FOCUS In Sound #9: Bijan Pesaran Interview
Welcome to FOCUS In Sound, the podcast series from the FOCUS newsletter published by the Burroughs Wellcome Fund. I’m your host, science writer Ernie Hood.
On this edition of Focus in Sound, we meet a young investigator who is conducting groundbreaking basic research to characterize how the brain coordinates our movements and how it operates when we’re making decisions. These are both very basic but complex neural operations, and understanding their mechanisms will not only shed light on brain function, but will contribute to development of brain-machine interfaces.
Dr. Bijan Pesaran of New York University received the Burroughs Wellcome Fund Career Award in the Biomedical Sciences in 2004, which funded his study on cortical mechanisms for hand-eye coordination.
Bijan earned his Ph.D. in Physics at the California Institute of Technology in 2002, where he served his post-doc until 2005. He has been on the faculty at NYU since 2006, where he is an Assistant Professor of Neural Sciences at the Center for Neural Science. Among his other honors, he has received the James D. Watson Program Investigator Award, an Alfred P. Sloan Research Fellowship, a McKnight Scholar Award, and the National Science Foundation CAREER Award.
Bijan, welcome to FOCUS In Sound…
Thanks for having me, Ernie, it’s a pleasure to be here.
Bijan, would you fill us in a bit more on the broad outlines of your work to characterize brain functions associated with movement coordination and decision-making?
The major theme in my research is trying to bring together our understanding of how different parts of the brain process information and understand how the different parts of the brain work together to process that information. We know a lot about how sensory information enters the brain, and we know a lot about the connections between the brain and the body, which are necessary for us to move. And what we found out is that there are many different regions within the cortex, the part of the brain, which is involved with thinking. We know that those regions in the cortex are involved in computing the transformations that take that sensory information as it comes into the brain and converting it into movement commands that allow us to move. So these many different regions of the brain are working together, but we know almost nothing about how they work together. We just know that they exist, they are active, and they’re critical for the process, but we don’t know how it all works. So what I’ve been trying to do is find ways to piece together the brain networks that are involved in guiding these movements. The networks are complicated. There are many different parts of the networks that are obviously millions of neurons working together in guiding any given movement that we make. And so it’s really been a technical challenge to try and find a way to make sense in the complexity of the circuits.
I know that much of your work focuses on the neural phenomena known as spiking and local field potential, or LFP, activity. Since those concepts are so central to your findings, would you describe each of them for us in simple terms?
Yes. This is really the crux of the approach that we’re trying to develop. Neurons fire action potentials. Action potentials are little spikes in their activity, and when you’re listening to the activity of a neuron on a headphone or on a speaker, you literally hear it as clicks or pops in the sound. So we hear spiking of neurons as being a cacophony of chatter in the brain. These action potentials are what neurons send to each other. These signals can be sent over great distances. In the most extreme case, you have a neuron in the cortex of your brain sending action potentials down to your toe to flex your toe muscle. And similarly, these signals get sent across the brain, throughout the different parts of the brain.
So this is how neurons talk to each other, and individual neurons create these action potentials or spikes. The challenge has been that we don’t really know how these different neurons are working together, because there are so many neurons that are involved in any given computation. Also the distances between them are so great that we don’t know which neurons are specifically working with which other neurons. So what we would like to know is that one neuron is specifically connected to another neuron and is changing the activity of that second neuron directly. That’s the way we think the brain works; that’s the model that we have based on overwhelming evidence. But when it comes to studying the brain in that way in a behaving organism, when we’re thinking about moving and actually moving, finding the connections that precisely between different neurons in the brain is basically impossible. So how do we make progress in that kind of situation? Well, what we want to do is to find ways to extract a signal from the neural network in a way which doesn’t depend on exactly which neurons we’re recording from. Because if it does depend on exactly which neurons we’re recording from, we’re never going to find the right sets of neurons; we’re never going to find the specific neuron that’s connected with another one. There are literally millions of neurons in the brain all connected to each other – billions in the human brain. If we were dependent on studying the brain at this microscopic level – neurons are smaller than a human hair – then we would basically find the problem to be intractable.
