John von Neumann Distinguished Lecture Series, 2015: David Berson

my name is Barry Connors I'm the chair of Neuroscience here at Brown and my main job will be to introduce the introducers and I'm going to start by introducing uh Ed Hart who is the Alva away University professor of Medical Science professor of Medical Science associate dean of medicine and biological sciences here at Brown and he's going to give us some opening remarks before we kick off with the main show thank you Barry and thank you all for coming through this exciting uh days uh day of lectures and discussion and culminating the the week of Stellar presentations and again discussion and the the exciting format of the sweatbox where you know students can Dive Right In and query their uh their The Luminaries in the field about the underpinnings of their uh of their pronouncements and uh and scientific conclusions so I stand here today to and represent Jack Elias who uh is our dean of medicine and biological sciences who unfortunately couldn't be here today because of other scheduling uh uh conflicts and uh and I represent the division of biology and Medicine which uh as many of you know is uh the faculty of which is engaged in the education of not only the medical students but also the undergraduate biology concentrators and the Neuroscience concentrators to and those undergraduates together account for about 20 percent of the graduating class every year and so we're excited to have the you know and I'll say a few more comments about some of the contributions that the undergraduates have have uh have made in in areas that we're we're very interested in uh in partnership with the faculty in computer science many years ago over 25 years ago uh the biology faculty and uh and computer science faculty organized an undergraduate concentration in computational biology which is one of the earliest if not the first such major undergraduate concentration in the country and that's still going going very strong and it's also been expanded Into The Graduate area so that we have in within the division of biology and Medicine we devote some of our resources to support graduate slots in computational molecular biology and that particular program is organized under will Fair brother who was a member of the ccmb center for computational molecular biology and Department of molecular cell biology and I believe they filled filled their slots this year and uh and overall in fact we've had a very successful year of graduate recruiting in the division of biology and Medicine we at the latest count have 44 students who are will plan to matriculate in September and join our various programs and there may even be another one that's uh that uh we're we're going to accept due to a computer error on College net he emailed us and he said I hit the yes button and so we're dealing with that right now um so um about that that group of 44 so we're bringing in 44 students overall in our eight graduate programs and those account for that's actually setting a record for our our division of biology medicine that's the largest class we've ever been brought in uh On a related front I'm happy to announce that we have recruited uh a uh a pair of new faculty who into the into the medical school who will lead the effort on biomedical informatics so again we in biology really understand the importance of computational approaches and now with the latest advances or you know or the legis the legislation and the and and the uh extension of Obamacare across the country and and there's now a larger greater incentive to utilize the massive amount of data that is a that is being accumulated in in the Health Care system and our partner institutions lifespan and Care New England are going to be introducing uh electronic medical records lifespan has already done that as of April they went Bing Big Bang and all their all their clinical facilities are now using uh epic the system called epic and uh which technologically I'm told is not not terribly advanced but it's it's now become the standard across most of the country and and about half of the academic medical centers in in the country are using this system so there's going to be a huge store of electronic medical records that in theory should be accessible and in in in in in uh one and could be the subject for uh not only improvements in healthcare delivery but also leading to new hypothesis about uh uh disease causation and and and and and and highlighting potential scientific questions that are that our basic science faculty can can address as well so we're excited about these two hires they're gonna they're starting in uh July 1. they both received their PHD there so it's a husband and wife team and they receive their phds a number of years ago at the at Columbia in the uh in their PHD program in biomedical informatics and and they've spent the last few years uh running the center of biomedical informatics at the University of Vermont medical school so we're are looking forward to that they're also going to be very much engaged in the Rhode Island quality Institute which is a Statewide effort and it's a health information exchange and I don't know how many of you know that about half just close to half the people in Rhode Island have signed up to share their electronic medical records for for research purposes and for and for emergency care so that these 440 000 people if if they any of them require Emergency Care the attending physician in the state of Rhode Island any in any at any facility can go online and look up their their medical records look up their diagnose their lab tests their MRIs their medical their their pharmaceutical you know their medication history through CVS so CVS is a partner with us and so this is the only health information Exchange in the country where it's the entire state is is uh is is involved in this and uh recently I heard that Connecticut had a similar effort that was funded a number of years ago through the health information technology program and they've basically given up and some of the counties in northwest Connecticut are now asking our Rhode Island quality Institute to step in and take over and uh and and run their efforts so one of our I know I'm going on a little bit on on this and this uh you know sort of the big data in healthcare but I think this is very much in line with the with the underlying theme of this week in terms of the importance of computation and and I think these electronic medical records that are that are going to be available to probably more uh uh you know with some variability across the country but I think Rhode Island has some unique characteristics to it that will allow us to take more much more advantage of it and I think we have trained a lot of the expertise in this area as well uh there's a new Department of biomedical informatics at Harvard Medical School that's going to be starting July 1 when it's going to have five actually the inaugural chair of that department is someone named Zach Kahani Zach kohani was an undergraduate here at Brown and did a double concentration in biology and in computer science and then he went on to do his PhD elsewhere I got his uh uh MD and PhD elsewhere and then his heading above the Harvard Medical School effort in biomedical informatics a week from tomorrow at the on the commencement weekend one this I think the Ruth Sauber so-called Ruth Sauber lecture will be given at around nine o'clock in the morning next Saturday a week from Saturday as part of the commencement weekend and the speaker will be a tool Butte a tool Butte is an MD PhD he was a computer science major here at Brown and then he was part of the plme program which means that he was part of that eight-year program which led to an MD from the from the medical school here and most recently he was leading the biomedical informatics effort at Stanford up until about three months ago when when UCSF recruited him away from Stanford to head up the entire University of California biomedical informatics system so he's going to be speaking here a week from Saturday as part of his reunion Class Medical class so Brown has certainly been noteworthy in terms of making major contributions this area in terms of trainings undergraduates and we really look forward to the next stage and bringing in faculty expertise now in biomedical informatics and we have a large Grant application that we've submitted through the division which was very uh uh a multi-disciplinary and integrative it's uh it's a proposal for to set up a Center for Clinical and translational research and uh it's a it's a 20 million dollar award if we're so fortunate as to receive it over five years and one of the key elements of it is to enhance this the the composition and strength of a center for biomedical informatics and to build build a really strong Bridge with the Rhode Island quality Institute so that we would actually have an employee within the the firewall of the Rhode Island quality Institute who would have access to all that data so um I think that's about everything I wanted to say in terms of doing some plugs for what uh for what division of biology and medicine has been contributing over the years and very supportive of efforts in in enhancing the various applications of computational approaches to scientific advances and we really appreciate Soarin and in this team for organizing this Symposium uh I've I've been to each of the two previous ones and they've been exciting and uh and and this one has uh it also continued in that in that tradition and I really look you know look forward to the to the next one now that you know we we're about to finish this one but uh I look forward to the speakers exciting speakers this morning and uh you know pass it over to back over to Barry thanks Ed um okay on to the brain and Neuroscience and I've asked Jim McElwain to introduce our first Speaker Jim is uh colleague of mine professor emeritus of Neuroscience former Fox chair in Ophthalmology and vision sciences and uh and a close colleague of David Burson thank you Barry it