’s brain signals into what they’re trying to say was featured in the journal Nature Communications. One of the lead researchers of the project, Dr. Gregory Cogan is a professor of neurology and neurosurgery at Duke. He explained that patients who suffer from motor disorders like ALS (amyotrophic lateral sclerosis) may have an impaired ability to speak or make them noncommunicative. The new speech prosthetic could help people who are unable to talk due to neurological disorders regain the ability to communicate through a brain-computer interface.
Current tools available today which allow noncommunicative people to speak are incredibly slow. To illustrate, the average person typically speaks about 150 words per minute. Meanwhile, the top speech decoding rate technology allows only roughly 78 words per minute. This delay is largely because it is difficult to fuse many brain activity sensors onto a very thin piece of material laid on top of the brain’s surface. The fewer the sensors, the less legible information there is to decode and communicate.
That is why Dr. Cogan, along with a group of neuroscientists, neurosurgeons, and engineers all teamed up to improve the lag limitations of current decoding technology. The team developed a speech prosthetic using Dr. Jonathan Viventi’s (Duke Institute for Brain Sciences faculty) biomedical engineering lab which specializes in making ultra-thin, flexible brain sensors. The team was able to pack an astounding 256 microscopic brain sensors onto a piece of medical-grade plastic the size of a post stamp. The technology can make predictions about intended speech based on its reading of different activity patterns that take place when coordinating speech such as motor cortex activity required to coordinate muscles in the lips, tongue, jaw, and larynx.
Once the speech prosthetic was created, it was tested on four noncommunicative patients who needed brain surgery for different reasons such as tumor removal. Dr. Cogan and his team had to work as fast as possible in order to keep the operating procedure relatively normal and timely. The team had to implant the device, test it, and remove all within a mere 15 minutes.
Testing the accuracy involved simple listen-and-repeat exercises. The patient would hear words and sounds and then try to repeat what was heard out loud. Because of the need for a quick test turn-around, the spoken data was only about 90 seconds long for each patient. After the procedure and test, a machine received the data and calculated how accurately the learning algorithm could predict what sound was being made based on the brain activity.
The decoder had 84% accuracy for certain sounds such as those with three letters that started with “g.” The technology’s accuracy was lower when there were similar sounds like “p” and “b.” The inconsistencies made the decoder have a total accuracy of about 40%. Additionally, the decoding is still significantly slower than the average person’s normal talking speed. Nonetheless, the team plans to advance the technology by making it faster, more accurate, and inventing a wireless version. Luckily, the cordless option and other improvements should be possible given the recent $2.4 million grant from the National Institutes of Health.
Sources:
Brain implant may enable communication from thoughts alone
Brain Implant Translates Thoughts Into Words Faster Than Ever
Join the conversation!