My interest in becoming a neuroscientist was really focused on tackling that grand challenge: the problem of piecing together the circuit when we have real trouble being able to study the circuit at the fine, microscopic level. The approach that I brought to the table was to use the local field potential. The local field potential is created by neurons and the action potentials that they fire, but is not created by an individual neuron. Local field potentials are created by groups of neurons. It turns out that they can be created by groups of maybe hundreds or even thousands of neurons, but not millions of neurons. By studying local field potentials, my goal has been to try and understand whether we can relate the activity of individual neurons, which we can record in one part of the brain, to the activity of groups of neurons in another part of the brain that might be connected with the neuron that we’re studying. So the local field potential is our way to try to coarse grain or average together the activity of small, interconnected groups of neurons that may be working with another neuron in another part of the brain.
The way we think about it is as if the neurons are sending action potentials between each other and receiving action potentials individually. We can’t resolve how neurons receive action potentials from other neurons in living animals, in awake animals. But what we can resolve is how a group of neurons may be receiving signals from a single neuron, and that’s really what the spike field approach that we’re using is. The spiking and local field potential approach is letting us ask how a small group of neurons in a part of the brain is receiving signals from an individual neuron in another part of the brain. And then what we can do is, of course, ask how that neuron is talking to that group of neurons; and what, when we’re thinking about moving, is that neuron saying to that group of neurons; and ultimately how is that group of neurons listening to the neuron; and whether or not the decision to move is really being formed together or being imposed on another group of neurons – how these processes come together when we’re thinking of moving.
So what types of methodologies do you use to measure these phenomena?
The advantage of studying signals like the local field potential along with the spiking of action potentials is a way of recording electrical activity in the brain. It can be done using the same electrodes, electrodes which have been refined and developed over decades and just have not been used in this way before. We’re using microelectrodes, so we take a fine wire electrode that has been sharpened to a very small tip. It’s carefully inserted into the brain, and since the brain has no pain receptors, the brain can’t feel the electrode at all. We’re in a position in that situation to listen to how the brain is working, and since we can do this in different parts of the brain at the same time, we can listen to the activity or the chatter of a given set of neurons in one part of the brain and then how that chatter is related with the chatter and the activity of another part of the brain. This is a traditional approach; the trick is really how we bring together the analysis of the neural signals. So once we record these neural signals, the major challenge is in interpreting them and understanding them. We use somewhat sophisticated statistical techniques to make sense of the wide range of signals that are present in the brain as it’s working to support our behavior.
You’ve arrived at some quite unexpected and enlightening results – would you describe some of your key findings for us?
The first major finding was [that] the local field potential itself was worth studying. When I started analyzing local field potential signals, the dogma amongst my colleagues and in the field was that they really didn’t tell us much about the activity of small groups of neurons. People thought the local field potential was like the EEG, which is a signal that’s measured on the scalp outside of the brain using quite large electrodes. And they thought that, like the EEG, the local field potential measured the cumulative activity of millions of neurons spread over centimeters in the brain and couldn’t really tell us the precise things that we need to know if we’re going to understand how movements are being controlled by the brain. It turned out that the local field potential was very informative about the precise things that we’re doing when we’re moving, in a very unexpected way. So that really kind of inspired me to think, well, maybe this is a signal which we should be studying more. Instead of just assuming that it’s unrelated to the properties of small groups of neurons, maybe we should test that idea and see how much information can be extracted from the field potential and how it’s related with the activity of individual neurons.
