The Future of Retinal Implants

0:00Daniel Palanker:This is Stanford Engineering's The Future of Everything, and I'm your host, Russ Altman. Since we started this show eight years ago, it's become an archive of amazing and impactful work by my Stanford colleagues. Research is not something that just happens in the lab, and as you'll hear on this show, the research at Stanford can impact areas like health, technology, law, and business, and many other topics that can affect everyday life. We hope you'll tune in to learn more about how research has the potential to help your life and to help the lives of people you care about in your family and your community.

0:32So PRIMA does stand for something. It stands for photovoltaic retinal implant. Okay, good. And so what it does, it basically tries to replicate the function of lost photoreceptor. So it's just a photovoltaic array. So it's a little, you can think about it as little solar panels, basically. Solar panels. So little pixels. So we have pixels in this array, photovoltaic array, that convert light into electrical current. The current flows in front of the electrodes from active to return electrode in each pixel and polarizes neurons sitting just in front of it. So the only requirement is proximity. As soon as you close enough and the current is strong enough, you polarize the nearby neuron.

1:24Daniel Palanker:this is the future of everything and i'm your host russ altman thank you for listening if you're enjoying the show please rate and review it we like to get a good rating and we'd love to see your comments it helps us form the show and make sure it's as good as it can be today daniel palanker will tell us that he can restore the vision of those who have lost it with electronic Arrays that are implanted into the retina. They're light sensitive and they talk to the brain. It's the future of retinal implants. Today, we're going to continue our feature called The Future in a Minute. At the end of my interview with Daniel, I'll ask him a few quick questions and he'll give me a few quick answers.

2:01Daniel Palanker:Also, thank you for listening and please rate and review. Give us a five if we deserve it. That helps the show.

2:13Daniel Palanker:So the loss of vision can be life-changing for individuals who it happens to. This can happen from diseases of aging. You may have heard of macular degeneration, which affects older people and can lead to loss of vision. There are also genetic diseases where children lose their vision early in their teenage life. So this is a terrible problem for many people and something that needs to be addressed. Well, in the past, there have been retinal implants, but they really only led to flickering light that allowed you to see light and dark, but they didn't really allow you to get to things like reading, writing, maybe even facial recognition.

2:51Daniel Palanker:Well, there's been some breakthroughs. And today I'm speaking with Daniel Palanker from Stanford University, where he's a professor of ophthalmology and electrical engineering. And his group has developed an implant that actually allows people to read and write. And they're looking at clinical trials that may advance to not just grayscale, but color, and also to the ability to see faces. So Daniel, to get started out, how did you decide to devote your career? You're an electrical engineer by training. How did you decide to devote your career to vision and ophthalmology and all things about eyes?

3:29Well, I'm a physicist by training. I was torn from the very beginning after high school to go to physics or medicine. And it turned out that I went to physics and then decided to apply to medicine. So that's a full circle. And I think it was the right decision. It also happened that I did my PhD in Hebrew University of Jerusalem in the field of applied optics with applications to ophthalmology, in particular retinal lasers. and that when I came to Stanford as a postdoc, I was in the Department of Physics, but we met on occasion with a new chair of the Department of Ophthalmology at Stanford, Mark Blumenkrantz, and he was looking for a physicist who can help him develop the field of ophthalmic lasers, and so he hired me in that capacity.

4:29So this is how I came back to ophthalmology at Stanford.

4:33Daniel Palanker:Fantastic. So we have a lot to talk about and including some recent work that's just quite amazing. But before we even get to that, for people who don't think about eyes and vision, I wonder if you can just give a summary of like what the key things we need to know about how the visual system works and the kind of the key vocabulary that you might be referring to when you describe some of your recent work. Yeah, eye is a beautiful combination of optics which projects the world onto the retina down to diffraction limit of about 3 micrometers. And neural tissue, the retina that converts this image into neural signals and it encodes it eventually in a digital format that propagates through the brain.

5:17So the retina itself is a three-layered neural tissue. The first layer of photoreceptors convert light into changes in cell potential, hyperpolarizing or decreasing potential of photoreceptors, basically proportional to light intensity. And this is, in the language of electronics, you would call analog electronics. It's basically a signal proportional to the input. And then there is a second layer of integration of that signal in bipolar cells primarily, also regulated by amacrine and horizontal cells, that keeps it still in the domain of analog electronics, but it enhances edges, it splits it into on and off pathways, and this adds complexity, but still it's analog electronics, and this is where we are hacking into, by the way, into that second layer when putt receptors are gone.

