Chapter 12 Chapter 10 Primate Visual System

"I squinted one eye and looked secretly, and it turned out to be like this..." - Children's games "Seeing" itself is a rather complicated process.So it's no surprise that the visual part of the brain isn't that simple.They are made up of a large primary system, secondary systems, and many higher-level systems.Each system receives input from millions of neurons.These neurons are located in the back of the eye and are called ganglion cells.The primary system is connected to the neocortex via the lateral geniculate body of the thalamus.The secondary system is to project to the aforementioned superior quadrigemlic colliculus.

The general structure of the eye is shown in Figure 38. It has a freely adjustable lens, at least for people under the age of forty-five.There are also pupils that can change the size of the aperture.Under stronger light, the aperture becomes smaller.The lens focuses the image in the field of view onto a thin layer of cells at the back of the eye called the retina.On one of the layers are four different photoreceptors that respond to incoming light quanta.Named after their respective shapes, eg, rods and three types of cones.There are more than a billion rod cells in each eye, which respond to dim light, and there is only one type.Cones, numbering about seven million, respond to bright light and come in three types, each responding to a different range of wavelengths of incoming light.Because of this, we can see different colors.This point has been introduced in Chapter Four.

The first processing step takes place when incoming information passes through the retina, which is, in fact, itself an extremely tiny part of the brain that is easier to study than the neocortex.American physiologist John Dewling called it the window to the brain.It may just be the first step toward a complete understanding of the vertebrate brain.Although its structure may be well worth studying, I still treat it as a "black box" and only describe the relationship between its input and output.The so-called input refers to the light entering the eye, and the output refers to the firing of ganglion cells. ①

Cone cells for photopic or daytime vision are extremely dense near the fovea of ​​the eye.Therefore, we are able to see extremely small details.That's why when you want to see something that interests you, you fixate on it.On the contrary, when you can see an object clearly in the dark, it is precisely because there are many rod cells in the retina. The eye moves in different ways, it can jump or move, called saccades, usually 3-4 times per second.The eyes of primates can track a moving target, which is a process of "smooth tracking".What's incomprehensible is that when you're trying to move your eyes smoothly along a still scene, it's next to impossible, and if you try to do it, your eyes will jump around, You can also do various continuous micro-movements.No matter what method is used to keep the image on the retina completely stable, this visual sensation will still disappear after 1-2 seconds. (This issue will be discussed in more detail in Chapter 15.)

The cells that carry signals from the eyes to the brain are called ganglion cells.Any specific ganglion cell can only respond to the opening and closing of a small light spot at a specific position in the field of view, as shown in Figure 39.Since the lens focuses the point of light near the ganglion on the retina, it must be at that particular location.But it also depends on where the eyes focus. (Just like in a camera, the response of a particular point on the film is related both to its position on the film and to the direction in which the camera is focused.) The area of ​​the field of view that can affect the activity of a single cell is called Feel wild.

In total darkness, ganglion cell firing is often low and irregular.This release is called the background release rate.There is a type of ganglion cell called the ON center type, that is, when a light spot is projected to the center of the receptive field, its firing increases suddenly.Beyond this small center there is a circular extent around them.On this area, if it is also stimulated with a small light spot, the opposite effect will occur.If the light spot falls completely on the annular area, the background emission stops completely.And when the light is turned off, there will be a cluster of pulses (see the left side of Figure 39).

Assuming that spots of light of various sizes are placed on the retina so that their centers lie in the middle region of the cell's receptive field, as we have seen, the cell fires strongly when stimulated with a small spot of light, and the larger the spot's diameter Its response is smaller.When the spot is large enough to cover the center and the ring around it, the cell doesn't fire at all.In other words, the response of the central area of ​​the receptive field is opposite to that of the periphery, which means that any particular ganglion cell has a strong burst of light stimulation in the right place, while a uniform light stimulation of its entire area does not. No response.The retina is to remove part of the redundant information transmitted to the eye.What it transmits to the brain is the information of interest in the field of view, where the light distribution is uneven, and what is to be ignored is the almost constant part.

