Heroes in Engineering: Prosthetics and Neural Implants

There is a moment in nearly every conversation about prosthetics and neural implants when the gap between what people imagine and what is actually possible becomes impossible to ignore. Popular culture and sci-fi movies have spent decades priming us for a future of seamless bionic limbs, of paralyzed patients rising from wheelchairs and reaching for a glass of water with a thought. The reality, as the scientists working in this field will tell you, is considerably more complicated and interesting.

According to a 2024 prevalence study commissioned by the Amputee Coalition, approximately 2.3 million Americans are currently living with limb loss, a figure researchers project will grow substantially in the coming decades as rates of diabetes and vascular disease continue to climb.

An additional 5.4 million Americans live with some form of paralysis, according to a population-based study commissioned by the Christopher and Dana Reeve Foundation, the result of spinal cord injuries, strokes, and conditions like ALS that sever the connection between a functioning brain and a body that can no longer respond to it. For decades, the best medicine could offer these patients was mechanical prostheses that are functional, sometimes elegant, but entirely disconnected from the nervous system. The limb moved when the patient moved it. The brain was not part of the conversation.

Lee Miller, PhD, a distinguished professor of neuroscience at Northwestern University’s Feinberg School of Medicine, has been working at that intersection for the better part of three decades. A trained physicist, he then moved into biomedical engineering, and later pursued a PhD in neuroscience, drawn deeper and deeper into a single animating question: how does the brain encode the intention to move, and what can that encoding tell us about building systems that restore movement to people who have lost it? “I am a physicist who’s gotten lost in a biological problem,” he says. It is, he will admit, a pretty good problem to get lost in.

What Dr. Miller brings to this work is a perspective shaped less by the biology of individual neurons than by the information those neurons carry. The distinction matters. Understanding the biochemistry of how a neuron fires is one kind of problem. Understanding what the pattern of firing across thousands of neurons actually means, what movement it encodes, what intention it represents, and how signals in one part of the brain relate to signals in another, is a different kind of problem entirely.

It is, as he puts it, a problem of math and physics as much as medicine. And it is a problem that, after decades of incremental progress, is beginning to feel solvable, though perhaps not in the way the movies suggested.

Ask Dr. Miller how close we are to prosthetics moving like original limbs and his answer is unambiguous. “We are light-years away,” he says. “What we can do with our biological bodies is phenomenal. The robotics is getting there, the computation is getting there. The interface to the brain is the huge problem. We’re scratching that surface.”

Keep reading to learn about the history of brain-computer interfaces, the hardware barriers that continue to slow progress, where AI is beginning to change the equation, and what it truly takes to work at the edge of neuroscience and engineering.

Featured Online Engineering Programs
ECPI University	Electronics Engineering Technology (Bachelor's)	Visit Site
ECPI University	Mechanical Engineering Technology (Associate's)	Visit Site
ECPI University	Mechanical Engineering Technology (Bachelor's)	Visit Site
American Public University	Engineering: Electrical Engineering (BEng)	Visit Site
Southern New Hampshire University	Online MBA - Engineering Management	Visit Site
University of Washington I-School	Online MS - Information Management (MSIM)	Visit Site
sponsored × When you click on a sponsoring school or program advertised on our site, or fill out a form to request information from a sponsoring school, we may earn a commission. View our advertising disclosure for more details.

Meet the Expert: Lee Miller, PhD

Dr. Lee Miller is a distinguished professor of neuroscience at Northwestern University’s Feinberg School of Medicine, where he holds appointments in the Departments of Neuroscience, Physical Medicine and Rehabilitation, and Biomedical Engineering.

A former president of the Society for the Neural Control of Movement and a fellow of the American Institute for Medical and Biological Engineering, Dr. Miller has spent his career studying how the brain encodes the intention to move and how that knowledge can be applied to the development of prosthetic limbs controlled directly from neural signals. His lab at Northwestern sits at the intersection of physics, neuroscience, and biomedical engineering, where the central question is not just how the brain works, but what the brain is saying.