The way we talk about it is decoding movements. When we understand how the brain is working, we think that we can show that, because we can decode what the brain is doing. We can predict what the subject is thinking about; we can predict where they are going to move. I think one of the great successes in neuroscience is that we can record the activity of individual neurons and with great precision reconstruct what the subject is going to do in the future. So when you’re listening to the activity of these neurons, you can literally predict what the subject is going to do next. And it turned out that we could do the same kind of analysis based on the field potential, and we could also predict with as much accuracy what the subject was going to do next as we could when we listened to individual neurons. So that was kind of an important finding and has kind of stayed with the research program since then.
The next kind of important step in my research program was asking, well, OK, so the field potential’s informative – can we use it to study brain networks? Can we use it to understand how one neuron is talking with another group of neurons? Again, no one had looked at things this way before, the assumption being, of course, [that] the signals were too coarse in the local field potential, and it wouldn’t be able to resolve and define? how one neuron is communicating with another group of neurons. But it turned out that you could identify groups of neurons that were actually communicating with other neurons in another part of the brain by using the local field potential and the activity of the different neurons together. Those are two sorts of major findings, that the field potential is specific and that neurons in other parts of the brain are actually communicating with the part of the brain that you’re recording the field potential from, and that you can resolve this experimentally. Those are the key findings that allow us now to move forward and ask questions about how the brain is actually working across the different brain regions sending signals to each other.
You are one of the rare researchers who work both at a very basic and at an applied level – tell us about some of the practical applications associated with your work…
When you say things like, we can predict what a subject is going to do or we can predict what they’re thinking about, they are quite compelling statements. They have a lot of power to them, and they can lead to really quite dramatic new technologies which can treat people who have movement disorders and ultimately, other mental health disorders if you can tap into their thoughts like this. The problem is that those technologies have yet to be realized because of the difficulty of recording from the individual neurons. Just like it’s difficult to resolve how one neuron is communicating with another neuron, it’s also relatively difficult to record from an individual neuron itself. The brain doesn’t have pain sensors, but it does have inflammatory responses, and when you place a microelectrode into the brain, ultimately, the brain wants to reject that foreign body. To record from a neuron, you need to get really close to it, because they’re so small and the signals are so faint. So you need to be within 50 microns, which is, again, finer than the size of a human hair. The disconnect between what we can do and what we could do is quite a challenging problem that has attracted my attention. And again it comes back to the field potential, because the field potential is a lot easier to record than individual neurons. Reporting from small groups of neurons, we don’t need to be right next to a specific neuron. When we made the earlier finding that the field potential could be used to predict what we were thinking about, it kind of addressed a very practical issue. , That was, well, maybe now we can interface with the brain and decode people’s thoughts in a very realizable technology that may actually help people.
The promise of that approach, to use field potentials to monitor our thoughts, has driven another line of research in my lab, which is to kind of demonstrate what we know about the brain by building brain-machine interfaces – by building machines that are listening to the brain, and letting subjects control the machines by changing what they’re thinking about. In that way, we can have subjects simply thinking about movement without actually being able to move necessarily and then translate those thoughts into actions that would be realized by a robotic device or by a computer interface. In the more sort of science fiction directions, you could even have a system which is correcting for brain activity patterns which are abnormal and could treat a mental health disorder, because we’re now tapping into the brain directly and treating its abnormal patterns of activity. All of those ideas depend on, of course, being able to interface with the brain and being able to do so for the lifetime of the subject. We don’t want to have brain surgery every six months to fix a broken electrode. We want to have a system which is reliable and works for decades. But the field potential has the potential to do that. I wanted to have an applied line of research in my lab to explore these ideas and to validate their potential.
So how close are we to having the basic knowledge sufficient to actually be able to design one of these working brain-machine interfaces for humans?