6:06But then on a third layer of ganglion cells, the signal is converted into digital format of binary spikes. I call duction potentials. And this is much more complex code. It spreads into multiple types of ganglion cells, about two dozen, each of them specializing in different features of visual scene. Some sample with high spatial resolution and low contrast midget cells. some with low resolution and high contrast parasol cells some are responsible for color vision we have three sensors of color we can call it red, green and blue and direction sensitivity and so on so all these two dozen streams propagating parallel they sample the whole visual field in parallel with mosaics of cells covering completely visual field And all the parallel streams propagates through optic nerve, one million wires, axons to the brain.

7:09And then it spreads further into areas which specialize in different aspects of visual scene until it all converges into person. So the code, as we go further and further away from the origin from photoreceptors, becomes more distributed, more complex and more abstract like in any neural network. And therefore, if you want to hack into the system and reproduce it, it becomes more and more challenging the further away you are from the origin.

7:36Daniel Palanker:Yes. And that's why when photoreceptors are lost in the diseases called retinal generation, we address it by introducing something that just replaces photoreceptors and stimulates the second layer of neurons, which is still analog, you know, a kind of electronics, analog processing of the signal. And code is relatively simple. not you know uh uh very simple but it's doable uh we managed to replicate it pretty well but going further away is more complex okay so a few pieces of anatomy that i want to make sure we so of course everybody knows that we have a lens in the front of our eye and it goes through a it goes through a transparent material the vitreous we call it then it hits then the light hits what you're referring to as the retina and in case and in case it comes up i just want to make sure that we talk about the macula, which as I understand it is the center of the retina that is kind of the most sensitive to light.

8:33Daniel Palanker:And I know we're going to talk about diseases of the macula in a second, but that's the part where you get your, tell me if I'm wrong, your highest acuity vision and your ability to detect small differences. And then as you said, there are these layers of cells, they go to the optic nerve, which goes to the brain, which is complex. Okay, so that's great. So now I want to talk about the things that go wrong in the eye that require the kind of amazing technologies that you're working on. So I know macular degeneration is one of the diseases that many people have heard of. Can you just give us a summary of what is going on in these key diseases that you're targeting?

9:13Yeah, so we target The disease is called retinal degeneration, and there are several kinds of them. The most common one is age-related macular degeneration. And this disease, patients gradually lose photoreceptors in the very center of the macular where we have highest resolution. And that's a very important and peculiar feature of vision that we are aware of, that acuity drops with eccentricity. So we have highest acuity right in the center of the visual field within a few degrees, but 10 degrees away and we are legally blind. Acuity is 10 times worse on access. And then, you know, 40 degrees away, we cannot even count fingers.

9:53So it drops very quickly. We are unaware of that because we scan and we, you know, look at what we want to pay attention to, reading and so on. But we become acutely aware of that as soon as we lose it. And patients with disease of HL8 macular generation, when they lose central vision, they cannot read. They cannot even recognize faces. And so the reason for this deficiency is that a supporting layer of photoreceptors called retinal pigment epithelium, that dark layer under the retina that actually is visible, retina itself is very transparent. But that dark layer with age becomes less and less functional and cannot support metabolism of photoreceptors.

10:37And photoreceptors are metabolically the most active cells in the body. so as a result retinal pigment epithelium starts dying off in photoreceptors as well and the gap of that it's called scotoma, blind spot, is growing over time expanding over time and people basically at some point cannot read, recognize faces that's one kind of disease another class of retinal generation is inherited There are many, you know, tens of hundreds, maybe even versions of genetic, you know, mutations that cause various aspects of loss of visual function for the receptors. And that kicks early in life, usually. So very often it begins in teenage, you know, or in college.

11:32And that's even more devastating because basically your productive phase of life is to live blind. And so we've started with age-related microgeneration because for two reasons. First of all, it's a bigger market and for companies it's important. And second is that retina because it starts late in life. retina is relatively well preserved, unlike the diseases that kick in earlier and retinary wires. And it's very important for us to maintain retinal structure because we want to retain retinal encoding. We want to stimulate the second layer of neurons after the receptors, but we want the rest of the retinal network to do its job.

12:13And that was our assumption from the very beginning. And I think it paid off because it shows that we retained many features of normal retinal processing with our implant. And that, I think, is the reason why our patients see form vision. Unlike all previous attempts of restoring sight, all patients saw were flickering lights, not really form vision. That, I think, is due to preservation of code. And that is in large part due to preservation of inner retina. Great.