Another type of cells with the same number as ON central cells is OFF central cells: roughly speaking, they are just opposite to the first type of cells, that is, when the light spot is withdrawn from the center of the receptive field, it will have a strong Release (see the right picture in Figure 39).This accounts for the rather general property of many neurons that they can send these spikes down the axon. A neuron does not generate negative-going spikes.So how do they transmit negative signals? It's not easy to find a fast background firing rate, say 200 Hz, in the thalamus or cortex.If such a cell is present, a positive response is produced by increasing its firing rate to 400 Hz, and a negative response is produced by reducing its firing rate to zero.Often, this neuron is replaced by two other fairly similar classes of neurons that both have a low background firing rate, one that fires when a parameter is increased and the other that responds to a decrease in it.When no stimulus is applied, the neuron usually does not respond, let alone 200 Hz, presumably to conserve energy.

If the brain were to send neural activity that varied sinusoidally at a certain point, then one neuron would fire when the signal was positive, and another neuron would fire when it was negative.But a caveat is that you can't use too simple a mathematical function to describe everything that happens: Moreover, a real neuron often responds to sudden changes in the input with an initial burst of firing.And this temporal firing pattern varies from neuron to neuron, and neurons have not evolved to mathematicians' convenience. The receptive field sizes of ganglion cells are quite different.The receptive field in the central area of ​​the eye is smaller than that in the outer periphery.Ganglion cells are relatively close together, so their receptive fields overlap, and a spot of light on the retina will usually trigger a group of adjacent ganglion cells to fire, even if they fire to different degrees.

There are not just two main types of ganglion cells, ON centers or OFF centers.They actually have many categories, and each category contains its subtypes. In mammals, this classification method is slightly different between species. For macaques, there are two main categories, ①sometimes called M cells and P cells (M cells refer to Magno, meaning large; P cells refer to Parvo, meaning small).The ganglion cells of the human eye are very similar.M cells are larger than P cells anywhere in the retina and also have large receptive fields.They also have thick axons, which allow signals to travel faster.At the same time, the M cell is sensitive to small differences in the distribution of light intensity, so it handles low contrast very well.But their firing rates saturate at high contrast, and they are primarily used to signal changes in the visual scene.