We’re very close. I mean, there are labs and hospitals in the US these days which are implanting electrodes to demonstrate the proof of concept. So we have systems which are used by paralyzed patients; patients who have lost the ability to move any of their muscles but have retained the ability to think about moving those muscles are involved in clinical trials right now. What we’re learning from those trials is that the proof of concept works, that we can interface with the brain, that we can record the activity of neurons in the brain, individually and groups of neurons, and decode that activity to help these patients in a relatively rudimentary but successful manner. What we haven’t yet resolved is the problem that when inflammation arises these electrodes get rejected after relatively limited amounts of time. And really it’s unpredictable, we don’t know which electrodes are going to work, we don’t know why precisely they’re failing; that’s really still holding back the field.
What we’re doing in my lab is mapping out the information that’s in the brain using the local field potential and using that to optimize our ability to interface with the brain. What we’re finding is that as we move electrodes in the brain and sample them at different depths within the cortex, the signals change within hundreds of microns, within a fraction of a millimeter. If you move an electrode, you’re going to get a new pattern of activity which is going to tell you something different about what the subject is thinking about. And since we’ve never made these maps before, this is really sort of a basic discovery that we can then use to design optimized probes to tap into that information. I think that when we optimize our probes to better harness the information that is present in the local field potential, since the local field potential is more robust than the activity of spiking, it won’t be so sensitive. Then we may translate that into a really practical clinical technology. And I’m not the only one working on these things. I should be clear, there are many labs, especially in the neurosurgical arena, where having systems which are more reliable is very important than in a basic science arena. So there are groups in Washington University, at Stanford, and Harvard. There are many other research groups, at Brown University and Pittsburgh as well, looking at signals that are not necessarily generated by individual neurons, but are more robustly present in the activity of field potentials and how they can be used to help patients with various different types of disorders.
Well, Bijan, that’s all very exciting, and I know your findings will certainly accelerate the process of developing that working brain-machine interface, and we’ll certainly be keeping an eye out for that.
I did want to ask you also, as I mentioned in the introduction, you actually got your PhD in physics – so how did you happen to end up in neuroscience, pursuing this particular line of research?
That’s a very good question; I often wonder how I ended up here myself. I never would have predicted this, of course. Studying the brain in the way I study it is not the kind of thing I ever thought was possible as a student who was interested in the physical world. I was really interested in the beauty of nature and its mathematical description. What really drew me to physics was the fact that you could describe the world in a very powerful and quantitative way. How I ended up doing neuroscience was, in a sense, a lucky break, because I was in England, and I was an undergraduate at Cambridge. I had been doing research as an undergraduate in physics, and I found that the questions did not consume me in the way that I think that research scientists need to be consumed by their questions. So I was looking around for other questions. I wanted to find an area of research which would really draw me in. The brain is a powerful thing; it’s our most complex organ; we know very little about it; and while I was walking the halls as a physics student, I saw an advertisement for research scientists at Bell Labs in New Jersey. It turned out that the research group that I was working in as an undergraduate had a connection with that group at Bell Labs. So I wrote to the director of the physics department at Bell Labs and said that I would love to do research on the brain, which was what they were doing at Bell Labs, and they brought me across. So I spent a year after I finished my undergraduate degree in a physics department. It was actually a theoretical physics department as part of a group of people who were interested in using concepts and ideas and approaches from physics to understand the brain. That was really the perfect match for me. I guess the rest is history in the sense that that was my entry point into neuroscience, and I became able to make a contribution in neuroscience in a quantitative way by analyzing neural signals; by understanding the physical properties of the signals; and by defining the physical significance. So that’s really how I got into neuroscience from physics. It was really through a sort of formative research experience at Bell Labs after I finished my undergraduate degree.
Terrific. Well, Bijan, it’s been a great pleasure for us to get to know you and your work, which is shedding so much light on basic neural mechanisms while illuminating a path toward improved outcomes in the real world. We wish you the best of luck for continued success, and thanks so much for joining us today on FOCUS In Sound…
Thank you very much. It’s been a pleasure.
We hope you’ve enjoyed this edition of the FOCUS In Sound podcast. Until next time, this is Ernie Hood. Thanks for listening!