Subscribe to eTOC

Brain-Computer Interfaces Move Forward at the Speed of Musk

Article In Brief

A brain-computer interface (BCI), developed by a company tied with Elon Musk, offers an innovative new technology to move the field forward. But independent experts said some of the claims about what that technology can potentially do are somewhat hyperbolic at this stage. Other efforts around BCI are also featured.

Figure

The BCI implants, developed by Neuralink, have been granted a Breakthrough Devices Designation by the FDA.

It's going to blow your mind,” said Elon Musk, the billionaire founder and CEO of Tesla, SpaceX and Neuralink. Speaking via a webcast on August 28, Musk introduced the latest developments, or at least made the latest claims, regarding Neuralink's brain-computer interface (BCI), a computer-based system that measures and analyzes brain activity and then converts signals into commands that are relayed to output devices that carry out the desired action.

The US Food and Drug Administration (FDA), he announced, had given the company a Breakthrough Devices Designation, and “good progress,” he said, was being made toward clinical trials.

As evidence of that progress, Musk presented what he called “the Little Pigs demo.” Three pigs in the webcast looked equally healthy, but one of them had been implanted with a device that was wirelessly transmitting neural spikes from the snout area of its brain as it rooted around in a pen.

“There's a lot of function that this device could do related to monitoring your health and warning you about a possible heart attack or stroke, and convenience features like playing music,” Musk said. “It's sort of like if your phone went in your brain.”

Figure

“What we are seeing is a shift to Silicon Valley-style neurotechnology companies that attract venture capital and a lot of talent quickly. The pace of iteration in fields that adopt this approach—several new electronics designs every year and regular software updates—is too fast for journal review cycles.”—DR. GRAEME MOFFAT

He predicted that, one day, “you will be able to save and replay memories...Over time we could actually give someone super vision. You could have ultraviolet or infrared, or seeing radar, and actually have superhuman vision.”

Alas, none of those capabilities has ever been demonstrated by Musk's company or, for that matter, by any scientist or company. Neurologists and neuroscientists who have been laboring in the field for over a decade seek not to endow healthy humans with superpowers, but to help people with neurodegenerative diseases, strokes or traumatic injuries to regain speech and motor functions. Despite that disconnect, however, leaders in the field told Neurology Today that they are impressed by the legitimate progress Neuralink appears to have made.

The company's implantable device, they say, is doing things that none of their own devices have yet achieved: It communicates to a nearby computer wirelessly, rather than through wires; its low-powered battery, also recharges wirelessly, lasts up to a day; it bristles with over a thousand electrodes, compared to a few hundred in traditional arrays; and rather than attaching those electrodes to inflexible shanks, the Neuralink device uses flexible “threads” so delicate they are implanted by a robotic “sewing machine.”

“If indeed they have made a device that can detect a thousand channels with good fidelity, and it can scan through this wireless technology, that is an important development,” said Karunesh Ganguly, MD, PhD, associate professor of neurology at the University of California, San Francisco. “At some point, you want to see it peer reviewed to know that what they're identifying as a neuron is actually a neuron. But they do seem to be pushing the technology ahead.”

More evidence of progress in the BCI field came with a September 7 report by Dr. Ganguly's group in Nature Biotechnology. Unlike other approaches for invasive BCIs, his “plug and play” system based on electrocorticography (ECoG) has permitted an individual with tetraplegia to maintain control of a computer cursor without daily recalibration and retraining.

Figure

“If indeed they have made a device that can detect a thousand channels with good fidelity, and it can scan through this wireless technology, that is an important development.”—DR. KARUNESH GANGULY

While he and other scientists in the BCI field expressed some skepticism about some of the claims made by Neurolink and Musk, they told Neurology Today they appreciate the interest that Neuralink is bringing to the field.

“It will be really important, as these technologies become available, hopefully over the next few years, for neurologists to become familiar with how BCI can help the patients they see in their daily practice,” said Leigh Hochberg, MD, PhD, FAAN, professor of engineering at Brown University, senior lecturer in neurology at Harvard Medical School, and director of the Center for Neurotechnology and Neurorecovery at Massachusetts General Hospital.

“As a neurologist, I am completely focused on developing and testing technology that will help patients I see who have paralyzing disorders to maintain or regain their ability to move and speak.”

Companies like Neuralink, he said, are essential to achieving that goal.

“At the end of the day, neither universities nor academic medical centers make the final marketed and supported medical devices that become available to patients outside of clinical trials,” Dr. Hochberg said. ‘That is always done by a company. The engagement of companies in this field will ultimately benefit our patients.”

Research on Pre-print Servers

Although Neuralink has not published a description of its technology in peer-reviewed scientific literature, and neither Musk nor any of the scientists working for him responded to requests for interviews, some detailed descriptions have appeared on the preprint server bioRxiv.