P cells are more numerous and their responses are more linear, ie proportional to input, than most M cells.And they are more interested in details, high contrast and color.For example, the center of the receptive field of P cells responds strongly to green wavelengths, but the peripheral area surrounding the center is more sensitive to red wavelengths.It is for this reason that the center and periphery are sensitive to different colors of light, and P cells can be divided into several subtypes, each of which is sensitive to different color contrasts.Here again, we see that the retina is not just transmitting the raw information that falls on the photoreceptors, it actually begins to process the information in a variety of ways. Ganglion cells mainly include M cells and P cells, each of which has ON-center and OFF-center receptive fields, and they transmit signals to the lateral geniculate body of the thalamus through axons, and then transmit information to the neocortex.Also, the retina also projects signals to the Superior Colliculus, but P cells do not project there, although some M cells and various other non-primary cell types do.The superior colliculus is colorblind due to lack of input from P cells. In most vertebrates, the ganglion cells of the right eye project almost exclusively to the optic tectum of the left brain (roughly equivalent to the mammalian superior colliculus), whereas the opposite is true for the left eye.In primates the projections are more complex.Each eye projects to either side of the brain, but the left-middle side of the brain receives only input related to the right half of the visual field. So, what you see with the fovea of ​​your right eye, is sent to the left lateral geniculate body, then to the left visual cortex, see Figure 40, and also to the left superior quadrigeminal colliculus.Of course, the two hemispheres of a normal brain are connected to each other by several nerve fiber bundles, the largest of which is the corpus callosum.If it were cut off for medical reasons (discussed in Chapter 12), the person's left hemisphere only sees the right part of the visual field, and the right hemisphere only sees the left part of the visual field, which would produce Some very strange results, almost as if there were two people in one head. Let us begin with a brief introduction to the secondary systems that project to the superior colliculus.This is the main visual system of lower vertebrates (such as toads); in mammals, many of its functions have been completed by the neocortex, while the remaining main functions seem to be the control of eye movements, and may also include some visual attention aspect. The superior colliculus is a layered structure with three main layers called upper, middle and lower.The upper layers receive various inputs from the retina, but also from the auditory system and other sensory systems.The various inputs have a rough mapping, although the details of this mapping vary from species to species.The input of the lower layer is more diverse. It is important to note that some neurons in the lower layer connect to the superior colliculus on the opposite side of the brain, a pathway called the intertectal commissure (which survived the split-brain surgery described in Chapter 12).Neurons in the lower layers also connect to neurons in the upper brainstem, which control the activity of muscles that focus on the eyes or neck. What properties do these neurons have?Many cells in the upper layer are selective for movement.In rhesus monkeys they are colour-blind, that is, they have no selectivity for the wavelength of light emitted by humans.They are very interested in faint stimuli, but not very sensitive to the details of the stimulus.They respond instantaneously to changes in light, whether it is lighted on or off.These are probably the keys to the generation of unconscious attention.They send out something like "Attention! Something's in there". Anyone who has given a speech has probably experienced that when a sudden change occurs, for example, a door opens to the speaker's left or right, and all eyes of the audience are turned in that direction at the same time, this immediate reaction is largely above is unconscious.I think the superior colliculus is the main factor that produces these eye movements. How on earth do the eyes know where to jump?That's thanks to ingenious experiments devised by David Sparks, David Robinson, and others.Now we have a better understanding of eye movements.In fact, the upper layer of the superior colliculus may be regarded as the projection of the sensation, and the middle and lower layers correspond to the projection of the motor system.In these regions, the firing of neurons encodes the direction and amplitude of eye changes in order to make the eye follow the target in a hopping manner.The signal is more or less independent of the eye position in the instant before the jump.This signal is sent to the brainstem to determine how large and in what direction the jump needs to be made. The signal doesn't express itself in the way the engineers guessed it would.A neuron might encode a particular jump direction, and its firing rate might encode the jump distance.Thus, in this way, a small collection of neurons can encode all directions and distances.Another way is that each neuron can encode the jump vector, that is, direction and distance.Actually not.To produce a jump, a patch of neurons in the superior colliculus fires rapidly.Broadly speaking, it is the active center of the motion map that determines the jump vector.Such a particular superior colliculus neuron may participate in many very different jumps.It is these activated neurons as a whole that determine the jump vector properties.In short, an eye movement is controlled by many neurons. ① What controls the speed of