A Brief History of Brain-Computer Interfaces

The idea that the brain’s electrical signals could be captured and used to control external devices is older than most people realize. As far back as the 1970s, researchers were beginning to explore whether the firing patterns of neurons could carry enough information to drive a machine. But it wasn’t until the 1990s that the field began to take shape in earnest, with pioneering work from researchers like Miguel Nicolelis and John Chapin demonstrating that neural signals recorded from rats could be used in real time to control a robotic arm. The implications were immediately clear, even if the path forward was not.

The early 2000s brought the first human implants under the BrainGate research program, which demonstrated that people with paralysis could use implanted electrodes to control a computer cursor using only their thoughts. These early results were remarkable for their time, but they also revealed how steep the climb ahead would be. The devices were crude by today’s standards, the signals were noisy, and the gap between what a paralyzed patient could do with a brain-controlled interface and what they could once do with their own hands was vast.

What has changed in the decades since is not the fundamental approach so much as the sophistication of every component involved. The electrodes, the decoding algorithms, the computing power, and, most recently, the application of artificial intelligence to interpreting neural signals have all moved forward rapidly. The hardware, as Dr. Miller will explain, remains the most stubborn obstacle. But the field has traveled an enormous distance from those first experiments, and the pace of progress is accelerating.

The Hardware Problem: Why the Biggest Barrier Is Getting Into the Brain

Ask Dr. Miller what the single biggest technical barrier is in this field, and his answer is immediate and perhaps surprising. It is not the algorithms. It is not the computing power. It is not even our incomplete understanding of how the brain encodes movement. “The biggest problem,” he says, “is sticking these wires in people’s brains.”

The Utah Array, developed at the University of Utah roughly four decades ago, is a four-by-four millimeter grid of 100 electrodes, each one and a half millimeters long. Researchers lower it directly into the surface of the brain’s cortex, where it sits among the neurons it is meant to record from. It is, as Dr. Miller puts it plainly, the workhorse of the field. It is also, by his own description, a bed of nails. “It’s these silicon spikes,” he explains. “They’re stiff, a million times stiffer than the brain. There are 100 electrodes, so we can record something like 100 neurons. That’s out of a million neurons that are going down to move my arm normally.”

The yield is extraordinarily low, and the limitations compound from there. Because the array protrudes through the skin, patients live with an open, persistent wound at the implant site. The signals the device captures degrade over time. The longest-running implant in Dr. Miller’s collaborative work, at the University of Pittsburgh, recently passed the ten-year mark, a milestone that required rewriting the original research protocol because no one had anticipated the device lasting that long. The results, however, tell a more complicated story. “The signals from that person are not very good anymore,” he says. “They’re detectable, they’re usable, but not nearly as good as an implant that is a year or two old.”

In his own animal research, the bar was lower still. “When I put the same kind of electrodes in my monkeys, we were really disappointed if we couldn’t get six months from it. We were pretty happy if we got two years,” he says. For a technology that is meant to restore function to people with paralysis or limb loss, that kind of timeline presents a serious problem. Someone who is desperate enough to accept the risks of brain surgery may be willing to work within those constraints. Someone looking for a decades-long solution, one that provides not hundreds but thousands or tens of thousands of recordable neurons, is going to need something fundamentally different.

“It’s going to have to have flexible kinds of electrodes that don’t carve up the brain if it moves relative to itself,” Dr. Miller says. “There are all kinds of materials issues, as well as electronic issues that clearly need to be solved.”

The elephant in the room, as Dr. Miller readily acknowledges, is Neuralink. Elon Musk’s company has poured resources into the hardware problem at a scale that academic research programs simply cannot match, and the results have been notable. Neuralink has implanted devices in multiple human patients, representing a meaningful measure of clinical progress.

Dr. Miller is measured in his assessment, noting that the company does not publicly share detailed results, which makes rigorous evaluation difficult. But he does not dismiss what they are doing. “We just need a lot of money to throw at the problem,” he says. “There are other companies that have not had as much success as Neuralink has, but there are a lot of people getting into the game.”

