Tutorial: Cognitive Neuroscience (27:14)
Date Posted:
August 11, 2018
Date Recorded:
August 11, 2018
CBMM Speaker(s):
Frederico Azevedo All Captioned Videos Brains, Minds and Machines Summer Course 2018
Description:
Frederico Azevedo, MIT
Download the slides (pdf , pptx )
FREDERICO AZEVEDO: Cool. So I'm going to start. My name is Fred Azevedo, as you may know, and my part of the tutorial talk a little bit about cognitive neuroscience. Basically, I'll have to, in 30 minutes, tell you everything that there is about cognitive neuroscience, and if you can, please contact me anytime during the course, or you can just send an email to this address.
So I will start saying to you that there are two classic books of neuroscience that are introductory, but definitely worth it. The first one is Principles of Neural Science, the Kandel one. There is more. And the Gazzaniga is more about cognitive neuroscience, and it's also as big as this one. And it's interesting-- a remark to say that Michael Gazzaniga was the one that coined the term cognitive neuroscience, together with George Miller-- they are cognitive neuropsychologists-- and when they were in the backseat of a taxi in New York.
So basically, and obviously, the term cognitive neuroscience is a combination of the terms cognition, which is the mental action or process of acquiring knowledge and understanding through thought, experience, and the senses, and the term neuroscience, which is the science of the structure and function of the nervous system. Basically, if you want to put it in simple words, it is the science that tries to understand how the brain enables the mind.
And just a little review from anatomy to function-- I'm going to concentrate in the brain. So the brain is composed of cerebellum, brainstem, the encephalon, and cerebral cortex, as it's depicted in the picture. But there is also the amygdala, the basal ganglia, the reticular formation, which are not depicted here.
The cerebral cortex is the part that is a main player in cognitive neuroscience, basically in cognition. And it can be further divided in four to six lobes, depending on the author. So the four common lobes are the frontal lobe, occipital lobe, temporal lobe, and parietal lobe, but you can also find the limbic lobe and the insular lobe. The insular would be here, after you make a hole there.
And the cerebral cortex is composed of layers, between six layers of neurons, and are also organized in a columnar fashion. And it's very useful to take the cerebral cortex in this-- as it is an-- opens it and displays in a flat map, like I'm presenting here. By doing so, you can see that there is interesting properties arrive.
And one of these properties that you can see here is that functional areas, they tend to stick together. So if you put-- if you do a scheme of the functional areas, you can see that the visual areas, they dominate the occipital part. And then you have this-- some other sensory areas, and then the motor areas, and it's color-coded.
And so now, talking a little bit about the core cognitive processes-- I'm going to be very briefly when I'm talking about perception, action, attention, and learning and memory. And also, there are other topics that I'm not going to talk about, but are-- they're as important as the other ones, like language, emotion.
And so it's starting with perception. There is an interesting process-- an interesting property of the brain, which is that, when the neural input arrive in the lower order sensory areas, normally, they dichotomize, and they're sent to the association areas.
For instance, in this case, you have the visual signal, which goes to the parietal association cortex and also the temporal association cortex. And this property of dichotomization of lower eyes-- of those streams in the lower eye-- from lower order visual areas to upper, higher visual areas seems to be a property of the sensory systems. It's not only in the visual system, but also in the somatosensory and auditory system.
And after, from the parietal and temporal association areas, they reach the frontal association cortex. And then the signals, which are processed in this bottom-up manner, they can be, depending on the objective, sent down, modulating the new neural inputs. Normally, the parietal stream is more associated with the processing of spatial features, while the temporal stream is more associated with the identity of the sensory object, and also related features, for instance, the emotional balance.
And the first stream was-- difference, high parts of the dual stream scenario was proposed by Mishkin and Ungerleider in 1982, I think. And they were studying the visual cortex in monkeys, and then they identified the "what" pathway, which is the parietal path, which is the-- sorry-- the "where" pathway, which is the parietal pathway associated with spatial processing. And the "what" pathway would be the temporal pathway associated with object identity and so forth.
So this is a-- Francisco Wright told a little bit about the visual processing, basically, as very simple. The information reached the retina, and then goes through LGN-- is the Lateral Geniculate Nucleus, and then V1, V2, and here you can see the dichotomization that I was talking about. So there is the dorsal pathway and the ventral pathway.
And it's interesting to note that, in lower order sensory areas like V1, you have simple features, the processing of simple features like direction and orientation. And as you go up in the hierarchy, that they're associating more complex features, like complex shapes. And, of course, the visual processing is not that simple, as I was telling.
And here is a diagram made by Felleman and Van Essen in 1991, and, basically, it shows the visual areas at the time that were known-- that are known to be functionally connected when processing visual stimuli.