is attempting to think about the eye as developing from a single Ur cell that flowers into the organ of site that we know but that's not how it happens the eye is actually constructed a symbol uh during a Developmental and elaborate developmental process that brings together tissues from different uh genealogies and one of these tissues is the neural tube a piece of the brain the developing brain that is as it were exiled banished to the periphery to live inside this organ that sits in the orbit uh this bit of brain comes out there and it forms eventually several things including the retina which is the topic of this morning's talk uh the retina this this piece of the brain that's been exiled to the periphery uh stays in touch with the brain that's stuck inside the skull by virtue of some Pathways that are occupied by what are called retinal ganglion cells you're going to hear much more about them as we along the retinal ganglion cells have been the subject that has occupied Dr David Burson who is our speaker and his team they have made fundamental contributions to our understanding of what these cells do how many types there are with their physiology is their connections and so forth and some of these uh contributions are in fact groundbreaking even revolutionary the team and Dr Burson have been recognized widely for this and I just will mention two recent examples of this last year David received the Brian boycott award named for one of the Premier retinal neuroanatomists of the 20th century this is uh given by the fascib and for work that has advanced our knowledge of the retina and its role in vision just this year he received the free friedlander freelanders friedenwolf sorry friedenwald award from the association for research and vision and Ophthalmology for outstanding contributions outstanding research that contributes to the basic our clinical knowledge related to Ophthalmology so given that I'm sure you want to hear what Dr Burson has to say about the brain that is in your eye so it gives me great pleasure to introduce my colleague David Burson we've got the livelier um what Jim didn't said say was that I was a postdoc in his lab for three years and I would not be standing here today if it weren't for the training that he gave me so I'm deeply grateful to him for not just the introduction but for all of the uh instruction and inspiration that he's given me back then and through the years since he still comes to my lab meetings which is a great Boom for my students as I'm sure they'll they'll tell you um okay so uh as uh Jim has said uh the retina is actually a piece of the brain and here's this developmental process that he was describing here is the early embryo and on this lateral side here you can see the beginning of the optic cup forming and if we were to take a slice through this embryonic uh neural tube this developing nervous system you can see that in fact the retina which is right here is in direct continuity with the brain in fact it is a piece of the brain that has been pushed out into the periphery and uh this in a sense is our window on the world but in fact the world can look back in through the front of the eye and see what's going on uh at least in this one piece of the central nervous system this piece of the brain so even with simple um uh optimal optimal off the ophthalmoscope images of the the retina through the front of the eye you can actually see the blood vessel pattern and so forth but you're actually looking at the surface of a piece of um of brain uh there are many Imaging methods that have been devised to allow you to see much more detail than uh is possible with the simple ophthalmoscope and one of these is a method known as OCT it's actually a method that allows you to see the layers of the retinol have much more to say about the layers as we go along but in addition there are methods now for visualizing individual cells within the living human retina this is known as Adaptive Optics Imaging and what you can actually see is the cone Mosaic the three different kinds of cones arrayed out on the surface of the uh the retina and their their distribution cone type by cone type so this is a very convenient um a piece of the the brain to be working with as it's it's visible even in intact organism now the real payoff of course is if you begin to remove the uh retina from the eye and and put it in a dish now you're in a position to actually manipulate the cells there in um multiple ways that allow you to piece together the circuitry of the retina and the processing that's going on through its various cellular layers so the challenge here is to think about what's uh in store in the next 250 years in in my case in vision science and that's a tough order um but I think there's uh maybe a way to go about this which is to roll the clock back to 150 years and see where we were back then see all the ways in which things have changed in the intervening years and then maybe try to do an extrapolation out to the Future to see what where we might end up uh so um fortunately already 250 years ago 1764 we knew that this uh Theory the emission theory of vision was wrong that something came out of the eye some form of energy came out and played across the surface of substances in the world and told us somehow magically what was out there something like the x-ray vision of of Superman so this was an idea promulgated by Euclid and a lot of the other classical uh natural philosophers and it really wasn't until the Advent of Arabic and Islamic science that it became clear that that was really not the way that it worked at all and in fact it was photons entering the eye of course they didn't know about photons but light entering the eye and somehow interacting with our bodies that under underlay Vision now in this case you're looking at al-hassan who is one of the Arabic natural philosophers who figured out that this was in fact what was going on and really Al Hassan's Insight came from his own interest in Optics because he was studying the Stars he was looking up at the sky understanding how Optics worked and you recognize that the eye itself was an optical instrument and that this must somehow create an image of the visual world inside our our eye now uh alhasan knew that the image should be upside down if it was a Standard Optical instrument and that bothered him so he actually kind of put his finger on the scale and flipped it back over again not quite knowing how it worked this didn't seem to bother Kepler who was another early astronomer who understood a lot about Optics and was interested in the way the eye worked and basically what Kepler said was yeah it's upside down on the retina but somehow we figured it out anyway it's not really a problem and I think in that he's absolutely right so here's just a an image of this inverted representation of the external World on the uh the retina of course the brain figures this out uh so we knew about retinal images um and uh Descartes was really the one who pushed the idea that it was actually the tissue at the back of the eye that was the key sensor here and uh Descartes view was that this sensory representation was somehow carried forward through the fibers of the optic nerve the connection between the eye and the Brain to allow the soul to somehow appreciate the nature of the external world so in uh Descartes idea the projections LED from the retina back into the brain and he was a little fuzzy on exactly where these terminations were he actually was thinking that they might be the ventricles of the brain but the key Insight that he made was that the projection the the carrying forward into the brain of the retinal image was re-instantiated in a new representation and so you can see this retino topic or topographic mapping from I back to the brain in Descartes View and this is obviously a very key Insight so the earliest idea of representations in the brain of the visual World dates back as far as Descartes and this is pretty much all we knew as of 250 years ago so everything that we've learned since is obviously uh huge and if the process that we're looking at now the advancement of techniques and approaches different uh Fields bringing their expertise to bear on this problem continues to accelerate as we think it has to this point it's going to be really hard to imagine where it will be 250 years from now so just by way of sort of introducing the work that I'll tell you about what's going on in my lab let me um use some of the key things that we've inferred in the last 250 years as a way of bringing up to speed on the the key parts of the story here so the first is that the retina is not simply a a sensor it's actually got some internal structure and this was obvious already to Trevor honest in 1830 or so and you can see with his earliest histological um studies of the retina in this case from the crow that the retina is a highly laminar structure different elements at different levels and this is key to understanding its organization now for trevoranas the idea was that light would come in and strike the inner surface of the retina and that would be the only logical place to put the sensory elements that would detect light itself and that somehow these sensory elements would then carry fibers back into the optic nerve into the brain now he got a lot of things wrong here uh the optic fibers actually run over the surface of the retina and as you'll hear in a moment the sensory receptors that respond to light are actually deep in the retinas far from the incoming light as they could be although as you'll hear later in the talk Trevor honest was right for the wrong reasons in at least one respect that is that there are some photoreceptors in the inner retina the recognition that uh photoreception the absorption of photons and the conversion into a neural signal actually occurred in the outer retina as far from the incoming light as possible it was made by Heinrich Mueller in about 1850 or so and what Mueller saw when he took fresh slices of frog retina was that there was some color out in the outer