12:44Daniel Palanker:Okay. So we're going to get to your implant. And this is great. And you've given us a little bit of a clue. Just to say that under normal circumstances, these retinal receptors that you're talking about, if I know this, that they're able to respond to single photons. So they're extremely exquisitely sensitive to light. But as you've described, we lose those. And so you have to, they are the ones who begin the electrical signaling, and you're going to have to bypass them. So now let's get to the exciting work. It's a system called PRIMA. I don't actually know if PRIMA stands for anything. You can tell me that in a second.

13:18Daniel Palanker:So tell us what is PRIMA? How does it work? And then we can back out some of the problems and the challenges and also the victories that come from it. So PRIMA does stand for something. It stands for photovoltaic retinal implant. Okay, good. And so what it does, it basically tries to replicate the function of lost photoreceptor. So it's just a photovoltaic array. So it's a little, you can think about it as little solar panels, basically, converting light into current. So little pixels. So we have pixels in this array, photovoltaic array, that convert light into electrical current. The current flows in front of the electrodes from active to return electrode in each pixel and polarizes neurons sitting just in front of it.

14:05So the only requirement is proximity. As soon as you close enough and the current is strong enough, you polarize the nearby neuron.

14:13Daniel Palanker:So let me ask you before you go on. Many of us know about pixels because we have cameras in our phones and they're getting bigger and bigger. For this first generation, how many pixels are we talking about? Right. So the first generation in humans have pixels of 100 micrometers. And our implant is 2 by 2 millimeters in the first generation. So it's basically 20 by 20. It's about 400 pixels. Okay. But relatively small. Now we have implants with 10 ,000 pixels and more in red, not yet in humans. But in humans, we started very conservatively with 100 microns. And I'm glad we did so because it works well and you want to start with success.

14:52Of course.

14:53Daniel Palanker:And do these pixels, do they detect black and white light or color? And is it yes, no, or is it a continuous range of intensity? Right. So the way we encode information is continuous, it's not just black and white. But software, so let me explain how the system works and where software comes in. So photoreceptors, as you mentioned, are amazing cells. They can amplify light or signal in terms of ratio of photons per second per cell to ions flowing through the membrane per second per cell. Can amplify by a factor of up to a million. That's why we can see single photons looking at the stars. And our photodiodes have no amplification.

15:39So the maximum, all they can do is convert one photon into one electron hole pair. So we, to generate enough light, enough current, sorry, to polarize cells, we need much more light. And that's, this is where augmented reality goggles come in. We cannot just convert ambient light, it's not enough. We amplify it. And so we put augmented reality glasses. There is a camera that captures an image. we can process it and project back into the eye using more intense light. But we don't want this light to be visible by remaining photoreceptor, photoreceptor surrounds the implant in the periphery. So we make it invisible.

16:18Daniel Palanker:It's 880 nanometers. It's near infrared wavelengths. It's not visible. So it's like your remote control on TV. You don't see it, but TV set does, right? Yeah. So that light is much more intense. It's still safe, but it is much more intense than normal. And that's how we compensate for the amplification that photoreceptors were providing. Gotcha. There's another feature that photoreceptors, together with layers that right on top of them, horizontal cells, provide some initial signal processing, contrast enhancement, edge enhancement. And that is also lost because horizontal cells are disconnected.

16:57There are no photoreceptors to regulate them. We do it by software. So, between the camera and projector, we can do pretty much anything. So we can do contrast enhancement, we can do obviously autofocusing, adjustment to a wide range of dynamic range of illumination. That's what also photoreceptors do, but what diodes don't do well. And so, camera does all, and image processing does all of these features. Now, when the image is optimized for the range where the implant works well, we project that image onto the into the eye part of it which falls on the chip is converted into image and this is how patients see so basically what patients do they look at the little screen on augmented reality glasses like google glass right yes or any other augmented reality glasses remaining now on the market and they see an image and it is gray scale it's not color because we stimulate to to provide color vision you need to be specific to each bipolar cells that carry specific color information and we don't have that resolution we stimulate several of them at once so what patients see they describe it as a sun color it's kind of white yellowish color so it basically responds to a mixture of all colors but it's not black and white it's gray scale we can modulate brightness however in the current clinical trial all the company did was to show that they can read and write and that requires just black and white so our software currently basically delivered black and white you know images of letters you know patterns and so on and our next step will be actually to utilize grayscale capability of the system and why it is important because the top uh on a wish list of patients is reading and writing but the second right after that is face recognition.

18:48They want to see faces. And faces do require dynamic range. You cannot present faces well in black and white. So our next trial that we actually started is optimizing software for grayscale representation. And in prosthetic vision, it's not as wide as natural. So we optimize it towards a range that they can see. So we measure that contrast sensitivity curve and optimize software to fit all the information within the range that they can see in a great scale.