eye movement?This may be related to the firing rate of neurons in the activated area.The stronger they fire, the faster the eyes move.Thus, the final direction of jumping depends not only on how fast the neurons involved fire, but also on the location of the effective center of this population of active neurons on the motor system localization map. You may find this arrangement unique, but it's an excellent example of how a population of neurons encodes relevant parameters such as the speed and direction of eye movements.Its advantage is that if some neurons are not active, the whole system will not stop working. No engineer can design such a system unless he already understands how the brain works.When these signals reach the brainstem, they must be transmitted in different sets of signals in order to control the muscles of the eyes.Exactly how to do this properly is for further study. Let us now consider the primary visual system projecting to the visual cortex through the lateral geniculate body.The lateral geniculate body is a small part of the thalamus.When I went to the Salk Institute in 1976, I inherited an office overlooking the ocean that belonged to the late Bruno Bronowski (creator of the TV series "The Ascent of Man"), and a double A colorful plastic model of a real brain.What I started with was figuring out where the side knees would go on the model.I had no trouble finding the thalamus, but it took me a while to find a small protrusion labeled lateral geniculate body, but that's not surprising since it's made of only 1.5 million neurons. To understand the lateral knee body, you need to grasp two points. The first point is that it is just a transit station.The second point is the opposite of the first point, and it does a lot of more complicated work that we haven't understood so far. The predominant neurons in the lateral geniculate body are principal cells, which generate excitatory responses.In addition, there is a small subset of suppressor cells that have GABA receptors.The lateral geniculate body is called a staging station for both anatomical and physiological reasons.Chief cells receive input directly from the retina and transmit it axonally to the V1 area of ​​the cortex, with no other neurons on this pathway.Therefore, it is called a "transit station".These axons rarely have collateral connections to other principal cells or to other parts of the lateral geniculate body.In other words, these neurons tended to remain isolated rather than communicate with their peers.In addition, the input from the retina is mapped to the lateral geniculate body such that each layer on the lateral geniculate body slightly distorts the mapping from the field of view.The receptive fields of neurons in the lateral geniculate body are larger than those of retinal cells, and they are very similar.At first glance, the lateral geniculate body simply relays information received by the retina to the visual cortex. The word "MAP" has two slightly different interpretations in the visual system.Its general meaning derives from the fact that those neurons that are not too far apart in the donor connect directly to the termini of axons that are close to each other in the recipient domain.This produces a rough mapping of the donor domain in the receptive domain.More strictly, it refers to "retinal mapping", in which neurons adjacent to each other in a certain visual field tend to respond to the activity of adjacent points on the retina, that is, to move adjacent points on the retina from the visual field The upper 3D information is converted into a 2D projection.As the higher levels of the visual system are explored further, the retinal map becomes increasingly disjointed due to many, many steps of approximation.However, the mapping from one region to the next remains fairly well preserved. The lateral geniculate body of macaques has six layers, as shown in Figure 41, two of which are composed of large cells (called Magnocellular), which receive input from the right or left eye respectively, but have little interaction with each other .And the input is mainly from the M cells of the retina.It is natural to think that the P cells of the retina also project to the other two layers with many small cells (called Parvocellular) in a similar way.However, it does not just have two floors, but a total of four.Their inputs are from the two eyes separately and always remain separate. What is the difference between the large cell layer and the small cell layer?In two laboratories, conscious monkeys trained to perform various visual tasks were subjected to localized small lesions on the lateral geniculate body.These experiments roughly suggest that neurons in the small cell layer mainly carry information about color, texture, shape, and disparity, while neurons in the large cell layer mainly detect moving and blinking objects (see ref. 2). So far we have only discussed excitatory chief cells.Inhibitory cells are mainly divided into two types, which include cells in the lateral geniculate body itself and the reticular nucleus of the thalamus.The reticular nuclei are a thin layer in the thalamus, not to be confused with the reticular formation in the brainstem.This thin layer of cells surrounds most of the thalamus, and the neurons are all inhibitory.They receive excitatory input from axons that pass to and from the neocortex, and they interact with each other.Their output is in turn mapped immediately below them to the part of the thalamus.If the thalamus is regarded as the gate leading to the cortex, then these reticular nuclei are like the guards guarding the gate. Neurons in the lateral geniculate body also receive feedback input from cortical V1 area.Surprisingly, more axons feed back from V1 than ascend to the cortex, but these descending axons synapse with dendrites