retina that he's surmised might be the pigment that absorbs the photons and leads to this photosensory process he Then followed this up with an extremely clever experiment in which he visualized the shadows of his own retinal blood vessels as he moved a bright light around in the periphery of his visual field and he worked out through simple trigonometry the apparent angular extent of the motion in the visual world for a given angular movement of the um The Illuminating source of light and worked out that the depth of the photoreceptors had to be far enough away from the blood vessels that it had to be in the outer retina where we now know the rods and and Cones lie now a major uh breakthrough in understanding retinal organization came with the Advent of methods for staining individual retinal neurons this is the Golgi method there were several other methods as well but this is the one that really uh broke the game wide open and the scientists who really deserves the greatest credit for exploiting this method to allow us to understand the fundamental organization of retinal circuits is Santiago Romani kahal a Spanish uh neuroscientist who eventually won the Nobel Prize along with Golgi who is the man who devised the method that that Kahala used and one of the key understandings of course of kahal was that the retina was made up of individual neural elements connected together by synapses that is that we're not dealing with the Syne of cells a a connected network but rather with individual elements units that are communicating through what we now know to be both chemical and electrical synapses so from kahal's work we worked out we began to understand the basic canonical retinal circuit from photoreceptors which you see at the top the rods and cones as we call them now through an intermediate neuron the bipolar cell and then eventually to the ganglion cell sitting at the innermost layer of the retina and these are the output cells so these are the ones that give rise to those axons that run across the surface of the retina and then eventually punch back through the eye itself to form the optic nerve running back to the brain so this is the canonical circuit there are several other interneurons here but the thing to recognize is that there are two layers where these synapses occur and we'll be focusing in some detail at the inner of these two the ones between bipolar cells and ganglion cells and the modulation that occurs at that level okay so uh now I need to try to introduce for you a fundamental physiological concept which is the notion of the receptive field so if you imagine one of these bipolar cells which is carrying the signals from the photoreceptors to the ganglion cells you have to understand that the processes from which it receives its input from these first order photoreceptors extend widely enough in the retina that more than one photoreceptor talks to that bipolar cell there's a region of the retina which when activated by light will make this bipolar cell respond we call that the receptive field and this can be defined in retinal space here how far across the retina can you be and still activate your bipolar cell but one could also as easily Define it in visual field coordinates because of this isomorphism between the visual world and its projection onto the flat retina so any given neuron in the retina or for that matter elsewhere in the visual system of the brain will have a recipe to feel an excited excitable region now the receptive fields of bipolar cells and higher order cells in the retina are actually not quite as simple as shown here and this is due to the interaction with inhibitory interneurons at several levels of the several levels of the of the retina one of which is shown here the horizontal cell so this is an inhibitory neuron that spreads laterally in the retina and can cause a bipolar cell driven by excitatory input from its own receptive field Center to actually suppress that response when there's stimulation out here beside the the central region so we now have an antagonistic Center surround organization which actually is a very clever design for a contrast enhancing transformation so just to give you a quick sense of that this is the work of kefir Hartline and Steve kufler back in the uh 30s through the 50s this kind of lateral inhibition centers around organization ensures that these output neurons are actually conveying not critical information about the light intensity but in fact are doing kind of a spatial derivative that they're looking for changes in Illumination in different regions of the visual scene so if you scan a like dark contrast motor like this with Center surround antagonistic units the output of the cell actually ends up giving you sort of an enhanced version of this so that you can really if you imagine the brain looking at the array of these cells the brain would be much better able to detect the contrast boundary than it would where the cells without their receptive field antagonistic surrounds so you can see yourself the perceptual consequences of this if you look at an image like this and at least under the right lighting conditions you may see a kind of scalloping in the luminance distribution there in fact that's entirely happening in your own visual system if I just isolate this third bar from the left by itself you can see that when it's not up against the other bars that you don't see that scalloping so essentially what's happening is these contrast enhancing elements in your visual system are lying to you in a sense about the number of photons coming from that region of the visual field in order to allow you to see that boundary between the stripes more clearly it also accounts for the funny little dim spots that you may see at the intersections between the the white Grid in this figure as well and I won't spell that out for you but it's basically the same phenomenon okay so the emphasis here is on the output of the retina what the what the retina is telling the brain and when we think about that we need to be talking about the ganglion Solstice because I said before these are the cells whose axons make up the optic nerve and already at the time of Kahala in in the late 1800s it was clear that these ganglion cells were incredibly diverse so if you look carefully you can see that there are many cells stratifying near the top of this synaptic layer in which they get their input from the bipolar cells some towards the bottom some spread widely across the retina some more narrowly and kahal was aware that these were fundamentally distinct types of neurons although they shared the property of communicating with the brain although he had very little to say about what might be different in terms of the function of these cells uh here's another view of this this is now taken from the macaque retina from the work of Dennis dacey and again you can see the cells schematically shown in side view here as kahal demonstrated it and you can see the wide range of different stratification levels some are stratifying more than one level some only in one um and uh you can also see maybe at least those of you close enough to the screen that the the branching pattern of the cells looked at on Foss also varies widely some are sparsely branching some much more densely so now the reason the stratification matters so much is that the lower order retinal neurons that interact with these ganglion cells themselves make synaptic contacts at different levels within this plexiform layer this synaptic layer so here are the bipolar cells at the top and you can see that different bipolar cells drop their axons down to different levels this suggests that there are different channels conveying information from the rods and cones to different subsets of ganglion cells by different populations and bipolar cells if diversity is even greater when you look at the other class of inhibitory neurons the amricin cells these are neurons of the inner retinal like the ganglion cells but they come in a huge variety of different forms the guess now is somewhere on the order of 40 in a typical mammalian retina so obviously these synaptic interactions are complex and any individual ganglion cell has the potential to look at a modified version of the low order retinal image as viewed by the photoreceptors through these complex synaptic networks so we anticipate that there'll be a lot more functional diversity than simply the kind of Center surround organization that you've you've seen so far the other thing that I wanted to mention about the anatomy of the ganglion cells is that they tend to form these very regular Mosaic like arrangements so that the locations of the cell bodies tend to be non-randomally distributed they tend to be separated from one another by a characteristic characteristics distance and then the the dendritic Arbors spreading across the surface of the retina tend to have a a Arrangement at which they're essentially covering the entire retina efficiently and in many cases are essentially tiling uh efficiently covering that surface and what this means then is that any given ganglion cell that is viewing the world in a particular way and I'll have something to say about what I mean by that in a minute uh every one of these classes of ganglion cells is seeing every place in the visual field so we have multiple representations of the visual world embodied in the array of ganglion cells of different types so what about function so one of the key distinctions between a ganglion cells has to do with receptive field size that is how much of the visual world does any one ganglion cells see in other words what is its window on the world right it doesn't see the whole visual World it sees just the small region uh related to its dendritic Arbor and the photoreceptors that that excited so we have large ones we have small ones another fundamental distinction is that some ganglion cells are excited by increments in illumination others are inhibited and still others are excited by either increments or decrements so these are the on off and