Breakthroughs in Communication: Where the Field Is Actually Winning

While the hardware problem continues to constrain what is possible in motor control, another corner of the BCI field has seen remarkable progress in recent years. The ability to restore communication to people who have lost it, whether through ALS, a brainstem stroke, or severe paralysis, has accelerated faster than almost anyone in the field anticipated.

Dr. Miller points to the work of two research groups in particular: Eddie Chang’s lab and the BrainGate consortium led by Leigh Hochberg. What makes their results striking is not just the speed of the decoding but the fact that the two groups are achieving similar outcomes using completely different hardware. One relies on a small intracortical array and the other on a much larger set of electrocorticography electrodes laid across the surface of the brain. “The communication interfaces have just exploded in the last few years,” Dr. Miller says. “We’ve gone from being able to work things out character by character every ten seconds or so, now to having almost conversational speed with pretty good accuracy.”

The ingredient that changed everything, Dr. Miller says, is the large language model. The raw neural signal, no matter how well recorded, is full of errors. On its own, it is not enough. But run those imperfect signals through a model that understands the statistical structure of language, and the system can make intelligent guesses about what the person is most likely trying to say. “The way it works is they can decode likely outcomes from the brain with respect to what you’re probably saying, but they’re full of errors until you run it in the large language model,” he explains. “Then it says, ‘what they’re probably saying, based on what I know about the statistics of language, is this,’ and it’s remarkably successful.”

The natural question is whether the same approach can be applied to movement. Dr. Miller thinks it might. His hope is that a foundation model built on muscle-activation statistics could do for motor prosthetics what large language models have done for speech. “With an EMG model, what I would hope we could do is say, Well, what they’re probably doing is this, this, or that, looking at the statistics of the way muscles are activated,” he says.

“I think that has the possibility of making the kinds of BCI prosthesis that I’m interested in significantly more successful.” He is careful to add that this remains a possibility, not a certainty. The extrapolation problem, teaching a model to predict movements it has never seen before, is genuinely hard. “Predicting the future is hard,” he says. “There’s nothing obvious now that is a clear path to making that work better.”

Ethical Questions the Field Is Still Working Out

The patients who volunteer for BCI implants are not a random cross-section of the population. They are people living with spinal cord injuries, ALS, brachial plexus damage, or brainstem strokes, conditions that have left them with little or no ability to move or communicate. The decision to undergo brain surgery in the context of a research study, with no guarantee of personal benefit, is not one anyone arrives at lightly. For researchers like Dr. Miller, ensuring that the decision is fully informed is both an ethical obligation and a genuine source of complexity.

Informed consent in BCI research presents specific challenges that go beyond those typical in clinical trials. Participants must understand that the procedure is experimental, that the surgical risks are real, and that any benefits are likely to accrue to future patients rather than themselves. “You obviously need to volunteer for the surgery, so there needs to be accurate informed consent,” Dr. Miller says. “We need to tell them this will be of no benefit to you, although it may benefit people in the future. Making sure you represent the risks accurately, without any kind of coercion, those are standard problems with consent for any kind of procedure.”

What researchers have found, rather unexpectedly, is that many participants report meaningful benefits even within the constraints of an experimental study. There is a community of implanted patients who share their experiences online, and the psychosocial dimension of participation appears to matter more than anyone anticipated. “Those persons who have been implanted, for the most part, enjoy doing it,” Dr. Miller says. “There is some kind of benefit to them psychologically, socially.”

But that same connection creates a problem the field has been slow to reckon with. When a research grant ends, what happens to the people who have been implanted? Dr. Miller describes this as an unresolved issue his team is actively navigating with the NIH right now. “What do you do when your funding runs out?” he asks. “Do you say we’re done, we’re going to take it out, that’s it for you? How do you pay for the explant cost? That’s something we’re dealing with right now with NIH, because they have no mechanism for paying for something beyond the length of the grant. We didn’t think about that at the beginning.”