farther from the cell body.Therefore, their influence will be greatly weakened.The exact function of these reverse connections is unclear (see Chapter 16 for some guesswork about their function). Of course, it also has inputs from the brainstem that modulate the behavior of the thalamus, especially the connections of the reticular nucleus.This means that neurons in the lateral geniculate body are free to transmit visual information when the animal is awake.However, when the animal is in slow-wave sleep, this transmission is blocked. Some neurons and various types of synaptic connections related to the thalamus have been described in some detail here, but the characteristics of the lateral geniculate body should be discussed. Can express the incomprehensible combination of simplicity and complexity. Chief cells in the lateral geniculate body project to the visual cortex (see Figure 40), cat axons can reach several visual areas, but macaque and human axons almost all connect to visual area 1 (1). (In the monkey's cortex, it is weakly connected to other areas, a problem related to Yushi, discussed in Chapter 12.) If all of V1 in a human or monkey is severely damaged, all of his (it's) visual field Half are nearly blind. At first glance, any part of the cortex appears disorganized.There are about 100,000 neurons per square millimeter; axons and dendrites are interlaced, and many supporting glial cells and microvessels are mixed together, completely in a chaotic state.They are not like the neat arrangement of transistors and other structures on a computer chip.If you take a closer look, you will find that some of its structures are ordered.The general arrangement of neurons remains much the same in many different regions of the cortex.Let's first look at what these commonalities actually are. The cerebral cortex is a thin layer whose vertical thickness is much smaller than its length parallel to the layer's surface, and where the arrangement and appearance of neurons is asymmetrical.The direction perpendicular to the surface of this thin layer is called the vertical direction (this is like flattening the cortex on a table).The other two directions are called horizontal directions.For example: almost all pyramidal cells have dendrites that rise vertically to the surface of the cortex.In contrast, cells at the level of the cortex have fairly similar properties to each other.This is somewhat similar to the arrangement of trees in a forest, with the vertical orientation distinctly different from the horizontal orientation. The most striking property of the cortex is that it is layered.It is important to understand these layers and the different functions of the neurons in each layer.For the convenience of description, it can be divided into six layers.In fact, there are several sublayers in the layer, as shown in Figure 42. The uppermost layer is the first layer, which has few cell bodies, mainly composed of pyramidal cells located in the layer below it. It consists of dendrite terminals that extend upwards and the interconnected axons between the terminals.So it's all these neural wirings and very few cell bodies.Below it are 2 and 3 floors, which are often collectively referred to as the upper floor.In these layers there are many pyramidal cells.Layer 4 is composed of many excitatory stellate cells and few pyramidal cells.Its thickness varies considerably in different cortical areas, and in some cortical areas there is almost no such layer.Layers 5 and 6, called the lower layer, contain many pyramidal cells, some of which have dendritic terminals that reach layer 1. Not only are the neurons in different layers quite different, but more importantly, the way these neurons are connected is also very different, as shown in Figure 43. Cells in the upper layers (layers 2 and 3) communicate only with other cortical areas.Although some of these neurons can be connected to cortical areas on the other side of the brain through the corpus callosum, their projections as a whole do not extend beyond the cortical area.Although some neurons in layer 6 have lateral axons that connect to layer 4, among them the neurons mainly project backwards to the thalamus or claustrum, which is a nucleus attached to the cortex located in the subcortex, and to the middle of the brain.Layer 5 is a very special layer of the cortex, only neurons in this layer project completely outside the cortex, that is, they do not project to the thalamus and claustrum, although there are some neurons that project to other cortices Area.So, in a sense, layer 5 sends information that's been processed in the cortex to the rest of the brain and the spinal cord.All these connections away from the cortex, and even back, are excitatory. Of course, the cortex also has many inhibitory cells.However, excitatory pyramidal cells accounted for the majority in number, inhibitory cells using GABA as a neurotransmitter accounted for about one-fifth of the whole, and the rest were mainly spiny stellate cells.The axons of these excitable spiny stellate cells are rather short (about 100-200 μm) and are only able to communicate with horizontally adjacent cells.All suppressor cells share this property, but there are some exceptions. ① One class of suppressive cells does not seem to exist.The axons of pyramidal cells often extend down to regions quite distant from the cortex.Before that, it usually shoots out several branches, which are called side branches.In some cases, these lateral branches in turn form many local forks, and they extend horizontally for a considerable distance, on the order of several millimeters, within the same cortical area. If we think of the cortex as capable of computational functions, it should have a special type of inhibitory synapse like a "gate". It needs to be able to To allow information to