on off type ganglion cells we'll have more to say about what builds that distinction a bit later um there are differences in temporal coding as well some cells respond only when there's a transient change in illumination so for a step in in light the cell will briefly Spike and then shut up other cells carry a consistent representation of the level of light intensity tonic or phasic cells and then there are a bunch of very surprising um highly specific features that you see among different ganglion cells some for example might be selective for the direction of motion of a contrast boundary across the receptive field others might have a very high rate of resting discharge and only be suppressed by activation of the receptive field some are orientation tune there are cells that are wavelength sensitive that is their color opponent and so on so we have a huge variety of functional types that presumably map onto these morphological distinctions and then when we trace the projections of these ganglion cells back to the brain through the optic nerve we now recognize that there are dozens of distinct Targets in the brain um these clearly have different functional roles to play uh in some cases these are very well established in many of these smaller nuclei not so much as known but it's clear that some are involved in oculomotor functions still others in high level uh perceptual uh functions in regulation of um sleep and circadian rhythms and so on so the the this is more or less where we stand now we understand that there are distinct ganglion cell types but we don't really fully understand how many there are um we know that ultimately we'll be able to relate particular ganglion cell types to particular Target structures and to understand their physiology in terms of what they're doing for the visual system but our ability is um at the moment is somewhat limited and we haven't really gotten as far down this road as we would like to um we don't understand really all of the synaptic circuits that underlie the specific response Properties or receptive field properties that individual ganglion cells have we're beginning to reconstruct these circuits in some cases pieces of the story are understood but much still to be done and ultimately of course you'd like to understand the circuit in totality you'd like to be able to understand each transformation that occurs at every synaptic level uh and that clearly is is beyond us at this point for for most ganglion cell types so this is really the focus of my own uh research program so if we're going to go about this if we're going to try to relate ganglion cell types their physiology and morphology and projections to functional roles how do we go about this and I think for the most part the way the field has approached this is to describe as many properties for as many ganglion cells as possible and to try to think about what those properties might be good for that's a perfectly appropriate sort of bottom-up Brute Force approach that I think can be very valuable and I've engaged in it in a major way in my own research program but I think there's an advantage to reversing the questions kind of flipping it on its head and saying let's just start from the point of view of what aspects of vision are going to allow us to survive in the world and reproduce and carry on our genes to the Next Generation that is taking essentially a darwinian approach to this what is it that animals use to survive in the world find their mates and and reproduce successfully what what is uh sort of um evolutionary Fitness in the context of the organization of the the visual system um and in the context of sort of vision and uh Evolution it's useful to revisit something that Darwin himself said about uh the eye so he was blown away by the gorgeous complexity and articulation of uh this biological Optical instrument and he confessed as you see in this quote here that uh although he was obviously a big believer in his own theory that if there were anything that would give him the Willies and make him worry that maybe he had it wrong it would be that the eye was so beautiful that it almost had to be designed okay so I think this this is very much an anatomical Viewpoint and it's very much thinking about the eye as an optical instrument but I think it can be extended to the retina as an evolved biological structure designed to help animals survive in the world and I think it can be extended in to the brain I think it can also be carried over to a thinking about uh physiology about circuits and circuit functions now one obviously we would ultimately like to understand the physiological basis of visual perception and many people are working on this but this turns out to be a tough problem and I owe Jim McElwain a lot for pointing out to me the tremendous utility of working instead in an aspect of the visual system that is more cut and dried hardwired and you might say reflexive where the relationship between sensory input and motor output is more straightforward and where one might have more tools to be able to try to unpack this whole thing to relate this to specific neurobiological phenomena and of course when we're dealing with reflexes the great grand did you all have to sort of bow down to in this role is Sir Charles sharington who worked on the spinal cord back uh in the early part of the 20th century working on spinal reflexes sharington himself had really very little to say about the visual system but what he said about spinal reflexes some of which is listed here I think really does relate very well to this attempt to try to bring reflexology to a study of the visual system and it's sort of integrated organization so he said that reflexes these have been paraphrasing here large the reflexes are dedicated circuits so you need to start with the sensory input ultimately you're going to have a motor output but somewhere in between you've got some set of interneurons or some modulation there's some integration going on there and that when you're thinking about a reflex circuit you want to think about the whole thing start to finish he said that reflexes are purposive in a darwinian sense you know this is kind of interesting because sharington comes at a time only 50 years after Darwin's theory is promulgated and after a long period where many physiologists were sort of waving their hands about um the purposes of of uh behaviors and reflexes and so forth and thinking about these in teleological terms that made a lot of people very nervous so this is in a sense a um a radical proposal of sharington at the time he's saying really because we're thinking now about evolutionary Fitness it's actually quite appropriate to think about the quote purpose not as if there's some designer up there you know choosing how things will work because we need to get something accomplished but because if you evolve a way of solving an important problem for the animal you're more likely to survive so it's appropriate to think about purposes and in fact reflex action cannot really be intelligible to the physiologist until she knows its aim okay so this is a worthy Pursuit here so the question is how do we go about thinking about this in the context of division sharington didn't really help us out here the people who did I think were the neuroethologists because what they were studying were very characteristic instinctive reactions of animals to visual input that could be studied reproduced and ultimately they felt had some sort of biological underpinning some genetic and ultimately neurobiological underpinning so here's one good example of this some of your I can see Smiles there so some of you seen this video before so I watch look at what the frog is doing here it's orienting it's seeing this black object moving against an otherwise stable ground orienting and and leaping so this is this is food for the animal right this is how it survives this is a useful visual reflex for keeping it alive but be careful how you use it um so uh this kind of reflex response uh was well known to two important physiologists who were studying ganglion cells back in the 1950s Jerry let fin at the bottom left here and Horace Barlow at the top right they were recording from frog optic nerve fibers they're recording these Spike outputs of individual ganglion cells and they found a variety of cells so this is the early reflection of functional diversity of retinal ganglion cells and uh one particular kind of cell that they found had a characteristic receptive field it's a little different from the one that I showed you the receptive field Center was an on off center so either increments or decrements in light would cause the cell to fire but around that was a region that would suppress the activity from the center if simultaneously illuminated so a small spot flashed on or off in the center or moving within the center was good but if you had Global motion for example if the whole animal moved and the visual World swept across the retina this is not going to activate this cell very well so you could think about this as a object motion detector or a local detector of a stimulus in the world this was was distinguishable from motion produced by the Animal's own behavior so I won't go through the quote here but both letfin and Barlow felt that this might be viewed as a Fly Detector or um something like that a bug detector this is the terms that are thrown around and this may be pushing things a bit far but in the context of something that's hardwired like this I think it makes a certain amount of of sense uh so letvin actually was interested also in how this worked at the level of the brain and here you see him lecturing in the 1950s and what he's pointing at here is an image of the Frog's brain stem which I'm clearing up for you here a little bit and what he's pointing to is this green structure in the dorsal surface of the midbrain which is the optic tectum or the superior colliculus which is the major Target of retinal output in the Frog so these centers in the in the midbrain are receiving direct input