One consequence of these distributed hierarchical processing is that in the higher-- or high-order areas, you are going to have areas specialized to process very complex objects, for instance, faces. And this is one study that I really like. It was made by Doris Tsao and Freiwald, and here they show the two monkeys images of faces and objects, and then they identify the face patches of the monkey first. And then, you can see here that this area is preferentially activated by the image of faces.
And after that, they insert an electrode exactly in this area, and then they show again the stimuli, and they see that there is an increase and a spike in activity, which is very selective to faces. So if you show faces, you have an increase. But if you see any other kind of category, the spike in activity is not that intense. And in the paper, they claim that the selectivity of neuron was 95%, which is incredibly high.
So another thing is that-- of visual processing, is that context matters. So for instance, here in this figure, it's-- you can see that even though there are just six elements, three Pac-Man style figures and three angles, you interpret these immediately as three spheres and two triangles.
In the second case, you have also just three spheres, but do you think there are three spheres plus one direct line? And if you remove just the impression that there is another line here, you immediately decouple these images. And here also there is one [INAUDIBLE] line, that is also an apparent effect, is an illusion. It's not real there, but our brain immediately interprets that.
And we can also see the neural correlates of such illusions. Here in this experiment, the authors were recording from the receptive field of the neuron, which is this circle here, and then they make this rectangle of light move. And then every time that the border reaches the receptive field, there is an increase in the neural activity.
Now, the interesting part is that they put another rectangle of light here, so there is basically no change in this line, and no change on receptive field is. But when they move this, the overlaid rectangle of light, when the imagined border reached the receptive field, you would also see a significant increase in-- a spike in activity. And when you just showed half of these images, you wouldn't see any kind of activity.
And another illusion that I like a lot is the color illusion. It's-- here, you cannot see very well, but I can show to you, and I'm going to show you that this blue is-- and this yellow are actually the same color. I just put some blank yellow-- white boxes that I show to you, that-- oops-- come on. Yes. So it's hard to believe, but it's true. And similarly, the-- and this happens because the blue and yellow background.
The second one that I like is this-- very probably you have seen that-- is, again, because of the difference-- the apparent difference in brightness, your brain interprets this as gray or blackish, and this as white. But actually, they are the same pattern of gray, the same tone of gray.
So the second thing that I would like to talk about is action. Basically, action is also distributed in a hierarchical way, so the planning of the action, it starts in the premotor and supplementary motor cortex regions.
And then, together with the motor cortex, the cerebellum, and basal ganglia, the plan of action can be transformed in a command to be executed. And then that is passed to the brainstem and the spinal cord, and then the spinal cord innervates the muscles, and then the movement can be really executed.
And nowadays, we understand a lot how this-- how the neural signal is processed during-- in the motor cortex. For instance, it-- some authors, they taught a monkey to move this lever to one of these eight positions. And they showed it to us-- relatively feasible to be called to which direction the monkey was moving the lever, only based on the neural signals. And one directly application of this is brain-machines interfaces.
So in this video, you are going to see that the author implanted a Utah array, which is basically a matrix of electrodes like this one, and here is an image of a similar implantation. And they collect the neural data, and then decode, and pass this to a robotic arm. And basically, the function of the monkey is to move the robotic arm, get the marshmallow, and bring back. And we can see that he's not using anything else, except the signal of the motor cortex.
And nowadays, this technology has been also applied to humans. So there are lots of patients which are already moving arms with their thoughts and using also the Utah array. It's very similar.
AUDIENCE: Sorry. Is he thinking about moving the arm?
FREDERICO AZEVEDO: Yeah, he's thinking and moving that. Now, this is-- the arm is just connected to a computer, and he's-- here, they strategically and conveniently put this thing in front, but it have-- he has an electrode and a cable that you can see here, that's connected up on computer.
AUDIENCE: The experience, how you can see the wires.
FREDERICO AZEVEDO: Yeah, exact.
AUDIENCE: It was also making him do this.
FREDERICO AZEVEDO: Yep.
AUDIENCE: They're training him--
AUDIENCE: Training him.
AUDIENCE: --but, to me, they're adjusting it over there.
FREDERICO AZEVEDO: Hmm, yeah.
AUDIENCE: He's also-- because he's been moving his hands-- like moving his arm now.
FREDERICO AZEVEDO: Yeah, he's-- yeah, you can see that the monkey sometimes moves their hand because it is part of the training. And there is-- when the monkey's also trying to move the arm, sometimes he'll move the hands.
But there is another scientist, Miguel Nicolelis, basically, he did something-- same thing. And then the way that he trained the monkey, after the monkey was implanted, well, that, in the beginning, the monkey was training with a lever and moving an avatar or robotic arm. But then he taught the monkey that he should keep the arms constrained and just use the thought. And then this-- it took-- the same thing happened.
