Both groups’ approaches have tradeoffs. Implanted electrodes, such as the one used by the Stanford University team, record the activity of individual neurons, which tend to provide more detailed information than recordings from the surface of the brain. However, implanted electrodes are also less stable as they move around in the brain. Even movements of 1-2 millimeters cause changes in recorded activity. “It’s difficult to record from the same neuron for weeks at a time, much less for months or years at a time,” Slutsky said. Also, over time, scar tissue forms around the implanted electrodes, which can also affect the quality of the recordings.
Surface arrays, on the other hand, capture less detailed brain activity but cover a wider area. According to Slutzky, the recorded signals come from thousands of neurons, which makes them more stable than spikes from individual neurons.
At the conference, Willett said current technology is limited by the limited number of electrodes that can be safely placed in the brain at one time. “Just as a camera with more pixels produces a sharper image, using more electrodes gives us a clearer picture of what’s going on in the brain.” he said.
Lee Hochberg, a neurologist at Massachusetts General Hospital and Brown University who collaborated with a research group at Stanford University, believes that 10 years ago, people could be deciphering speech simply by recording their brain activity. says most people would never have imagined. “We want to show patients with ALS, brain stem stroke, and other neurological diseases and injuries that we can restore their ability to communicate easily, intuitively, and quickly,” Hochberg said.
Betts Peters, a speech pathologist at Oregon Health and Science University, writes that while these new BCIs are still slower than typical speech, they are faster than existing augmented and alternative communication systems. These systems require users to use their fingers or eye gaze to type or select messages. “Being able to follow the flow of a conversation can be of great benefit to many people with communication disorders, making it easier to fully participate in all aspects of life,” she told WIRED. told by email.
There are still some technical hurdles to creating implantable devices with these capabilities. First, Slutzky says that both groups’ error rates are still fairly high in everyday use. By comparison, current speech recognition systems microsoft and Google The error rate is around 5%.
Another challenge is device longevity and reliability. A practical BCI would need to record the signal continuously for years, Slutsky says, without the need for daily recalibration.
The BCI should also be wireless, without the clunky cables required by current systems, so that the patient can use it without having to be connected to a computer. Companies like Neuralink, Synchron and Paradromics are all working on wireless systems.
“The results are already amazing,” says Matt Angle, founder and CEO of Austin-based Paradromics. He is not involved in the new paper. “I think we will start to see rapid progress towards medical devices for patients.”