from the retina including from these funny a bug detector cells and of course what we know now is that in the superior colliculus which is a major Target of direct retinal input there is a systematic topographic mapping of the sort that Descartes had already envisioned so single points on the retina project their axon to single points within the tactile map and ultimately then you have a very precisely organized retina topic representation of events in the visual world world well where do we go from here well it turns out that in the deeper layers of the tectum you have neurons that are connected to centers of the brain stem that are responsible for uh constricting the eye muscles the extraocular muscles so this is a way of redirecting gaze and depending on the organism it may also involve head movements or whole body orientation so there's a motor map down here in the deeper layers and this is systematically organized so that neurons in one particular place in the map will fire before reorientation to a particular region of visual space now this superficial layer visual map and the deep deep layer pre-motor map are in alignment so that a given stimulus will activate a patch of The Superficial visual map which sits above just those cells which need to be active for the cell for for the organism to make the appropriate orienting movement to now align its gaze or its body towards that external object so you can think about this as a fixation or a gaze shifting machine and this works for many sensory modalities Beyond Vision as well but in the context of the present a conversation this is essentially the way in which the so-called lug detectors projecting through the tectum or leading to that orienting movement that you saw the Frog make a minute ago now this is the one point at which I can make a link to John Von Neumann because leadfin was part of a very interesting uh crew of characters working at MIT at this time including Walter Pitts and Warren McCulloch who were as all of you know uh the four runners for all of computational Neuroscience and neural network modeling and they were building the simplest earliest forms of neural net at this time and trying to understand what they might do for visual processing now vonorian was very excited about all of this but he was also somewhat skeptical about how far you could really take this and the quote that I extracted from a letter that he wrote to Norbert weiner who's another character obviously who was in in the mix here was responding to McCulloch and pitts's work was that the situation is rather worse than better than before meaning before the first neural network models McCulloch and Pitts have demonstrated that anything and everything can be done by an appropriate neural mechanism and by neural mechanism here what they mean is a modeled computationally modeled neural mechanism so their their models could compute that was clear but nothing we may know or learn about the functioning of the organism can give without microscopic psychological work any clues regarding the further details of the neural mechanism in other words you can build a neural net that will do something but if you really want to understand how the brain does it you got to get your hands dirty so I take a certain amount of satisfaction in that since that's how I spent my whole career okay so that was a digression let your let me remind you that I've talked about one visual reflex so far that links a particular class of ganglion cells through a certain Pathway to an appropriate motor output um now for that reflex to work which obviously requires information about uh local object motion as distinct from Global motion image quality is essential right so if you've got a blurry image the system is not going to work because any motion is going to cause some amount of uh movement over a wide retinal area in the cells that I showed you before will simply shut down so image quality is important and there's a second visual reflex that helps to ensure image quality and that's the pupillary light reflex so uh what you know of course is that as light levels increase your pupil stops down and uh the advantage of this of course is that it increases your depth of field this is like the f-stop on a camera as you increase the f-stop things that are not perfectly in Focus are reasonably in good Focus whereas if you've got an open pupil like in the top images things that are in Focus essentially only at the focal plane in front and behind that things get very fuzzy so the the operation of the bug detector system requires image quality and that requires a pupillary light reflex the reason you need a reflex like this is that of course as you stop the pupil down you're letting less light into your eye so if you're in dim light conditions you stop your pupil down now you won't have the sensitivity you need to see things right but in bright light conditions you can afford to stop the people down very efficiently so there's a sweet spot for any given lighting condition where you have optimal image quality and sensitivity so the reflex does this for you but how does it work what does it need to know in order to generate the appropriate pupillary constriction obviously it needs to know how many photons are entering the eye per unit time this is what we call irradiance it's light energy striking the retina okay so the more light energy there is the more the pupil stops down and this is a a reflex that's mediated through the the brain stem in response to input from from the retina now it turns out that the pupillary light reflex is really only one of quite a number of reflexes or behavioral responses hard right of physiological responses that are related to the absolute light intensity as distinct from contrast boundaries within the visual field movement pattern or whatever okay so among these are the entrainment of your biological clock to the day night cycle I'll have more to say about that in a minute the amount of melatonin circulating in your your bloodstream sleep weight cycles alertness and of course very bright light can be aversive so we blink people who have migraines are photophobic so it's it's painful for people and so on all of these relate not to the structure of the image not to any object information or or pattern or color but rather to simple light intensity now back around the 1990s it became clear from the work of Russell Foster and his colleagues that in animals with severe degeneration of the outer retina where the rod and Cone photoreceptors live many of these photic responses to Bright Light persisted now if the eyes were removed that fell away so it clearly suggested that somewhere in the visual system there might be another photoreceptor and maybe the most spectacular demonstration of this came from work on the Circadian system so let me just sort of walk you through this for a minute what you're looking at is a plot of activity over a 24-hour period actually it's probably more like a 48-hour period so an animal is in a dark light cycle Dawn and dusk in the laboratory environment essentially by turning on and off the lights and what you can see in this animal which is actually a mouse so it's nocturnal is that most of the activity marked by these little vertical tick marks occurs during the dark period that's when the animal is active searching for food in the light period it's inactive largely sleeping okay so the there's a very close close constraint of the activity to the day night cycle now if you plunge the animal into constant Darkness at this point so you're going through days here now in the dark what you see is that there's still a consolidation of the activity pattern and it still has about a 24-hour cycle but it's not quite exactly 24 hours it's drifting a little bit it's like 23.5 so it drifts a little bit every day so this is the free running circadian rhythm there's still a rhythm that's generated by a part of your hypothalamus known as the suprachiasmatic nucleus but the exposure of the animal to light cleans up the clock it takes it from 23.5 hours a day to exactly 24 hours so now the animal is entrained we say now if you take the animal while it's in constant darkness and you pulse it with light at an appropriate circadian phase what you'll see is this jog in the onset of the activity so that's another indication of the effect of light on the circadian clock how does this affect come to be well it turns out that there's a direct projection from the retina to the essentially clock chip in your brain the little nucleus in the hypothalamus that actually serves as the Circadian pacemaker we know it's the pacemaker because if you take this structure out of the brain and put it in a dish it's still cycling at 24 hours it doesn't need any other input it's actually a molecular biological feedback circuit that's causing this thing to oscillate with a period of about 24 hours so it's this retinal input the retinal hypothalamic tract to the hypothalamus that is responsible for this entrainment for this perfect alignment of the activity rhythm with the external world now what I've told you is that in animals with severe degeneration of the outer retina this process still works if you take the eyes out it doesn't so this suggests that although we've thought that the only photoreceptors in the eye were the rods and cones in the outer retina there has to be something else and a key breakthrough came in the year 2000 with the discovery of a novel photo pigment Gene it was named melanovsen because it was first found in the dermal melanophores of frogs but it was later shown to be present also in uh the retinas of mammals and in mammals it's found only in a very small population of cells in the inner retina in the layer of the ganglion cells suggesting that this might have been the missing photoreceptors this is what has inspired Us in 2002 to go after the cells that were feeding the suprachiasmatic nucleus this clock chip region of the hypothalamus and we did that by injecting retrograde Tracer into this region of the brain in in rodents and tracing