Journalist: Luming Cao
Luming: Welcome to SciSection. My name is Luming and I'm your journalist for this episode! We're joined today with Dr. Michael Beyeler. He is an assistant professor at UC Santa Barbara who directs the Bionic Vision Lab. Thank you so much for taking the time to meet with me today.
Dr. Beyeler: Yeah, thank you. It's a pleasure to be here.
Luming: Awesome. So could you talk about your academic background?
Dr. Beyeler: Sure. So I'm originally from Switzerland. I went to ETH Zurich to get a degree in electrical engineering. And sometime around that, I got really interested in the brain, how it works. There was a cool lecture on computational neuroscience telling me all about how neurons work and how they communicate in the brain and that can give rise to vision and perception and decision making. And that's just so fascinating that I wanted to continue down that path and it eventually led me to the US. So I studied at UC Irvine, and now I'm at UC Santa Barbara as an assistant professor. And I've never looked back since.
Luming: That's really cool. So can you give a brief overview of biological processes that enable vision? Let's just start from the very basics of vision.
Dr. Beyeler: Sure. So yeah, we move our eyes to see, right? So light enters the eye and hits the back of the eye where the photoreceptors live. These are cells that respond to light like rods and cones, and then transform that light into an electrical chemical signal that is sent to other neurons first in the eye and then later in the brain in the visual cortex. And somehow there's millions and billions of neurons that communicate with each other to create a conscious part of the world so that we could just look out into the world and coherently make sense of it. And yeah, that stuff has always been really fascinating to me. We're a little closer now to understanding how some of the processes work, but in general, it's still a big mystery.
Luming: Yeah, that's really fascinating. I feel like for people like us who can see, we often just take vision for granted, but it's such a complicated and amazing process.
Dr. Beyeler: We do. And in fact, 25% of the brain is dedicated to vision. [It] comes so natural to us, but it's actually, yeah, some really heavy computations going on behind the curtains.
Luming: Yeah. So can you talk about what is bionic vision? What is a bionic eye, for example?
Dr. Beyeler: Yeah, so my research right now focuses on cases where vision breaks down. So it's really.. it's about degenerative diseases of the eye or the brain such as retinitis pigmentosa or macular degeneration. So these are hereditary diseases that affect you in the later stages of your life. They're very common and there's no cure for them. And so a bionic eye comes into play as a treatment technology. So it's an actual chip that is implanted in the eye of blind people. And the idea is to stimulate the surviving neurons in the visual system to sort of fool the mind into thinking that it saw something.
Luming: Hmm. That's really cool. So can you talk about what are some different types of blindness? So does the bionic eye only help wind specific kind of blindness or like, how does it work?
Dr. Beyeler: Yeah, so I know it sounds like science fiction, but there are already some devices out there and they work mostly for these two diseases, RP and AMD. So it's for diseases of the eye which are very common in the developed world, apart from like glaucoma and maybe cataracts in the developing world. This is like one of the causes of blindness. And so the bionic is targeting exactly these diseases by electrically, stimulating the cells, basically. So in these diseases, you have some cells that degenerate, and if the photoreceptors are among them, then you lose your ability to see, because the photoreceptors are the ones that respond to light. And so the idea of the chip is to replace the lost functionality with a microelectrode array, where we go in and stimulate individual neurons to make them active and communicate with the other.
Luming: Oh, so does the bionic eye require, kind of like, a surgery to put in some chips that replace photoreceptors?
Dr. Beyeler: Yeah, in general, that's the idea. So often the system comes with a camera that is embedded in a pair of glasses that the patients are wearing. And so the camera's constantly recording video. And the idea is to translate that video into electrical pulses that are delivered to the retina via the electrode array. So it's really, it's a problem of translating a video basically into a series of electrical pulses that the brain can now understand.
Luming: That's really cool. So how good is the technology right now? Can, what kind of images are the patients seeing?
Dr. Beyeler: Yeah, I would say it's very early days still. So the currently available devices will produce very low-resolution, kind of foggy vision as far as we understand. And so it's very different still from natural vision, but that is exactly where we come in. So we are really interested in understanding what it is that these people really see when they use their devices. That has proved to be really hard to figure out just because we, you know, they're blind, right? How, how are they going to tell us what they see? So it is really been hard to get a scientific understanding of how this vision is qualitatively, fundamentally different from normal, natural vision, we will call it.
Luming: So how do you actually figure out what these people are seeing? What are some techniques you and your researchers use?
