Even in a city where high-stakes deals are made daily, few would expect that behind an unassuming Manhattan facade lies a team of neuroscientists developing a device that could change the relationship between machines and humans.
But change often hides in plain sight, undetectable until it isn’t.
Precision Neuroscience is developing a brain-computer interface, a system that creates a direct link between the brain’s electrical activity and a computer. This type of technology has been in development for decades, but thanks to AI, the latest generation promises to reveal a wealth of new information about the brain.
I met Precision’s president and chief product officer, Craig Mermel, in March at an AI conference. The company stood out as one of a few hardware companies among a sea of startups building productivity tools and coding platforms.
The AI boom, Mermel told me, had turbocharged modern medicine’s capacity to collect data about the brain. At a moment when many of the conference attendees talked about the technology’s implications in broad terms, Mermel made them seem real — and imminent.
Precision was founded in 2021 by a group that included some Neuralink alums, including Ben Rapoport, a practicing neurosurgeon and electrical engineer who now serves as its chief science officer.
The company’s initial goal is to “restore independence” to severely paralyzed people in a less invasive way than its competitors, like Neuralink. That means helping those with paralysis communicate with others, use computers, or even hold a desk job by channeling their neural activity into commands for external technologies.
“The set of conditions that we’re treating all have in common is paralysis. So it’s people who basically have the ability to think like we do, but not the ability to move, especially move their hands,” Rapoport told me during a visit to the company’s office.
How it works
Rapoport handed me a slip of yellow film, which the company calls the “Layer 7 Cortical Interface.” The name reflects its ambitious goal: to create a seventh layer that sits atop the six cellular layers of the human cortex, the brain’s outermost region.
The Layer 7 Cortical Interface is about one-fifth the thickness of a human eyelash. One end is embedded with an array of 1,024 electrodes that can record and stimulate brain activity. These electrodes have wires that run lengthwise, linking the film to custom electronics that process neural data and convert those signals into computer commands so patients can interact with the real world using thoughts alone.
The device is designed to rest on the brain’s motor cortex — a small region located behind the frontal lobe that translates thoughts into actions — and conform to its surface, never directly damaging the tissue.
Once in place, it generates a detailed view of the brain’s activity. Or, as Precision puts it on its website, “the world’s highest resolution picture of human thought.”
The competition
Rapoport was part of the founding team at Neuralink, Elon Musk’s high-profile neurotechnology venture. It launched in 2016 and has raised about $1.3 billion, according to PitchBook.
He told the newsletter Neurotech Futures last year that he left Neuralink in 2018 because brain-computer interface technology needed to be extremely safe and scalable for it to become a “clinical reality.”
“Some changes were needed in the way the technology was being implemented,” he said.
Neuralink’s N1 implant is a coin-sized, battery-powered device that sits on the skull and connects to ultra-thin threads — embedded with electrodes — that are woven into the brain’s cortex.
In an April 2024 update on its website about the status of its clinical trials, Neuralink detailed how the implant is inserted into the brain.
“At a high level, the surgery involved a neurosurgeon exposing the target region of the cortex (e.g., scalp incision, craniectomy, durectomy), the R1 Robot performing the insertions of threads of the N1 Implant, and the neurosurgeon mounting the body of the N1 Implant in the craniectomy and closing the scalp.”
In a follow-up email to Business Insider, Mermel said Neuralink’s system is based on “penetrating micro-electrodes, which cause damage when they’re inserted into the brain.”
Precision, he said, has shown that it is “possible to extract information-rich data from the brain” without damaging it.
Neuralink did not immediately respond to a request for comment from Business Insider.
In July, Neuralink announced it had successfully completed implant surgeries on its eighth and ninth human patients. The company is expecting to implant 20,000 chips a year by 2031 to generate at least $1 billion in revenue, Bloomberg reported.
Invasive forms of brain-computer interfaces typically involve implanting electrodes directly into the brain by way of a craniotomy, which involves removing a section of the skull to temporarily expose the brain, or a craniectomy, in which the removed bone is not replaced.
Blackrock’s Neurotech, which launched in 2008, makes a modified version of the Utah Array — a micro-electrode array that penetrates up to 1.5 millimeters into the brain. It’s been used for “studying neural circuits, investigating the mechanisms of brain function, and developing neural prosthetic devices,” according to Neurotech’s website.
Synchron, another company in the BCI race, makes a “Stentrode,” a thin, flexible electrode array that doesn’t require open-skull surgery. Instead, it’s inserted into the brain through the jugular vein. Its aim is to “restore control of a touchscreen in people with limited hand mobility using only their thoughts,” according to its website.
What’s different about Precision
Rapoport’s thesis is that placing the device on the brain’s surface itself is less invasive than its competitors and enough to gather valuable intel about how the brain operates.
“When we think about the ways that we want the brain to interface with the digital world, most of what we think about is the conscious thought that takes place in the brain,” he said. “Essentially all of that takes place at the surface” in the “outermost few millimeters.”
The company’s Layer 7 Cortical Interface is designed to be easily replaced or moved. It’s also modular, so multiple arrays can be combined to cover more regions of the brain.
The device is just one component, however. The method of insertion is another.
Precision is also developing a minimally invasive “cranial microslit” implant procedure to bypass the need for a full craniotomy, Mermel said by email. This involves making a small incision, less than 1 millimeter into the skull, through which the Layer 7 Cortical Interface is inserted and placed on the brain’s surface, he said.
Clinical studies
In 2021, the FDA gave Synchron a green light to start clinical trials. In 2023, Neuralink received approval to start trials. Precision, at an earlier stage in its road to commercialization, announced its first clinical studies in 2023.
Since then, the company has conducted clinical studies on 47 volunteer patients — all of whom were already undergoing brain surgery for other reasons. The devices were placed on their brains during surgery to read, record, and map activity on the brain’s surface.
In these clinical studies — limited to 15 to 30 minutes — one or two electrode arrays were typically placed on a patient’s brain while they engaged in activities such as speaking, playing rock-paper-scissors, or operating a joystick, so AI algorithms could learn the typical neural patterns associated with that activity.
In April, Precision said it received FDA clearance to use its electrode array in the “recording, monitoring, and stimulation of electrical activity on the surface of the brain.” The clearance authorizes the array for commercial use for up to 30 days. Studies are already underway, but the company declined to share the number of patients who are participating in these longer clinical studies.
The value of the data
Brain-computer interfaces serve as translators of sorts, converting the brain’s electrical language into the vernacular of machines — and eventually, into real-world action.
Precision says its system captures about 1 to 2 billion data points per minute from each patient. It analyzes the data in real time and leverages AI algorithms to translate the raw data and electrical signals into computer code.
The goal is to collect data from a diverse sample of patients, Rapoport said.
“In the entire field of neuroscience, we have never had such a diverse set of high-quality, high-resolution, long-term recordings from dozens of patients’ brains until we started doing this, and this includes Neuralink,” he said.
Between all of its studies, it has gathered enough data to begin building what he described as a “neural foundation model.”
The main focus of this machine learning model will be to decode speech and motor intention from the brain to ultimately help patients control computers and smartphones, Mermel said. However, he added, the company is exploring future use cases for its technology, including assisting surgeons during neurosurgery, treating conditions such as depression, and aiding in stroke recovery.
“When you have something that is safe and effective for a niche group of people, it begs the question: Are there ways to scale it beyond that original use case?” Rapoport said.
Read the full article here
