Ah, the ole' BMI. I'm going to be honest with you all here. I once went to a conference and ended up sauntering into a talk about BMI, thinking it referred to body mass index. I'm thinking ok, sure, I'll go to a talk about how the brain influences body weight. After all, I do find that research interesting. Besides, there wasn't much else to do in Colorado in winter- I was very pregnant and it was snowing. So the talk won. Alas, poor Past Melinda had not yet learned the acronym for brain machine interfacing, the field of study pairing technology with the central nervous system. A technology Future Melinda would find absolutely fascinating.
Boy, was my mind blown. I mean, it was one of those rare moments in life. You know the kind because you can count them on two hands. Maybe even one. Yes, this was a one-hand-count kind of experience. I remember the talk as if it were yesterday. Actually, even better than that because my memory is absolute crap, so I'm not even sure what I did yesterday. But I sure remember the talk given by Miguel Nicolelis, MD, PhD.
So who is Miguel Nicolelis? Because I'm big on science communication, I'm going to describe him in two versions. The first version is that he is a Brazilian MD, PhD who works on brain machine interfacing at Duke University. The second version is that he is a man who helped a paraplegic kick a soccerball in the 2014 World Cup. I'll let you decide which version is more exciting.
So, how did Nicolelis do this? Well, first let's back up and discuss BMI just a bit. How does brain machine interfacing work? Imagine the brain is a computer. Each neuron is a bit (a 1 or 0). It either fires or it doesn't. They don't fire with varying degrees. They just do or do not (there is no try). Sorry, had to throw in a Yoda joke. Anyway, if you measure these 1's and 0's, you can detect a pattern for each neural command. In other words, in theory, you could map which neurons are 'on' during particular tasks. These signals are sent to a computer, which 'learns' these patterns, which are essentially codes. The brain trains this computer over several sessions. What this means is that every action we do can be expressed as a binary code. This is pretty cool, right? With this code, the computer can then do any sort of actions we program it to do. For example, if the computer learns our code for picking up a cup with our hand, we can have that computer command a robotic arm to do the same action. Sounds like science fiction. Yet these experiments have already been done...
(Can we just take a moment to appreciate that paragraph ending? I have really been working on leaving hooks.)
The prestigious journal, Nature, published findings from one such experiment in 2002. In this study, researchers trained a monkey to move a cursor towards a moving target on a screen using a joystick. A computer was trained on the monkey's neural signals and coded to have the ability to move the cursor. Here's the really cool part- the scientists then had the monkey play the game again, but they disconnected the joystick and had the computer play the game based on the codes the monkey's brain was giving. Even typing that gave me the chills. I mean, it's amazing!
The theory behind this idea can be applied to prosthesis and any number of applications. Dr. Nicolelis, being the badass superhero of neuroscience and BMI that he is, took these ideas to the next logical step- Hey, why not build a whole robotic exoskeleton suit and have a paraplegic not only walk but kick a freaking soccer ball? I mean, these are the types of next logical steps when you're a genius, I suppose, right?
So this is what he did. Now, of course, he didn't do it himself. This project involved >150 scientists around the globe. The team built a robotic suit that recorded EEG information from a person's brain. These signals were used to train a computer that could make the robot walk. This already sounds complicated enough but there were very interesting challenges that the team had to overcome.
When we walk, we do more than activate leg and trunk muscles. Our brain also depends on feedback from the environment. When we take a step, our brain gets feedback from pressure sensors in our legs and feet. These let us know that we have successfully executed the step. Or, if the ground was a little higher, lower, softer, or less sturdy than expected, we can correct for it. Without this feedback, the pattern of walking doesn't work. Knowing this, the team decided to add feedback to the circuit. But there's a problem- the human brain needs this sensory feedback to execute the next step. Paraplegics usually do not have feeling in their legs and cannot give this type of feedback. So how could the team proceed? Their solution was truly genius.
The team installed sensors on the bottom of the robotic 'feet' to detect pressure. They then translated this pressure into a vibration that was applied to the person's arm. Because brains are so amazing and plastic, the individual's brain began to associate arm vibration with pressure of a step. This completed the circuit enough to allow the brain to continue to 'walk.'
But walking isn't enough. Let's go for a soccer ball kick. Hell, let's go for kicking a soccer ball in the opening of the World Cup. (I imagine Dr. Nicolelis sitting around thinking this with some Brazilian pastry - do they eat pastries in Brazil? Maybe he was sitting around with a stick of roasted meat, I've heard those are delicious)
That's exactly what happened. Mr. Juliano Pinto - a man with complete paralysis of his lower body -kicked off the 2014 World Cup. This is what science can deliver, folks!
I hope you've enjoyed this read. Let me know if there are other neuro/AI topics you'd like to hear more about.
Your Average Neuroscientist, Melinda
Watn to learn more? Some follow-up links:
Miguel Nicolelis TED MED talk
Helpful scientific review article on BMI