This High Schooler Invented a Low-Cost, Mind-Controlled Prosthetic Arm

Ten years ago, when Benjamin Choi was in third grade, he watched a “60 Minutes” documentary about a mind-controlled prosthesis. Researchers implanted tiny sensors into the motor cortex of the brain of a patient who moved a robotic arm using only her thoughts. Choi was fascinated by the concept, likening it to something out of a Star Wars movie.

“I was really, really amazed at the time because this technology was so impressive,” he says. “But I was also alarmed that they require this really risky open brain surgery. And they're so inaccessible, costing in the hundreds of thousands of dollars.”

Years later, when the pandemic hit in 2020, Choi—a tenth grader living in Virginia—suddenly found himself with ample free time. The lab in which he’d planned to spend his summer researching aluminum fuels had shut down. But the documentary he had seen years earlier stuck with him, and he decided to use his spare time to build a less-invasive prosthetic arm himself.

In his makeshift laboratory on the ping-pong table in his basement (where he sometimes worked 16 hours a day!), Choi independently designed the first version of his robotic arm using his sister’s $75 3-D printer and some fishing line. The printer couldn’t build pieces over 4.7 inches in length, so Choi printed the arm in tiny pieces and bolted and rubber banded it together. In total, it took about 30 hours to print. This version worked using brain wave data and head gestures, and Choi posted instructions online for anyone to build their own.

He had some previous experience building robots and coding from participating in competitive robotics at the elementary, middle school and high school levels, even going to the world championships several times. Starting in ninth grade, he taught himself the computer programming languages Python and C++ by watching videos on Stack Overflow, a website for programmers.

After more than seventy-five design iterations, Choi’s non-invasive, mind-controlled robotic arm is now made from engineering-grade materials able to withstand weights up to about four tons. It operates using an algorithm driven by artificial intelligence (A.I.) that interprets a user’s brain waves. And it only costs around $300 to manufacture—a huge savings compared to what’s currently on the market. A more basic, body-powered upper limb prosthesis costs about $7,000. As of 2015, the very advanced, full-arm Modular Prosthetic Limb, which has 26 joints, hundreds of sensors and can curl up to 45 pounds, cost about $500,000. This prosthesis, paired with a surgery to reroute nerves that once controlled the arm, allows patients to command the limb with their thoughts and even feel texture through it.

The invention earned Choi, now a 17-year-old senior at the Potomac School in Virginia, a spot in the top 40 finalists of this year’s Regeneron Science Talent Search, the country’s oldest and most prestigious science and math competition for high school seniors. This year’s first place winner was Christine Ye from Sammamish, Washington, who developed a way to analyze the gravitational waves emitted from collisions between neutron stars.

“It means a lot to me to see that my work is recognized like this,” Choi says. “I'm definitely very grateful to be a finalist.”

An estimated 2 million people are living with the loss of a limb in the United States, and about 185,000 amputations occur every year. The World Health Organization states only one in ten people who need assistive products, including prosthesis and orthoses, have access to them, citing “high cost” and a “lack of awareness, availability, trained personnel, policy and financing.”

Choi’s arm uses electroencephalography, or EEG, to avoid the invasive techniques of other prostheses. EEG devices record the brain’s electrical activity using sensors placed on the head.

They are often used in medicine to diagnose epilepsy or other brain disorders.

His system uses two electrodes: a baseline sensor that clips onto the earlobe and another on the forehead that collects EEG data. The forehead electrode picks up brain wave information, which is sent to a microchip in the prosthetic arm via Bluetooth. An A.I. model that Choi created, also embedded in the chip, deciphers the data and converts it into a prediction of what the brain is thinking. The arm also moves using head gestures and stops with intentional blinks.

Mind-Controlled Prosthetic Arm - Regeneron STS 2022 Demo Video!

Watch on

Six months after he started building the arm, he posted a video on YouTube demonstrating its dexterity. The prosthesis caught the attention of Joseph Dunn, an upper-limb amputee from Pennsylvania. Choi began consulting with him remotely about the design.

“Working with Mr. Dunn made this project really impactful and really inspiring and motivating,” Choi says. “Maybe this sounds a little cliche, but you can really help people, I think, through engineering, through technology.”