leave the cell body through the axon and cycle several times in the cortical area, that is, it needs to realize several cycles of computation. For this, we need a strong inhibitory set of synapses, but it is not at the origin of this axon origin, but just before the axon exits the cortex. Although one theorist needed to construct such a type of synapse for his model to work, there is actually no evidence for their existence. In the axon There is no discovery at each bifurcation point. These show that the cortical area always seems to rush to send information without any circular processing.This also means that when the brain needs to establish a community of activities through repeated iterative operations, the connections between various cortical areas are as important as the connections within a single cortical area. How exactly is information passed between the layers of the cortex?This is an extremely complex issue, however we can gain some insight from the rough block diagram below (see Figure 43). The main, but not the only, access to the cortical area is on its 4th layer.But when it's small or non-existent, just go straight to the lower part of layer 3.Layer 4 is mainly connected to the upper layer 2J, which in turn forms a large local connection with layer 5 until it reaches layer 6 below it.Level 6 in turn returns to level 4 through short vertical links.Layer 1 also receives some major input from other cortical layers.These associate with the dendritic terminals of the upper pyramidal cells from the lower layer. None of the above has addressed the complex nature of the many axonal connections in small patches of cortex, and in particular the surprisingly long many connections from a layer to itself.Obviously, behind all these regularities, there are still some necessary connections.However, until we have a better understanding of the cortex, it is too difficult to spell out these regularities.The neocortex may be the crowning glory of humanity, so it won't let its secrets out easily. Finally, we will talk about the divisions of the brain.Initially, the division of the cortex was based on observing the shape of the stained sections under a high-powered light microscope (this type of academic study is called architecture).The striate cortex is so named because of its prominent horizontal textures that extend horizontally in all directions from the ends of large axons.These striations are large enough to be visually observed in stained microscopic sections, as shown in Figure 44.These textures suddenly disappeared at the edge of a large cortical area.Therefore, it is natural to give such a fairly uniform area a name or a serial number.Other areas of the cortex are slightly different.For example, the striated cortex has a thick layer 4, while the primary motor cortex has little if any.Unfortunately, the differences between adjacent regions are so subtle that neuroanatomists cannot agree on them. In the early 20th century, the German anatomist Korbinian Brodmann divided the cortex of various mammals, including humans, into several different regions and sequenced each region.He called the striated cortex area 17, the area adjacent to it was area 18, and the area adjacent to area 18 was area 19.Label the primary motor cortex as area 4.Other neuroanatomists, such as Oskar and Cecile Vogt, divided the cortex into more regions. ① While Broadman's division is largely correct, it is generally too rough a division.For example, areas 17, 18, and especially 19 are all related to vision.As will be discussed in the next chapter, Region 17 can be viewed as a single region, and Regions 18 and 19 also include many important subregions, so some of these terms are no longer used.Of course, in some medical articles, they still use this division of human cortex. All in all, the primary part of the visual system is highly parallel—many similar but different neurons are active at the same time.The retina at the back of the eye is the front end for processing visual input, sending this information along two major pathways to the cortical pathway to the lateral geniculate body and the superior colliculus involved in eye movement, as well as to several smaller brain stems. Small visual areas that are involved in eye movement and pupillary regulation.Information related to color is transmitted to the lateral geniculate body, but not to the superior colliculus.The information in these initial parts is quite local and simple.For us to see anything, this visual information must be further processed in different areas of the visual system. ① In mammals, there are few, if any, neurons that project from other parts of the brain to the retina. Of course, moving our eyes can affect the firing of retinal neurons. ① There is a third class, sometimes called "W cells," which include a considerable number of neurons and have various properties. ①However, please note that since the required output is only a simple two-dimensional vector, this method cannot be used when a region simultaneously processes more complex information. ① Also known as "striate cortex" and "area 17". ① One exception is a type of inhibitory neuron known as a "basket cell," whose axons extend much longer distances within the cortex, which can be a centimeter or more.When they connect with another neuron, multiple synapses are formed on its soma and nearby dendrites.So they can produce quite strong inhibition in important parts of neurons.Their exact function is not understood, and we have neglected here the function of a well-known suppressor cell called the chandelier cell.Its axons associate only with pyramidal cells and form multiple inhibitory synapses only at the initiation sites of their axons. ① It was Oskar Vogel who cut open and examined Lenin's head, authorized by the Soviet authorities for this purpose.