So moving forward, another important cognitive process is attention, which is the process by which organisms select a subset of available information upon which to focus for enhanced processing. Basically, in simpler words, what happens is that there is an enhancement of signal processing of things that you attend to, or space, or locations. And there is an inhibition of processing of the unattended stimuli, in very, very simple terms.
So for instance, I'm going to concentrate in two kinds of attention, which is the spatial attention and the feature-based attention. Again, I am concentrating in the visual system because I don't have time to go for every system, sensory system. But there is also auditory attention and the sensory motor attention, and so forth.
So in the case of spatial attention, for instance, you can fixate here in the middle of the screen, and you can select for any reason-- portion of the space to concentrate your attention to. And in this case is this region that is marked. And the other type of attention that I'm going to talk about is the feature-based attention.
So imagine now that you want to select only the yellow objects in the picture, like this cab and these two signs. What happens, that your brain is going to boost the signal in those regions, and then it's going to be easier-- well, there-- it's going to be easier to detect, let's say, in very simple terms again.
In fact, things are a little bit more complicated than that. For instance, for spatial attention here, is you have an increase in firing rate, but you have also-- while in feature-based attention, you have not only an increase in firing rate, but also you have a sharpening of responses.
And here in this example, what happens, we can see that there is a [INAUDIBLE]-- small spheres that are moving-- small circles that are moving one direction 100% coherently, but when you add noise, basically, you add some randomness in the direction of movement. And then you have-- if this happens, you have a widening of the tuning curve, which is basically the response to a certain direction.
And if you are paying attention to a certain feature, if you want to pay attention to this-- to the direction of the movement, for instance, for the stimulus, with attention, what happens, there is a sharpening of the response. So you are basically removing the apparent noise of the stimuli.
And here is a classical experiment that shows the neural correlates of a spatial attention. And, basically, what the authors did was they taught the monkey to fixate centrally, and then they were recording from a receptive field that, for instance, is here. And then, the stimuli that they showed was a classic Posner stimuli. Basically, the monkey is looking at this fixation point, and then there are two garbled patches that are going to flicker. And at some point, one of the garbled patches is going to rotate, and the monkey has to detect this change of orientation.
And then, you can prove that the monkey-- that there is an attentional effect when the monkey has increased in sensitivity. This is the behavioral curve of the monkey. Basically, it's how easy the monkey can detect this orientation change when there is a-- when the monkey becomes-- when the symbols become easier to detect due to attentional effects. You can see that there is a shift of the curve to the left.
And after that, if you put-- or record it from a neuron-- let's say in V4, like-- as they did, you can find that when the monkeys pay attention to this stimulus, there is an increase in the firing rate. So here in this curve, you can see a contrast of attended and unattended conditions.
This increase in firing rate and a widening of tuning curve is not the only thing that-- are not only-- the only effects of attention. For instance, the same author as before, Marlene Cohen, she showed that attention can improve performance also by reducing noise correlations.
Well, being very fast, noise correlation-- be very fast and explain what noise correlations is. Basically, when you have one neuron-- when they're recording from one neuron, and you present the same stimulus many times, you're not going to have exactly the same response as before. So we have a variability in the response, sometimes that you respond more, sometimes that you respond less. And then when you plot a tuning curve, you take the average of these responses.
Now, you can do this with many cells. In this case, if you take two cells, you can see that when the monkey is not attending, there is a correlation in the variability. And then, you can see that there is a correlation here, and what attention does, it's that this decouples the two cells. And, basically, there is a decrease in this correlation of the noises of the-- the neuronal noise, which are the noise correlations, and you can see a summary of this effect in C.
And also, we can see that there's a change in oscillatory neural synchronization in the brain. For instance, in this work here of Pascal Fries, he showed that there is an increase in gamma synchronization and a decrease in low frequencies. But I don't have time to talk a lot about this. If you want to know more details, please come talk to me later. And I also forgot to tell, this presentation is in our web page, and all my comments that I wanted to do that would be very more calm are there. So I'm doing just a very short summary because of the time.
AUDIENCE: Can I ask a question about the last slide?
FREDERICO AZEVEDO: Yeah, please.
AUDIENCE: So if you'll just slide it back-- well, one more roll past, one slide back. Yeah. so the reason you're getting correlation where you're not attending is just because there's more spreading of firing, and it's very together. It just--
FREDERICO AZEVEDO: Yeah, the way to understand it is that when you are not paying attention, basically, the neurons, they are, let's say, working kind of together. And then, when you are-- you start to attend, is that the neurons, they become-- start to become more independent of each other. So there's a reduction in these correlations.