The young inventor went on to win funding from the Massachusetts Institute of Technology in 2021 to continue his research and work with experts at the university. For about six months, he experimented with cloud computing to make the arm internet compatible.

“These A.I. models can get so big,” he says. “I was thinking I would store them in the cloud, and then have my arm communicate via Wi-Fi.”

But that didn’t work for two reasons. First, it took too much time for the arm to respond to a user’s thoughts.

“That's just not ideal, because especially for prosthetics, you want them to work in real time very quickly,” he says.

Second, a user would constantly need to be connected to Wi-Fi, which isn’t practical, Choi explains. Instead, he compressed his A.I. model—which contains several sub-models—and stored it in a dual-core microchip inside the arm.

To create his A.I. model, he worked independently with six adult volunteers for about two hours each, collecting their brain wave data in his school and home. While collecting the data via an electrode on the forehead, he asked each participant to focus on clenching and unclenching their hand.

He trained the A.I. to distinguish between the brain signals, and the A.I. model continuously learns from a user’s brain waves.

“The more you use it, the more it figures out specifically how you think, what your brain wave patterns are, until the accuracy really increases significantly for you over time,” Choi explains.

In total, the algorithm has over 23,000 lines of code, with 978 pages of math and seven completely new sub-algorithms. Choi’s algorithm performs with a mean accuracy of 95 percent. He says the previous gold standard for a similar artificial neural network was 73.8 percent.

Brock Wester, a biomedical engineer at Johns Hopkins University with a background in neuroprosthetics, says Choi’s technology is impressive, especially when considering the prosthesis in the “60 Minutes” documentary had a large team of researchers working on it, whereas Choi designed the whole arm himself.

“That he was able to build this limb, to develop the controls for it and the algorithms to decode his neural signals in real time to send those control signals, I think that's just remarkable,” says Wester. “He should continue to do research in this space, because I think he can make a lot of important contributions.”

Wester notes Choi still has some more engineering to do, especially when thinking about how the limb connects to a user’s body. At this point, the arm is connected to a fixed post on a platform. Choi says he’ll eventually design a socket, but that process will require custom-fitting for the user.

Last summer, Choi was selected as a Simons Fellow at Stony Brook University, where he worked remotely with Ji Liu, a professor in the department of electrical and computer engineering, on the machine learning algorithm of his A.I. Though Liu says he gave Choi some high-level advice about the algorithm, including how to construct more comprehensive training data sets for it, Choi worked on the finer details on his own.

The novelty behind Choi’s project is that he “applied state-of-the-art machine learning techniques to his robotic arm system,” Liu says, adding that the low-cost arm’s performance is comparable to more expensive and advanced machinery.

“He is not only very smart, but also working very hard and independent,” Liu says of Choi. “That’s also pretty outstanding compared with graduate students.”

Choi has also won awards in the Regeneron International Science and Engineering Fair, the Microsoft Imagine Cup, and the National At-Home STEM Competition. He received a manufacturing grant in October 2020 from PolySpectra, Inc., a company that produces durable 3-D printed materials, to produce his arm.

Outside of engineering, Choi is a nationally-ranked squash player, student body president at his school, a published short story author, a violin soloist with top finishes in several competitions, and the founder of a team of Potomac students that competed on the NBC quiz show “It’s Academic.”

Choi plans to study engineering in college, and wants to keep improving his prosthetic arm. He has his sights set on conducting a clinical study with patients with upper-limb losses. He’s already acquired two provisional patents for his invention: one for the neuroprosthesis, and one for the brain wave interpretation algorithm.

Choi says his algorithm could have uses beyond prosthetics, including controlling assistive devices like wheelchairs and helping patients with ALS communicate.

“Brain wave interpretation is a really big emerging field,” he says. “My algorithm is the best of all the algorithms reported in literature by a pretty significant margin. I think it could have big applications going forward.”

Margaret Osborne | | READ MORE

Margaret Osborne is a freelance journalist based in the southwestern U.S. Her work has appeared in the Sag Harbor Express and has aired on WSHU Public Radio.