In March of 2019, Philip Sabes, PhD, of the University of California, San Francisco (and a “founding team leader” at Neuralink) was the lead author of a paper describing the “sewing machine” his group had developed with funding by the Defense Advanced Research Projects Agency (DARPA).

The fixed, rigid metal arrays used since the 1950s to penetrate the brain, the paper noted, disturbs the vasculature and attracts immune cells. Thinner, more flexible probes, however, would not be stiff enough to insert into the brain directly. Dr. Sabes' solution was to design a system that works like a sewing machine, with stiff needle-like injectors that implant polymer probes with the aid of a neurosurgical robot.

Each of the system's 96 polymer threads, the paper stated, holds 32 electrodes, “for a total of 3,072 electrodes. We developed miniaturized custom electronics that allow us to stream full broadband electrophysiology data simultaneously from all these electrodes. We packaged this system for long-term implantation and developed custom online spike detection software that can detect action potentials with low latency.”

Five months later, in August of 2019, Musk was listed as the first author of another preprint on bioRxiv that described Sabes' sewing machine as part of Neuralink's BCI platform. “We have built arrays of small and flexible electrode ‘threads,’” Musk's paper stated, “with as many as 3,072 electrodes per array distributed across 96 threads.”

As described by Musk in his webcast in August, implantation of the company's device will require removal of what he called “a coin-sized piece of skull” by its robot. “Then the device replaces the portion of skull that we removed.”

Unprecedented ‘Financial and Intellectual Effort’

“Technological advances already made and anticipated from the Neuralink effort are remarkable and will serve the neuroengineering community well,” Lee E. Miller, PhD, Distinguished Professor of Neuroscience at the Feinberg School of Medicine of Northwestern University, said. “This scale of private investment of financial and intellectual effort is unprecedented in our field.” He added, however, that the company needs to demonstrate that its device is doing what it claims to be doing.

“They showed these rasters of brain activity on the webcast, with cool bloopy sounds,” he said. “For peer review, I would insist on seeing the actual signals they recorded. Although there is no reason to believe it to be the case, they could be recording movement artifacts.”

Figure

“It will be really important, as these technologies become available, hopefully over the next few years, for neurologists to become familiar with how BCI can help the patients they see in their daily practice.”—DR. LEIGH HOCHBERG

“My hunch is that it will not be as expensive as deep brain stimulation for Parkinson's disease, which costs tens of thousands of dollars. But it's going to be more expensive than, say, LASEK surgery.”

—DR. LEE MILLER

“When I was first approached about working on brain-machine interfaces, I thought, ‘This is crazy.’ But we went farther than we could have reasonably hoped to do. That's why I'm not completely closed to the wildest claims Musk makes. It's not insane that maybe one day we can replay memories.”

—DR. SLIMAN BENSMAIA

Dr. Miller also took issue with Musk's prediction that a BCI device could eventually cost as little as a few thousand dollars. “That's not going to happen any time soon,” Dr. Miller said. “My hunch is that it will not be as expensive as deep brain stimulation for Parkinson's disease, which costs tens of thousands of dollars. But it's going to be more expensive than, say, LASEK surgery.”

He also disagreed with Musk's over-the-top claims about recording and replaying memories.

“There's a lot of science in the realm of memory,” he said, “including Wilder Penfield's decades-old work that appeared to show existing memories being triggered by electrical stimulation of the brain's temporal lobes during neurosurgery.”

“That's a far cry, however, from suggesting it would ever be possible to record from a particular memory and play it back,” Dr. Miller continued. “While short-term, working memory very likely is based on reverberating neural activity and amenable to the intriguing hippocampus memory prosthesis that Ted Berger has been working on, long-term memories almost certainly require protein synthesis and structural changes to neurons that couldn't even be recorded, let alone ‘played back.’ That's pure science fiction, and to suggest otherwise sets up all sorts of false expectations.”

Sliman Bensmaia, PhD, the James and Karen Frank Family Professor of Organismal Biology and Anatomy at the University of Chicago, runs a laboratory there devoted to research in somatosensory neuroscience and prosthetics. Earlier this year, before COVID-19 restrictions began, he visited the Neuralink offices and gave a talk.

“There's a team of really great people working there, and the device they have come up with is really remarkable,” Dr. Bensmaia said. “The device that I work with, that almost everyone works with, the only device that has been used in humans so far, is the Utah array, made of metal microelectrodes. It's like a mini bed of nails that you press into the brain. Of course the brain doesn't like that, and the electrodes don't last. So the fact that Neuralink has these thin, flexible fibers should cause much less damage. And they have a lot more electrical contacts. The question is how robust and stable it will be. Will it last for decades? But it's pretty cool. It's way further along now than it was just six months ago when I visited them.”