Basically, to be-- they are working in a similar way. So if you have an increase in-- if you present sometimes a stimulus, and this neuron gives a higher response, the other neurons are also going to give a high response. But if-- and also, the same happens, if it gives a low response, the other one is also going to give a low response. So they're kind of coupled.
AUDIENCE: So that it's because without attention, the tuning curve is just wider, so it's more likely to plot around the same--
FREDERICO AZEVEDO: It is my interpretation, yes.
So being-- now, talking a little bit about memory and learning. So I would like to start with the famous case of Henry Molaison. I don't know if I pronounced it right. And basically, he was a guy that had many epileptic attacks when he was a kid, and then they tried to treat this with drugs, and nothing solved.
So at that time, there was a new technique that was being accepted, which basically was a removal of the focus of epilepsy. And in his case, the epilepsy was so intense that they decided to remove both temporal lobes bilaterally. So they removed basically almost complete, the inferior part of the temporal lobe.
So here is an original picture, an original MRI of his brain. And here is another MRI of a similar procedure, but this one is made just on one side. So you can see that basically almost all the amygdala, hippocampus, is removed, and some parts of the entorhinal cortex also. So you throw everything away on both sides.
And interestingly, what happened to him is that, even though he could still remember the information about his past, like-- I think it was five years, until some time in the past, he could not learn new explicit memory. Basically, there was a nurse that would come every day to talk to him, and he would introduce himself every time and say-- he'd ask, who are you? So he could not learn anything.
And in this picture, we can see more details of the regions that were removed, so 1, 2, 3 are basically these slices of the brain, to see that the-- how big it was, the removal. And so what we get from this is that these areas, the hippocampal area is very important for the formation of new declarative memories.
And I'm talking about declarative memories because you have different types of long-term memories. For instance, being very summarized again, there is the explicit, which is the declarative memories. Basically, the memories are facts and events, things that you can tell verbally. But there have also the non-declarative memory, which is the implicit memory, for instance, the memory of riding a bike, or some emotional response. Then for every region, you have-- every type of memory, it requires a specific configuration of brain areas.
And besides the long-term memory, you have also the short-term memory, which is the work memory, basically, a temporal memory of things that you need to be required to do in the-- in and out, in a short scale of time, like minutes. And, for instance, in this experiment, they taught the monkey to fixate in the center of the screen. And then they would show a sample of the monkey, and the monkey-- they need to memorize the identity of the object, basically, which object was that-- was there, and also the spatial location of the object.
And then, they recorded from neurons of their dorsolateral prefrontal cortex, and then they-- there they found object-selective neurons and location-selective neurons. Basically, the object-selective neurons, they would show an increase in activity because they were-- this is this phase in here, basically, the yellow.
And you can see that there is no object there, but you can still see that the neuron keeps the activity related to the sample. So there is a sustained activity during the delay period, and this happens for the identity of the object, but also for-- in the case of the location-selective neuron for the spatial position.
And a little bit about the learning. So you have also-- there is an experiment that was made from-- by the-- Kandel, the author of this book, that he showed that-- the habituation and sensitization in aplysia. Basically, what he did was, when you have the siphon in depletion, and every time you touch the siphon, the aplysia has a reflex of retracting the gills. It's a protective reflex.
And then, what they did was to touch repeated times the siphon to record from different neurons, and then they showed that there is a habituation, basically, an increase-- there's-- it's changed here. So basically, there is a synaptic depression, which is basically a depletion of the readily releasable vesicles.
So even though the action potential is shown here, it stays the same, the excitatory pulse now-- potential here decreases. And this happens because there is an increase in the number of vesicles that is released. And the opposite happens for sensitization. You can also make-- increase the amount of vesicles that are released by playing with these neuronal-- with the current injections.
And so what I just showed to you was the short-term plasticity, was basically synaptic enhancement and synaptic depression. But you also have long-term plasticity, which is basically-- the idea is the same, but it-- while you are-- the difference when you're playing with the release of vesicles, here you are changing more structural configurations, the cell, for instance.
If for a long-term potentiation, you can change the numbers of upper receptors. Basically, you increase the number of upper receptors, and at the long-term depression, you decrease it. And so by doing so, you have a weakened-- you can weaken or strengthen the synapses.
Wow, I did it in time. So basically, I went very briefly through all the cognitive process-- of some of the cognitive processes. I don't have time to go through language, emotion, but these topics are going to be discussed with more details during the course. There are specific speakers for that. And that's it, and time to save.
[APPLAUSE]
have an interactive transcript feature enabled, which appears below the video when playing. Viewers can search for keywords in the video or click on any word in the transcript to jump to that point in the video. When searching, a dark bar with white vertical lines appears below the video frame. Each white line is an occurrence of the searched term and can be clicked on to jump to that spot in the video.
LEARNING