back to the cells of origin of the retinal input to this clock region and then targeting those cells for for recording and just to make a long story very short the answer is that in fact those cells are directly photosensitive you can see that in the top trays you can see the spiking evoked under conditions where all retinal synapses have been blocked okay so this little step here is the increment of light and you see this causes a depolarization in the cell and it's it's spiking away in the bottom Trace you're looking actually at recordings made from a cell that has been isolated from the rest of the retina so we take enzymes and dissociate the living retina and record just from the red labeled ganglion cell that innervated the clock uh and sure enough that cell will give you a nice inward current and excitatory current uh in in response to light so this is a third retinal photo receptor but oddly it's a ganglion cell that also talks directly um to the brain now these cells have some interesting features that um suit them very well to their role in detection of light intensity and regulation of the circadian clock the first is that they're unusually sustained in their response to a step of light so most ganglion cells shown in red will adapt rather quickly and this is because most ganglion cells are interested in change either over space or time in light level by contrast these intrinsically photosensitive retinal ganglion cells that's what the iprgc acronym refers to there do adapt a bit but they really hold their response for very long periods of time in fact a form of postdoc in the lab has shown that they'll hang up there for at least 10 hours when it comes to encoding the number of photons raining down on the retina per unit time most ganglion cells are extremely poor so on the x-axis here you have increasing intensity and the schematized ganglion cell is really not showing much change in its firing rate uh over time these funny photoreceptive ganglion cells by contrast monotonically increase their firing rate as a function of light intensity they are encoding the very feature that we suggest is necessary for these irradiance driven responses and then finally whereas most ganglion cells have this Center surround organization that I described earlier which is reflected by a sort of inverted U function as you increase spot size you get more and more response out of the cell until it fills the receptive field Center and then as you increase the spot beyond that to encroach on the surround region you now begin to suppress the cell's response so there's an optimal size for the spot to activate the cell in the case of these ganglion cell photoreceptors you essentially see no evidence at all of surround antagonism and this is part of a general strategy these cells seem to operate under which is to is to essentially summon photons over space and time they're not differencing they're not interested in contrast and structure in the scene but rather in just the number of photons how do they get to be so good at this what what makes them different well one of the things that clearly is different is that their whole photo transduction um biochemistry follows a logic very different from that in the rods and cons and instead looks much more like what you see in the photoreceptors of insects and mollusks and so forth the invertebrate of photoreceptors and I won't have time to go through all the ways in which these are similar but the gangly itself photoreceptors are shown in the right column and they line up with the so-called rhabdomeric or invertebrate like photoreceptors in many many uh respects as shown here and not at all with the so-called ciliary photoreceptors the rods and the cones so in particular one thing that may be relevant here is this issue of pigment by stability so the the rod and Cone pigments are composed of a protein that binds a vitamin a derivative and when that photopigment absorbs light it causes a configurational change that leads to the light response in the photoreceptor but it also bleaches the pigment so you get a dissociation of the chromophore the vitamin a derivative from the protein that doesn't happen in rhabdumeric or insect-like photoreceptors and apparently it doesn't happen in the melanopsin that is the photo pigment of iprgcs either so this suggests that under constant right illumination the bi-stable pigment is constantly flipping back and forth between an excitable State and an excited state and it doesn't require ancillary tissues or slow enzymatic processes to reconstruct the photopigment is happening in real time all the time because of absorption of of light that's clearly part of the story that happens at the level of synaptic inputs to these cells so what you've seen already is that the canonical visual circuit is leading from classical photoreceptors through a synapse onto the bipolar cell and then from the bipolar cell to the gangly itself this is occurring in the so-called inner plexiform layer the major synaptic layer where bipolar cells convey their signals to ganglion cells now these funny ganglion cell photoreceptors like other ganglion cells have their dendrites in a synaptic layer so they are likely to be getting synaptic inputs and in fact we can show that that's true by eliminating the direct photo response through the melanopsin process or by simply dropping the light intensity so this is an intensity series brightest light at the top dimmest light at the bottom at dim light intensities the melanopsin system is not actually engaged at all the threshold is quite high for that system so at low light intensities you're essentially looking at synaptic drive to the same ganglion cell through the rod and Cone to bipolar circuitry so even though the cell can respond to lead on its own it actually does also blend signals from the classical photoreceptors now there's something peculiar about the contacts between the bipolar cells and this new funny ganglion cell and that has to do with the sign of the response of the ganglion cell to light so we know that there are basically two kinds of ganglion cells on cells and off cells on cells respond to increase the light off to to decrements and light and this is a reflection of the bipolar cells that drive them this is occurring in two different layers of the synaptic Zone the off cells send their dendrites to the top and get input from bipolar cells that are in are excited by dimming or inhibited by light and the on-center cells are stratifying here and they're getting input from a class of bipolar cells that has the opposite sign of response so essentially you have a segregation of these on and off channels now what you can see here is this gangly install photoreceptor is stratifying in what should be the off sub layer it should be getting inputs from the off bipolar cells and yet you can see that it's an on response it's excited by light so it's completely breaking the rules that we all thought uh applied to the on and off channels of the retina so it occurred to us that one possibility was that these cells are in fact receiving inputs from on bipolar cells but it synapses in a place that they really shouldn't be making any in the upper part of the plexiform layer and in fact that proves to be true I won't have time to show you much of the data but this clearly is part of what's what's happening here so I'm just going to show you now an image taken from some electron microscopic work that we've been doing on such synapses and in this image what you're looking at is a green bipolar Axon passing by the dendrite of one of these melanopsin cells and right here you see this very strong synaptic specialization each of these arrows is pointing to what's called a synaptic ribbon which is a structure that is really highly specialized for continuous fast release of synaptic vesicles onto the postsynaptic Target and we suspect that this very unusual synaptic Arrangement is a part of what makes the melanopsin ganglion cells so good at irradiance coating um one other peculiar aspect of this Arrangement is shown here so if we blow up the schematized synapse between the conventional bipolar cell and its ganglion cell Target what you can see is an arrangement in which you have two postsynaptic partners here's the synaptic ribbon these are synaptic vesicles and here's the ganglion cell that we're interested in but in addition there's a second partner this is a process of an amocrine cell one of these inhibitory interneurons and it's known that at these diad synapses the inhibitory interneuron driven by the bipolar cell is feeding both back onto the bipolar cell and forward onto the ganglion cell and what this means then is that you have the potential for temporal filtering essentially if you had a sustained Drive shown in green here from the bipolar cell to the ganglion cell but then you also use that same drive to drive the amoricron cell and that that amrican cell is now inhibiting either pre or postsynaptically this circuit through to the ganglion cell then what will happen if you sum this delayed inhibitory sustained input with the direct excitatory input is some kind of transient response which then sort of renormalizes so this is a temporal filter a high pass temporal filter you could say a change sensitive Circuit by contrast what we have found and others have have reported as well is that these peculiar synapses in the uh upper part of the synaptic layer these uh these funny uh synapses I showed you in the EM the electron microscopy a minute ago are what are called monad synapses there's only one postsynaptic Target in this case the melanops and ganglion cell and the absence of this inhibitory circuitry May mean that the sustained Drive can pass all the way through the circuit without the temporal filtering that is typical of other ganglion cells okay third reflex um and this baby that was challenging this is a little hard to watch but just see if you can see any uh butterflies flying through the air here anyone see anything okay let's do this again with image stabilization right