Dr. Beyeler: Yeah, so far it's, it's mostly been behavioral observations. So we would sit with these patients and we would turn out a single electrode and ask them to draw on a touchscreen, what they saw. So if you turn on a single electrode often people think of it as just turning on a pixel in an image. But when we ask patients to draw what they see, they very rarely drew pixels. Most of the times, they drew more complicated shapes, such as like arcs or triangles and all sorts of weird wedges and things. And so that is one hint that what they are seeing might be different from what we expect them to see. So similar to that, there's a whole bunch of observations you could do by turning on multiple electrodes, asking people to rate the brightness, for example, of an individual thing they see. And after you have collected all this different data, it's sort of our job to piece it together and build a computational model that would allow us to predict for any given stimulus what the patient should see. So at the end of the day, it really is very very interdisciplinary work. So it involves a lot of cognitive psychology, behavioral psychology, but also computational modeling. So you need to know the underlying neuroscience of how the eye works and how simulation might affect vision. And you also have to have computer science experience because you want to build computational models that can predict a vision for new stimuli. So overall it's a very exciting and application-driven field.
Luming: Yeah. So as you said, vision is really complex because it’s comprised of like depth, color, shapes, and all these kinds of things. So are these specific aspects of vision that are hard to, I guess, replicate for these patients?
Dr. Beyeler: Yeah. As far as we understand so far there's no color perception at all. Most people say what they see as is maybe grayish or maybe has a yellowish tint, but it's really hard to reproduce color. On top of that, shape is just lost or distorted, which makes it really hard to recognize some of the things. And, and so I guess the more general answer to that is it really depends on where you stimulate in the brain. So as we start in the retina individual neurons respond to very small neighborhoods of the visual scene. If you think of an image, one neuron might encode the top left corner of this image and tell you what color is in there, what edge, orientations, and motion, all these different aspects of vision, what is present for the upper left corner. But as you move on deeper into the visual cortex, neurons respond to more and more complicated things up to the point where you have a neuron that fires only when you see your grandmother—It's not entirely true, but that's sort of the idea that as you go higher up neurons become more complex and they're risk properties. So whenever you turn on a neuron like that, you tell the brain, “Hey, I just saw my grandmother,” and the brain has to piece that information together with other input it gets. And so it becomes really hard that you basically have to understand what each brain area is doing, how it is contributing to our conscious perception of the world, so that you can go in and kind of hack into the brain and turn on the right neuron at the right time.
Luming: Cool. So how do you stimulate different areas of the brain? Do you do it through the optic nerve or?
Dr. Beyeler: So the current implants that are commercially available or in clinical trials, they're mostly implanted in the eye itself. So under the retina, but there are other devices in development that target the optic nerve, they target the primary visual cortex and then higher up in the brain. That hasn't been very successful. But let's say up to the primary visual cortex is where you could implant. And so depending on the implant location you're dealing with a different population of neurons. So you have to understand what these neurons typically do in actual vision. So you have a clue how to activate them.
Luming: Cool. So you mentioned that neuro-engineering is very interdisciplinary. So how did you integrate all of your interests that led to where you are today and how do you coordinate all these types of fields in your research?
Dr. Beyeler: Yeah, it's a very exciting thing, but also challenging. So on a practical level right now, I have a joint appointment both in computer science and psychological and brain sciences. And so I try to recruit students also across domains because I think that's really necessary to make progress in this application domain—you have to have a little bit of everything. And I think I've always functioned like that. Even though, you know, a program might have been called computer science or computational neuroscience, really the skill you needed came from a broad range of areas. And so that's very much who I am as a scientist, and that's also who I want my students to be or become.
Luming: Great. So what advice would you give for students who might be interested in pursuing a career in neuro-engineering?
Dr. Beyeler: You definitely need a strong computational background, and I would say it doesn't really matter what you study, as long as you get the methods down, because it tends to be a little easier to learn the neuroscience after you already have the computational background. It's a little harder to go the other way around, at least in my experience. But at the end of the day, I think science nowadays is much more interdisciplinary as it's been maybe a decade ago. If we just think about, you know, the applications of machine learning and AI in, in different domains, and if you want to do application-based science sooner or later, you're gonna find yourself in a very interdisciplinary field. And that also takes a certain type of person or a researcher. If you like that stuff, then I think you're going to thrive. It's going to be really, really, really good.
Luming: Yeah. Great. So to finish up do you have a favorite work of science fiction?
Dr. Beyeler: Well, probably the dorkiest answer is I really like The Hitchhiker's Guide to the Galaxy, Douglas Adams, kind of classic. But I also like more contemporary stuff like Black Mirror for example, had actually had a lot of neurotech in there last season. So that was interesting to see, you know, what maybe the public thinks about the potential and possible implications of such technology. So it was very interesting to see from a science perspective as well.
Yeah, that's really interesting. And on that note, thank you so much for talking with me today. So that's it for this episode of SciSection! Please make sure to check out all of our newest interviews.