Chang’s group first worked to characterize the part of the brain that produces phonemes, and therefore language — an ill-defined region called the dorsal laryngeal cortex. The researchers then applied these findings to develop a speech decoding system that displays the user’s intended speech as text on a screen. In 2021, they reported that this device enabled a person unable to speak after suffering a brainstem stroke to communicate using a preselected vocabulary of 50 words and a rate of 15 words per minute. “The most important thing we learned,” says Chang, “is that it’s no longer theory, it’s actually possible to decipher whole words.”
Unlike other groups, Chang did not record the activity of individual neurons. Instead, he uses electrodes on the surface of the cortex that record the average activity of neuronal populations. The signals are not quite as precise as those from electrodes implanted in the cortex, but the approach is less invasive.
The most profound loss of communication skills occurs in locked-in patients, people who are conscious but unable to speak or move. In March 2022, a team led by neuroscientist Ujwal Chaudhary from the University of Tübingen was able to communicate with a man suffering from amyotrophic lateral sclerosis (ALS). The man had previously relied on eye movement to communicate until the disease gradually robbed him of the ability to move his eyes as well.
The team obtained family consent to implant a brain-computer interface in him. It then asked him to imagine movements in order to use his brain activity to select letters on a screen. When that didn’t work, the researchers tried playing a tone that mimicked the man’s brain activity—a higher tone for more activity, a lower tone for less—and taught him to modulate his neural activity so that the pitch increased for “yes” and decreased for “no”. In this way he could pick out a letter every minute or so.
From the lab to the market
According to Amy Orsborn, who researches BCIs in nonhuman primates at the University of Washington in Seattle, such case studies suggest the field is rapidly evolving. But while initial successes have attracted media and investor attention, BCIs are a long way from improving the daily lives of people who have lost the ability to move or speak. Currently, individual patients use brain-computer interfaces in short, intense sessions; almost all have to be physically wired to a series of computers and overseen by a team of scientists who are constantly working to improve and recalibrate the decoders and associated software. “What I want,” says Hochberg, “is a device that is available, that can be prescribed, that comes off the shelf and that can be deployed quickly.” Furthermore, ideally, such devices would last a lifetime.
Many leading experts in the field are now collaborating with companies to develop just such devices. Chaudhary, on the other hand, co-founded the non-profit company ALS Voice in Tübingen to develop neurotechnology for locked-in patients.
Blackrock Neurotech devices have been a mainstay of clinical research for 18 years. The company wants to bring a BCI system to market maturity in 2023, explains Managing Director Florian Solzbacher. The company cleared a major hurdle in November 2021 when the US Food and Drug Administration (FDA), which is responsible for approving medical devices, approved the company’s products for an accelerated review process to facilitate their commercial development.
A possible first product could consist of four implanted arrays connected by wires to a small device. However, Blackrock Neurotech is also working on completely wireless brain-computer interfaces; Neuralink and Paradromics have set themselves the goal of developing similar devices.
The two companies also aim to increase the signal bandwidth, which should improve the devices’ performance by increasing the number of neurons whose activity is recorded. The Paradromics interface, which is currently being tested on sheep, has 1600 channels spread over four modules.
Neuralink’s system uses very fine, flexible electrodes called threads that are designed to both conform to the shape of the brain and reduce immune responses, says Shenoy, who works as a consultant for the company. The aim is to make the device more durable and the recordings more stable. Neuralink hasn’t published peer-reviewed papers yet, but a 2021 blog post reported the successful implantation of the threads in a monkey’s brain, picking up signals at 1024 sites.
Apart from Blackrock Neurotech, only one other company has so far managed to develop a brain-computer interface that has already had long-term use in humans. The New York City-based company Synchron has developed a “stentrode” – a set of 16 electrodes arranged around a stent for blood vessels. The device can be implanted in one day on an outpatient basis; it is pushed through the jugular vein in the neck into a vein above the motor cortex. The technology was first used in a person with ALS in August 2019 and was fast tracked by the FDA a year later.
Similar to the electrodes Chang uses, the Stentrode does not have the high resolution of other implants and therefore cannot be used to control complex prostheses. But it allows people who can’t move or speak to control a cursor on a tablet to write text, surf the web, and control other devices.
According to Thomas Oxley, Synchron’s co-founder, the company is in the process of submitting a proof of concept in which four people will use the wireless implant at home in their everyday lives. The next step is now to test the device in a large-scale study and to investigate whether it noticeably improves the quality of life of the users.
future challenges
Most researchers working on brain-computer interfaces are realistic about the challenges that lie ahead. “BCIs are more complicated than any other neurological device ever built,” says Shenoy. “It will probably take a few more years of hard work before the technology matures.”
Orsborn emphasizes that commercial devices must function for months or years without expert supervision – and serve every user equally well. She reckons advances in machine learning will solve the first problem by allowing users to recalibrate their device themselves. It is more difficult to achieve consistently good performance across all users. “Human-to-human variability is the great unknown that we don’t know how problematic it can become,” says Orsborn. In non-human primates, even small deviations in the positioning of the electrodes can affect which neural circuits are sampled. In humans, there is also the fact that not everyone learns and thinks in the same way – and that the underlying disease that the patients bring with them could also have an influence on how the affected person’s brain works.
»BCIs are more complicated than any other neurological device ever built«(Krishna Shenoy, neuroscientist)
There is also agreement that brain-computer interfaces will raise ethical questions that need to be answered – from privacy to personal autonomy. As ethicists emphasize, users must always have full control over the output of the device. And while current technologies cannot decipher people’s thoughts, developers of brain-computer interfaces will still collect a wealth of data on users’ brain health and overall communication behavior. Also, BCIs represent a new type of cybersecurity risk.
The users of brain implants also have to worry that their devices will not be supported forever or that the manufacturers will suddenly close down. There are already such cases today in which the patients feel left in the lurch afterwards.
However, Degray hopes that brain-computer interfaces will soon reach more people. What he wants most is technology that would allow him to scratch his eyebrows again despite being paralyzed. “Everyone looks at me in a wheelchair and they’re always like, ‘Oh, the poor guy, he can’t play golf anymore.’ That’s really bad. But the real scare comes in the middle of the night when a spider walks across your face. That’s really bad.”