19:18Daniel Palanker:This is the Future of Everything with Russ Altman. We'll have more with Daniel Palanker next.

19:35Daniel Palanker:Welcome back to the Future of Everything. I'm Russ Altman. I'm speaking with Daniel Palanker from Stanford University. In the first segment, we got a beautiful lesson on how the eye works. And we heard initial description of this amazing implant that allows people to read and write and see for the first time after losing their vision from diseases. In this segment, I'm going to ask Daniel about the limits of that technology, how good can it get, and also some of the details about how those implants get in and how they're powered. And we'll end up with the clinical outlook. When will this actually become available to patients?

20:11Daniel Palanker:Don't forget, at the end of the segment, we'll do a future in a minute where I ask some quick questions and get some quick answers. So, Daniel, how do these things get inside the eye of these patients in the human trials? And how are they powered? Yeah. So, the device is completely wireless. It's powered by light. And so, pixels just convert light into current and there is no power supply, no wires, nothing. That's what makes implantation very straightforward, much easier than any other implant in retinal prosthesis domain. It's just a little thin 2x2 millimeter panel or array that is inserted under the retina.

20:53The way procedure goes is that you inject fluid under the retina to lift it, and that's a standard beginning of subretinal surgery, which is done for many reasons, like removing membrane and so on. Then you make a cut of two millimeters, usually peripherally, so it doesn't affect central vision. And it's called retinotomy. And then you slide or inject. We have an injector of that little panel under. And then surgeon pushes it gently until it lands exactly where she wants it in the central part of the macula. And then retina is reattached. And the rest is optical. You project light and you get the...

21:32In fact, in RETS, we measure acuity in the same day. In patients, usually they leave it for a couple months for the eye to heal after surgery. But in principle, the retina is retouched right away and ready to go right away. Wow.

21:45Daniel Palanker:Okay. What is also amazing is that you can actually remove the old generation like we did this in RETS. You remove the low-resolution first implant. You can slide under a version of high-resolution implant, and it provides high-resolution right away. And that's another beauty that it's a replaceable organ in a sense. I was going to ask about that because I didn't know if the volunteers who get the early ones are then going to miss out on, and it sounds like they won't miss out. They'll be able to continue. They're asking about it. The question is whether the surgeons will want to do another surgery in an elderly patient, that's another issue.

22:21But in RADS, we demonstrated that, yes, you can slide it out, put another one in exactly the same location, and it works with high resolution.

22:28Daniel Palanker:Okay. So this is very exciting. We have our glasses, which are amplifying the signal. We have our amazing array on the retina, talking to the neurons and getting the brain going for vision. So I know you're thinking about the future. And I was wondering, what are the limits to this? Because we all know that technology gets better. You've already told us that we had 400 pixels, but it's going to be, I think you said 10 ,000 already. And so, but I'm sure you're thinking there are physical limits. It's only going to go so far. Tell me how you think about that and what does that look like? Yeah. So the most fundamental limitations are geometrical, like in all loss of nature, actually.

23:13Geometry defines scaling and loss of nature. And when you make pixels smaller and smaller, the penetration, it's like little fountains that comes out of the pixel, the current. It also shrinks in height, and it just doesn't go deep enough to stimulate our neurons, which are about, you know, 40 microns kind of deep. And they are often separated from the implant by some debris layer of another 40 microns. So there is a certain height we need to achieve. And when you make pixels smaller than 100 microns, at some point, it basically doesn't penetrate deep enough. And you cannot brute force yourself through that because there is a thermal limit and a chemical limit.

23:55And so it has to be geometrical solution. That's the most fundamental kind of thing here, that our solutions basically to leave the flat geometry and go into 3D. So we create a three-dimensional electrode of different configurations, simplest one as a pillars, where cells migrate around. And that's another very fortunate aspect of the subretinal space, that it is actually very dynamic. If you introduce something that provides voids for cells to move, they will move. And this way, they actually come closer to electrodes. And our pillar electrodes penetrate these gaps that I mentioned about, 40 microns, and get right to the target neurons.

24:37So this way, we can actually scale pixels down from 100 microns all the way to 20. We demonstrated this in reds and provide acuity, you know, corresponding to pixel size. So that scaling is geometrical solution. And it also basically is the same on all other brain machine interfaces. The question is how close it can get to target neurons. And the great example is cochlear implant where the electrodes are inserted in the cochlea. They are separated by about one millimeter gap from the target neurons. And that's why making them denser than one millimeter doesn't really help. It feels divergent. And so cochlear implants exist now for probably 40 years, but they're still in a range of about 12 independent channels, 20 electrodes, but about 12 of them are independent.