so the handheld camera is now see the butterfly there it's the same images you saw before but the handheld camera was causing so much jostling there's so much Motion in your visual field that it actually made it difficult for you to detect the motion of this uh White butterfly going by so image stabilization turns out to be an extremely important feature of um visual system visual motor systems and you can see this in all kinds of species and this is one of the things you can do in your backyard if you're one of those Urban Chicken razors but basically in this case the animal is detecting the motion and uh using its head to stabilize the visual scene for us this is usually handled instead in response to self-motion that involves head rotation so as I move my head uh to the right my eyes are counter rotating to the left by the same same angular amount and this stabilizes the image on my retina so that I can detect motion out in the visual world most of this is being handled by the vestibular apparatus so you probably know the vestibular apparatus detects head rotation through the semicircular canals there are three fluid-filled tubes and these detect essentially the relative motion of the Bony canal and the fluid inside so if the head snaps briskly to one side the fluid stays behind while the bone moves the semicircular canals transduce this and use this to drive a visual reflex that stabilizes the the eyes so here's the basic certain structure of one of these canals the best axis for rotation of course is the axis that passes through the plane in which the canal sits there are three canals each with its own axis so if we take the canals away and just look at the axes these are the Cardinal axes of the vestibular apparatus three mutually orthogonal um axis of best rotation for activating these three systems so any arbitrary rotation of the head of course can be unpacked by the brain as the relative activation of these three channels now this system works great as long as the head is moving quickly so on the bottom axis you have head velocity the faster the head moves the better the system works so the gain of the vestibule ocular reflexes which is what this is called the amount of Icon or rotation for a given head rotation is a function of head velocity so at high head velocities pretty good as the head rotates more slowly the fluid does not move with the canal I'm sorry the fluid moves with the canal so you don't get any shearing Force you get no relative motion there and the system breaks down what that means then is you're going to end up with residual slippage of the visual scene across the retina that's bad for perception okay so fortunately the retina itself detects this motion essentially the error the shortfall of the vestibular apparatus results in a visual motion signal it's a global visual motion signal retinal slip that can be used by the nervous system to correct for this so if you fix an animal's head so it can't move its head at all the vestibular apparatus is not in the picture and then you sweep bars across its visual field as shown here you can see that the eye tracks slowly and then resets this is a nystagmus it's exactly the kind of thing that you get with vestibular drive it's the same motor apparatus but now what's being what's driving this is a visual input which is the what's left over from the slop of the inadequacies of the vestibular apparatus now it turns out that the tuning of these visually sensitive elements in the eye is complementary to that of the vestibular apparatus it works best for slow velocities of visual motion and because these two are converging onto a common motor output frame the sum of these two together give you essentially Perfect Image stabilization across the full range of head rotation velocities okay so how does this thing work out in terms of uh synaptic networks in the in the brain well we have the vestibular apparatus driving the vestibular ocular reflex the VOR through its connections to the vestibular cerebellum and brain stem and from there we get access to the oculomotor plan and the extraocular muscles for eye counter rotation and image stabilization what we know is that there is a visual analog optokinetic nystagmus that I'm just showing you and that the key site for the source of this visual information to the vestibular cerebellum is a part of the brain stem called the accessory optic system the details don't matter it's a brain stem Center that receives direct retinal input and one would imagine it's carrying signals about retinal slip This Global motion of the visual scene across the the retina now in the 1980s a guy named Jerry Simpson working at NYU made an absolutely brilliant Insight which was that the organization of visual motion signals in the accessory optic system was what he called a visual system in vestibular coordinates what he meant by that is that the visual motion representation in the system was designed to detect rotation of the head around the same axes used by the three semicircular canals okay what does that look like in terms of a visual receptive field so here's a rabbit sitting in an imaginary visual world and what Simpson found in the accessory Optics system were cells that responded to visual Motion in many regions of the visual field this could be a whole Hemi field where it was sensitive or it could be even panoramic both eyes but the direction of motion preferred varied according to where it was tested so in this case you preferred up in the front part of the visual field and down the back part of the visual field okay so you can see quickly that the best stimulus for activating this cell would be rotation of the visual world around an axis now the amazing thing that Simpson showed was that the the axes that emerge from an analysis of complex visual motion receptor Fields like this were identical to the ones used by the by the vestibular apparatus this is a vestibulocentric rotational motion visual system here it's essentially a visual system designed to speak vestibularis okay so Simpson found cells not like just like the ones I showed you but cells that matched the or preferred the optic flow arrangements for rotation on each of the three semicircular canals okay so we have a visual analog for each of the three semicircular canals here now how is this system built well what's very clear is that the inputs to the system come from a subset of retinal ganglion cells that are selective for visual motion uh and are responding only at light increments these are the so-called on Direction selective ganglion cells and classic studies suggested that they came in three flavors cells that like upward motion downward and slightly backward motion or forward motion okay three channels sounds sort of like the three vestibular channels but we actually have a bit of an issue here if you think about how this thing might be constructed if you project those directions of motion preferred by individual cells out onto the animal's visual world it would look something like this okay so either liking upward downward and backward or forward motion three types each one has a relatively small receptor field is looking at a small part of the visual field and arrayed across the whole visual world it would look like this we're trying to build a cell that likes this okay it likes rotatory motion around some axis pointing out at you this way okay one of the vestibular axes well let's overlay these things how could this work well perhaps a neuron in the brain that has one of these rotationally sensitive receptive fields is selectively sampling from different populations of Direction selective cells in different regions of the retina weighting the relative inputs from these different channels so as to reiterate reconstitute this rotationally um sensitive receptor field that could work uh but we had a radical alternative hypothesis which is that the earlier studies had gotten it wrong because they'd only really coordinated one place in the brain that instead it might be that one type of Direction selective ganglion salt in the retina actually adjusts its directional preference as a function of where it sits in the visual field that is where it sits on the retina so as to always match the optic flow that would be generated by rotation of the head around one of the semicircular canals why would that be useful well if you imagine a neuron in the accessory optic system that gets convergent input from all of those cells across the entire retina what you end up with is this kind of receptive field okay so the key thing here is the suggestion that the earlier argument that the directional preferences of individual ganglion cells is fixed it's the same directional preference regardless of where we record a soul of that type on the retina that instead what you have is a population of neurons that in the ensemble have a directional preference that matches the rotatory field but any individual cell might have a very different directional preference for one of its compatriots that sits in a different region of the retina essentially the computation by this analysis would be done in the retina itself who's here postdoc in the lab has been testing this crazy hypothesis using calcium Imaging in Mouse retina and what you're looking at here is a plot on a flattened Mouse retina of all of the cells he's recorded so far it's a heroic task it's like 500 cells or something like that and remember that the classic model suggests that there are three directions of preference now if you look at those arrows I challenge you to find those three directions in there it's a mess right it doesn't look anything like the classic Model in fact if we reproduce What the classic folks have done and sort of align all of these different directional preferences at a single point what you see is every possible direction is represented and if anything what you see is maybe a weak suggestion of four preferred directions so it's quite different now this is enough to I think throw the old model out but is it enough to tell us that this flow matching