Building the Field: What It Takes to Work at the Edge of Neuroscience and Engineering

The field that Dr. Miller works in does not reward narrow specialists. It demands fluency across neuroscience, engineering, materials science, data science, and clinical medicine, and the ability to move between those worlds without losing the thread. For students considering this path, his advice is practical and, coming from a distinguished professor at one of the country’s leading research universities, a little unexpected.

“From a slightly cynical perspective, I would say, learn how to write,” he says. “You can’t get funded, you can’t publish papers if you can’t write a coherent article. Many of them can write a coherent article. The number that can write a persuasive article and grant application is small.”

Writing, though, is only part of it. The other is knowing who to call. Before a single word of a grant proposal gets written, Dr. Miller convenes the people he wants to work with and simply talks through the problem. “I get that group together and we just talk about the problem,” he says. “What’s the nature of the problem? What are the possible approaches? What are the technical hurdles? What is it that we are never going to be able to do? I might spend half a year having that kind of discussion before we get down to putting metaphorical pen on paper, and that process is really fun.”

That collaborative instinct is something he works to pass on to his trainees. The days of the lone scientist pursuing a singular idea are not compatible with a field that requires neurosurgeons, data scientists, materials scientists, and roboticists all pulling in the same direction. “I’m reasonably intelligent, but there’s no way that I can cover the huge range of multifaceted skills that one needs in this field,” he says. “PhD stands for Piled Higher and Deeper. We tend to put on blinders, and to a certain extent you need those blinders, because it has become so specialized, but at the same time, you need to be able to talk the language of the people you need to collaborate with, or it won’t work.”

For all its difficulty, Dr. Miller has little trouble attracting the next generation of researchers to the work. The field sits at the intersection of science fiction and real-world clinical impact, and that combination is hard to resist. “People are drawn to it,” he says.

Kimmy Gustafson - Writer

With a keen understanding of the challenges and opportunities in higher education, Kimmy Gustafson regularly contributes insightful articles to OnlineEngineeringPrograms.com, providing readers with a comprehensive view on the evolving landscape of engineering education since 2019. She has interviewed many engineering experts on a range of subjects, including geoengineering.

Kimmy has been a freelance writer for more than a decade, writing hundreds of articles on a wide variety of topics such as startups, nonprofits, healthcare, kiteboarding, the outdoors, and higher education. She is passionate about seeing the world and has traveled to over 27 countries. She holds a bachelor’s degree in journalism from the University of Oregon. When not working, she can be found outdoors, parenting, kiteboarding, or cooking.

Related Features

Tech 2020: Non-Invasive Medical Nanotechnology

Invasive medical procedures often involve pain and permanent scarring. With a scale on a molecular level and with the ingenuity that comes with engineering, nanotechnology and biomedical engineering have great capacity to transition invasive medical procedures into non-invasive ones.

How to Get an Engineering Internship (2026)

Engineering internships are an increasingly important part of the transition from student to engineer. Internships provide an opportunity to put theoretical skills to work in hands-on environments. They also give engineering students valuable work experience, networking opportunities, and future career options.

Humanoids & The Future of Robotics Engineering

New AI systems designed specifically for robots are turning what were once purely programmed machines into embodied AI: systems that can perceive their surroundings, reason about tasks, and act in the real world.

Interview with an Expert: A Spotlight on Self-Driving Vehicles

Not long ago, self-driving cars were science fiction. Today, not so much. Influential companies like Tesla, Uber, Apple, and Google boast dynamic auto-drive programs, and many new startups are following their lead.

National Robotics Week 2022: An Expert’s Guide

The 12th annual National Robotics Week (RoboWeek) takes place April 2-10, 2022. Established by Congress in 2010, this tech-focused week is about demonstrating the positive societal impacts of robotic technologies, and inspiring students of all ages to pursue careers related to Science, Technology, Engineering, and Math (STEM).