Power in prosthetics: the future of neurotechnology

Blond man with a prosthetic limb running.
Jonnie Peacock at the 2016 Rio Paralympics. Courtesy The Independent

At the age of five, tragedy struck for Jonnie Peacock: a meningitis infection resulted in his right leg being amputated below the knee. His disability didn’t stop his love for running however, and he went on to win the gold medal in the 2012 and 2016 Paralympic Games. Jonnie’s story couldn’t have been possible without advanced the prosthetic limbs that enable disabled athletes to participate in the highest echelons of sporting events.

Despite their progress, prosthetic limbs remain far from perfect. In the United States alone, there are over two million people who suffer from limb loss, and more than 75% of them have to use purely mechanical prosthetic limbs that are imprecise, difficult to control, and require a great deal of training to use for complicated tasks.

That’s where Jose Carmena, professor of Electrical Engineering and Neuroscience at UC Berkeley, comes in. Professor Carmena’s research centers around brain-machine interfaces (BMIs), which allow for communication between the brain and an external computer. Carmena’s lab is working to use BMI technology to develop smart prosthetics that patients can control more naturally and with greater precision. The beauty of Carmena’s BMI technology is that the brain and the computer can learn from each other. The brain learns to control the prosthetic limb, while at the same time, the computer analyzes the brain’s activity and accelerates its learning process.

Diagram showing how prosthetics can interact with the brain.
The process for controlling a prosthetic limb using a BMI. Courtesy Science Direct.

As the diagram on the left shows, the BMI process starts with sensors that measure the brain’s electrical activity.

These signals are then translated by decoding algorithms into movements of the prosthetic limb.

However, what if the limb didn’t move quite the way the brain wanted it to? Carmena’s lab has developed a technique known as closed-loop decoder adaptation (CLDA) in which the parameters of the decoding algorithms are adapted so that the limb’s movements more closely match the brain’s intended movements. This decoder adaptation improves neuroprosthetic control and more importantly, accelerates the brain’s process of learning to control the prosthetic limb.

The beauty of the BMI is that the brain and the computer can learn from each other.

Within the BMI research space, scientists have developed two approaches to designing neuroprosthetic technology.

One approach is to train a computer algorithm to recognize correlations between brain activity patterns and certain movements of a limb. Then, the algorithm would control a BMI that moves a prosthetic arm based on its neural activity. The problem with this approach is that such an algorithm couldn’t be trained on patients who are missing their arms or who are paralyzed. In these cases, there wouldn’t be any training data for the algorithm to analyze, so it wouldn’t be able to learn how an arm is supposed to move. Furthermore, in such an application, the animal’s brain and the BMI would be two distinct entities; the brain would have no idea that the BMI even existed.

Instead of that approach, Carmena wondered why not use the inherent learning mechanisms in the brain? His approach focuses on incorporating the BMI and the brain into a combined neural representation. The brain would serve as the learner, learning how to use the prosthetic limb just like any other limb on the human body. The computer would help the brain learn how to control the limb faster, but it would never have direct control over the limb’s movements.

The ability of the brain to learn and adapt by tweaking its internal structure is known as neuroplasticity. Neuroplasticity plays an important role in BMIs because it allows the brain to incorporate a BMI into a feedback loop to constantly receive and send information. Carmena’s hypothesis was that animals should be able to not only develop new skills from using a BMI, but also retain these skills while continuing to learn new ones.

Why not use the inherent learning mechanisms in the brain?

Think of it this way: if one week you learned how to drive a car and the next week you started to learn about thermonuclear astrophysics, chances are you wouldn’t forget how to drive the car in the second week! This was the motivation behind learning about neuroplasticity and testing it on lab rats. For 20 days, Carmena’s lab would use the same algorithm to train rats to complete a certain task, and every following day they noticed an improvement in how much time it took the rats to complete the task. Afterward, the rats were given a new algorithm to learn a new task, and to Carmena’s delight, the rats still remembered the old task! This breakthrough paved the way for future research in neuroplasticity.

Knowing that the brain can adapt itself and retain these adaptations over time is important, but the fact remains that learning how to control a prosthetic limb is a very challenging task for the brain. Between a human’s shoulder and fingertips, there are twenty-four muscles that work together to move the arm and reach for an object. As Carmena explains, it would take half a page of equations to model a robotic limb that could move with two degrees of freedom. For five degrees of freedom, you’d need a book. Now imagine trying to model twenty-four degrees of freedom just to move one prosthetic limb. “It took millions of years of evolution, literally, for the first mammal to learn how to reach with a limb,” Carmena explains.

So not only does the brain have to relearn millions of years of learned movement, but it also needs to decide which neurons are important for moving the limb. One possible solution for the brain is something called the muscle twitch hypothesis. Imagine that there are neurons in the brain that, when fired, cause certain muscle twitches. So all the brain needs to do is find the right sequence of muscle twitches that cause the limb to move and voila, we have control…right?

Neural dust implanted on a nerve.
The “neural dust” microchip, shown here implanted in a rat’s nerve fiber. Courtesy Berkeley Science Review

Not really, explains Carmena. In the muscle twitch case, the brain is just invoking existing networks of neurons instead of learning to choose the right neurons to perform the task. The goal is to use BMIs to have the brain refine its choice of networks over time. Initially, when the animal first tries to move the limb, many networks will activate across the brain. Eventually, the brain will learn which networks are necessary for controlling the limb, and it will modulate these networks more actively than the ones that are less directly involved with the prosthetic.

The next step was to get the BMI inside the body so that it could be used as a longer-term clinical solution. The implanted device would have to collect neural data and transmit it wirelessly. And as if figuring out these two challenges weren’t complex enough, the researchers also wanted the device to be capable of surviving inside the body for a decade, which meant it couldn’t rely on a battery for power.

Luckily, Michel Maharbiz, professor of Electrical Engineering and Computer Sciences at UC Berkeley, who was collaborating with Carmena to solve this problem, had a revelation. Ultrasound waves — energy waves at a frequency higher than humans can hear — could actually be used to power implantable devices. This meant that the implants, dubbed “neural dust,” could be powered and integrated into the brain’s feedback loop for many years.

Carmena’s breakthroughs have ushered in great advances in neuroprosthetics. But after 15 years of working from his lab and publishing papers, he decided it was time to release his work to the public. Two years ago, he developed a company with Maharbiz called Iota Biosciences. The pair hope that their neural implants can decrease caregiver time and enable patients to perform tasks that once seemed impossibly out of reach. Carmena’s achievements have undoubtedly set him apart as a visionary among his contemporaries, and it is only a matter of time before neuroprosthetics come to a brain near you.

The work that engineers do shapes the world around us. But given the technical nature of that work, non-engineers may not always realize the impact and reach of engineering research. In E185: The Art of STEM Communication, students learn about and practice written and verbal communication skills that can bring the world of engineering to a broader audience. They spend the semester researching projects within the College of Engineering, interviewing professors and graduate students, and ultimately writing about and presenting that work for a general audience. This piece is one of the outcomes of the Fall 2019 E185 course.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store