idea is right maybe maybe not uh there's a lot of evidence on this sure I could tell you about it if you want to hear more detail I'm going to just show you one line of evidence and the idea is this if you want to figure out whether this complex representation it has embedded within it multiple flow matching sort of rotatory cell populations if the underlying logic here is rotational then what you ought to be able to do is start with this distribution of cells pick an arbitrary axis of rotation and reconstruct the optic flow that would occur if they had rotated around that axis so that's what's being shown here in the blue lines we've got flow fields on the flattened retina as they would occur if they had rotated around an arbitrary axis this Center of rotation corresponds to the projection of that axis out into the visual world okay now the question is how well do the directional preferences of individual cells we've recorded in the retina align with those flow Fields right and we can assess that by looking at the distribution of difference angles between the observed preferences of the individual cells and the prediction for a particular flow field so that's shown here and you can make a histogram of these difference angles and if you've got a pretty good alignment you might expect a piling up of observations here and we see that in this particular case and maybe see fewer at other orientations right so we for any given axis where we're testing the the fit of the the actual observed directional preferences to the flow fields we can assign an index index of goodness of fit and and essentially have one point in a heat map of a mapping of the goodness of Finn across the retina now we have to redo this for another axis and yet another axis and do this for axes separated by 10 degrees in Azimuth and elevation and then take a look at the heat map so what I'm going to show you in the next slide is a heat map generated that way and wrapped back onto a sphere and if there's a rotational structure to this uh directional preference distribution then you should see clear differences hot spots and cold spots and this is what you see so clearly there's a lot of structure here and it's very hard to get your mind around what you're looking at here I understand that but the struct the the the big gaps between hot and cold here and the numbers of hot spots are precisely what you would expect if in fact our hypothesis was correct that is if the directional preferences of these Direction selective gang lead cells feeding this image stabilization Network are actually already in vestibular Centric coordinates they're matching the flow fields that are generated by rotation of the head around a particular semicircular Canal so you just have to take take it on faith if you can't work through the logic here as I imagine it would be tough to do okay so just to give you one last look at this uh what I'm showing you here are three versions of that heat map overlaid on top of which you see the optic flows that would be associated with rotations around each of the three vestibular semicircular canals and what you should be able to see is that where the pole for each of those rotational systems sits that is where there's a center of rotation you tend to find a hot spot nearby so for this rotational system a hot spot here for this rotational system a hot spot back here a little bit hard to see at this orientation and for this last one it's a little bit weaker but there's some heat in here as well that presumably corresponds to that third canal so what I've tried to do uh by way of sort of pointing uh a way of thinking about how to break down the complexity of the ganglion salt population and the functional specializations of those cells and their projections to the brain we've gone through three different visual reflexes on the left gaze shifts and probably even attention shifts to objects in the world based on real world object motion the system I just talked about the image stabilization system and then on the far right the pupillary light reflexes one example of an irradiance coding system in the table I'm trying to make the point that different ganglion cells are feeding these different uh systems on Direction selective ganglion cells for image stabilization melanopsin cells for the pupillary light reflex and these bug detectors or local Edge detectors are sometimes called for object uh orientation to objects now the spatial Precision of these systems varies dramatically across the the board so for the collicular Gaze shifting system spatial uh Precision is essential um for the pupillary light reflex this uh melanopsin ganglion cell system not at all and fairly High also for the the image stabilization system retina topic organization also varies radically this is kind of hand in glove with spatial Precision so again very highly topographically organized system for uh for gaze shifts to objects uh in the case of the pupillary relaxed massive spatial and temporal conversion so that's that's kind of irrelevant in in that case and then for the image stabilization system a funny kind of hybrid because of course the individual neurons have small receptive fields and prefer direction of motion in one particular location in in the retina But ultimately what you're getting to is a global representation you get these panoramic visual receptive fields that are driven by inputs from neurons of this type across the entire retina that's how you're building the global representation so is that uh precise or imprecise it's a little bit hard to say there are differences in these systems in terms of temporal integration whether they're driven by the on or off pathway I haven't had a chance to talk about it but their dendrites sit at different levels in the interplexiform layer so they're presumably drawing inputs from different classes of bipolar and amerigrant cells and finally their relationship to the ocular motor system is kind of intriguing because the the Gaze shifting system is tapping into a rapid eye movement Network in the oculomotor system very rapid shifts of gaze the image stabilization network is sort of classic slow eye movement system like you would see during smooth Pursuit and even the pupillary light reflex system is linked to the ocular motor system but now it's the intrinsic musculature of the eye in this case the iris muscle okay so we're supposed to think ahead 250 years Leon would like us to know sort of where where we are to be 250 years from now I shudder to even make a guess at this but I think we can point in a couple of of specific directions the first is with uh the Advent of things like serial block face electron microscopy an image of which you saw earlier I think we are very much uh on the verge of being able to fully reconstruct synaptic circuits uh in at least restricted regions of the brain that's today so give us 250 years I don't think there's any question at all that we can completely reconstruct all of the synaptic connections of all of the cells of known type everywhere in the nervous systems of experimental animals and maybe at least in postmortem human material whether we could get that level of specificity in the living brain I have no clue that's a great question so storage is going to be a problem going forward um so this is just an example of what's already been accomplished in Mouse retina you know for a small volume every cell type every every example of every cell of every cell type and its synaptic connection connections have been worked out we have fantastic markers for specific cell types we know to some extent which genes are expressed in which cells that's just we're just seeing the dawn of that now we'll have that in in extraordinary detail I think in a matter of decades for virtually whatever assault class you want in in the mammalian nervous system we have the ability to use genetic methods to kill cells to excite cells to silent cells to introduce specific proteins into specific cells this is all now so 250 years down the line the amount of control that we're going to have over the brains of at least experimental animals is going to be phenomenal and of course we're already getting into the realm now of uh therapeutic approaches that exploit this kind of experimental power so for example we now have animal models that have severe Auto retinal degeneration we can introduce virally proteins into cells that would not normally be photoreceptive turn them into photoreceptors and actually recover visual function in animals that were functionally blind before we can cure single Gene defects in human blind patients now so you know the the future is is really promising I think in 250 years it would be shocking to me if it weren't the case that we could completely cure or at least um ameliorate any kind of functional blindness um through methods like this I think you know just as these days a cataract is no longer a cause for blindness because you can take the Cloudy lens out I think the same thing will be true for degenerative uh rental diseases uh as well um I think what's um even more mind-bending but maybe not beyond the realm of the possible is that with the advanced advances in methods for controlling the activity of neurons and living brains that it may not be simply a therapeuticutic sort of approach that is uh being used for these kinds of things but rather that it may even be elective that it may be that sort of the Advent of wearable Computing is going to quickly get into the realm of implantables and that the most remarkable heads-up display that you could ever design is going to end up being your own visual cortex so that would be my guess for 250 250 years down the line Leon would be that in the end you want a virtual reality system it's going to be the manipulation of your own cortex that's going to be you know creating this experience for you so with that I'll give it a give it up and take some questions if you will

Channel: Brown CS

Views: 78

Keywords:

Id: V6R5PK74DHE

Length: 88min 9sec (5289 seconds)

Published: Fri May 26 2023