Despite all that, Dr. Bensmaia added, “Then there is Musk and the way he talks about it. Some of the stuff he says is outrageous. It might be possible to achieve some of the things he's talking about one day, but it won't happen for a very long time.”

Even so, he said, the progress made in the field in recent years is already beyond anything he thought possible in such a short time.

“I participated in DARPA's Revolutionizing Prosthetics program,” he said. “When I was first approached about working on brain-machine interfaces, I thought, ‘This is crazy.’ But we went farther than we could have reasonably hoped to do. That's why I'm not completely closed to the wildest claims Musk makes. It's not insane that maybe one day we can replay memories.”

In fact, a 2018 paper published in the Journal of Neural Engineering described a study involving epilepsy patients with surgically implanted electrodes near the hippocampus whose electrical spikes were recorded and analyzed while they performed a memorization task. When scientists stimulated the CA1 region by playing back the sequence of neural firing made when the subjects correctly remembered a preliminary set of memorizing tasks, their performance on subsequent memorization tasks improved by 35 percent.)

Dr. Hochberg leads the BrainGate consortium, which includes researchers from Massachusetts General Hospital, Brown University, the Providence VA Medical Center, Stanford University and Case Western Reserve.

“Over the past few years, in our published research, the participants in our trial who had very little or no movement of their arm or hand have been able to control an unmodified tablet computer for email, for texting, for controlling their music players,” Dr. Hochberg said.

“I used to say it would take decades before a BCI is available to people outside of research trials that could offer a true clinical benefit. I now think we are just a few years away. Right now these systems often require the oversight or engagement of a trained technician to start the system and calibrate it at the beginning of each day. We need it to work 24 hours a day, seven days a week, in the absence of any technical oversight. On all those merits, we are on track to achieve that goal with a flexible, powerful and reliable system.”

Because Mass General has a clinical research support agreement with Neuralink, Dr. Hochberg said he should not speak specifically about the company. But, he said, “I'm excited by the entrants of multiple companies to the BCI field. The engagement of companies will ultimately benefit our patients who have neurological disease or injury.”

As for those neurologists who remain leery of a field in which companies like Neuralink are publishing accounts of apparent gains in preprints posted without peer review, a University of Toronto fellow said, essentially, this is a sign of things to come in this burgeoning world of technology.

“What we are seeing is a shift to Silicon Valley-style neurotechnology companies that attract venture capital and a lot of talent quickly,” said Graeme Moffat, PhD, a former managing editor of Frontiers in Neuroscience who now also runs a company developing non-invasive brain imaging devices.

“The pace of iteration in fields that adopt this approach—several new electronics designs every year and regular software updates—is too fast for journal review cycles. We'll see papers on the long term effects of new BCIs on the brain, but the peer-reviewed scholarly literature is just unsuited to reflecting the rapid innovation in devices like those that Neuralink is building.”

Disclosures

Dr. Ganguly has received a one-time consulting fee from Lightside Medical, a medical incubator company. Dr. Stavisky is a scientific advisor to Vorso Corporation and Broad Mind Inc. and has equity in both companies. Dr. Angle owns stock and is employed by Paradromics. Drs. Miller, Bensmaia, and Moffat had no disclosures.

Link Up for More Information

• Musk E. Neuralink update. Accessed August 28, 2020 https://www.youtube.com/watch?v=vxehbGLoar8. http://bit.ly/NLwebcast
    • Hanson TL, Diaz-Botia CA, Kharazia V, et al The “sewing machine” for minimally invasive neural recording. Pre-print posted March 14, 2019 https://www.biorxiv.org/content/10.1101/578542v1. http://bit.ly/NLsewing
      • Silversmith DB, Abiri R, Hardy NF, et al. Plug-and-play control of a brain-computer interface through neural map stabilization https://www.nature.com/articles/s41587-020-0662-5. Nat Biotechnol 2020; Epub 2020 Sept 7.
        • Musk E Neuralink. An integrated brain-machine interface platform with thousands of channels https://www.biorxiv.org/content/10.1101/703801v4. Posted August 2, 2019. http://bit.ly/NL-BCI
          • Wilson GH, Stavisky SD, Willett FR, et al. Decoding spoken English phonemes from intracortical electrode arrays in dorsal precentral gyrus https://www.biorxiv.org/content/10.1101/2020.06.30.180935v1.article-metrics. Pre-print posted July 1, 2020. http://bit.ly/Wilson-phonemes
            • Gallego JA, Perich MG, Chowdhury RH, et al. Long-term stability of cortical population dynamics underlying consistent behavior https://www.nature.com/articles/s41593-019-0555-4. Nat Neurosci 2020;23(2):260–270.
            • Hampson RE, Song D, Robinson BS, et al. Developing a hippocampal neural prosthetic to facilitate human memory encoding and recall https://iopscience.iop.org/article/10.1088/1741-2552/aaaed7. J Neural Eng 2018;15(3):036014.