LECTURE 6- Visual architecture
1-Why do we study perception?
One of the reasons why we should study perception beyond the fact that we want to understand how we perceive, how we see
things, how things are seen processed by our brain, is that it's a very good stepping stone for understanding the software and
hardware of the brain.
Research in perception is one of the first and the most detailed study of software and hardware, particularly the hardware of
the mind. And therefore, understanding this hardware should allow us to better understand the integrity of the brain or the mind.
2-Visual perception
2.1-Sensors
We have two sensors: left and right retinae.
They have different parts:
• Blind spot: part of the retina which does not have any sense, there are not any cones or rods
in it. It is blind, there are only blood vessels going in there and axons coming out.
• Macula: the most sensitive part of the retina, which contains the fovea. It has the highest
density of sensors, the cones in it, which allows you to see the high precision image that you
have.
When counting the number of cones and rods in your eyes, we have about 126 megapixels in each eye.
That is more than 250 megapixels in the two eyes, much more than what you and the most advanced
cameras nowadays use.
Unfortunately, we only use about 6 megapixels or 6 million cones in each eye during the daytime and 120MP or we have 120
million rods during the night.
We have a more rods than cones, most of our sensors in the eye are rods. But these rods are useless in the daylight, they are
completely saturated. What we really use is only six megapixels in each eye, so even ordinary cell phones are much better
sensors.
How is it possible that you have only six megapixels, but you see a much more complicated and richer world than what a cell
phone does? Fovea, which is the most sensitive part of the retina, has only 200000 cones (0.2 MP)! So, there are very, very
few sensors in our eyes. Yet our vision is very, very good compared to what a cell phone does.
These are pictures of foveae of different people. Each of the cones are marked by the kinds of
wavelengths that they're sensitive to.
The red ones are the L-cones, green ones are M-cones and the blue ones are S-cones.
There's a huge variation across different participants, different observers, but none of them have
any colour blindness.
They are very densely packed. The density of the blue cones is about the same, but the density
of red and green cones are variable across people.
These are all packed into a very small region known as the fovea, so one of the reasons why we do have better vison than a lot
of cell phones are even modern day cameras is because the 20000 cones are packed into a very small region in the fovea, the
density there is very high.
As the graph shows, in most of the retina we don't have a lot of cones, but in
the centre, you have a huge number and then it falls off. In blind spot region
there are no receptors at all.
Whereas the rods are very prominent in just outside the fovea and then it falls
off gradually.
But still, that doesn't explain why we are so much better, because it's easy to
increase the density of pixels in a camera as well. So, what is it that makes us
so good?
,One obvious reason is because it's not the eyes that make us good, but what's behind the eyes that makes it really good. We
seem to have dedicated a huge amount of power to processing whatever we get from the eyes. So, the eyes don't have a lot of
very good sensors, but then whatever it is we get from the eyes is being processed by a huge set of machinery, which is our
brain. 15-20% of the brain, the visual cortex in the occipital cortex, is purely dedicated to vision and does nothing else; and
about 40% is somehow in one way or the other involved in vision.
Your brain weighs about 2-2.5% of your body weight (less than 2kg), but it consumes 25% of the energy that you eat. So, a
quarter of everything that you eat is being used up by your brain even though it only considers 2% of your body weight. And a
large part of whatever the brain is doing is essentially dedicate to vision.
That possible is what allows us to use the small amount of sparse input that we get from the eyes to see this wonderful world.
In a sense, we are constructing a huge and rich word from the sparse input that we get from the retina.
3-Visual pathway
How the input from the external world reaches the brain.
The information in the left visual field is sent through both eyes
but ends up in the right hemisphere. So, the right hemisphere
processes the left information from the left visual field and few
information from the right visual field; and the left hemisphere
processes information from the right visual and few information
of the left visual field although the information is seen by both
eyes.
What happens once the information reaches the brain? How is it that you go
from the electrical impulses that come from the retina to the perceived
conscious world that you actually experience when you are opening your
eyes?
The visual system is organised into very specific regions and each of them
has a certain organisation and a certain function.
Most of these areas are in this occipital cortex and there are some which will
be on the underside of the brain, which is on the inferior temporal cortex.
4-Primary visual cortex or V1 or striate cortex
The first place that gets input from the external world through the eyes and the thalamus is called the primary visual cortex. It
sits around the Calcarine sulcus and receives input from the lateral geniculate nucleus (LGN).
Except in hippocampus, most of the cortex has six layers, included V1.
usually information comes in from the bottom, from LGN and through these layers up to layer 4 and then the information is
process and goes to the top layers and from there, it goes to further regions like V2 and V3 and so on and so forth. It also
receives a lot of feedback from the higher areas.
Most of the processing is very similar in the whole cortex, and so are the cells you find in all of it. They all have different
functions because they are sit in different parts, but use the same kinds of architecture. So, the kind of information processing
pathways are similar across different regions.
So, just like any other part of the cortex, it has different kinds of cells with different sizes in different layers. There are some
horizontal fibres or vertical fibres in different layers and these are all meant to connect, interconnect different kinds of cells so
that they can do all the information processing that is necessary.
, The reason it is called striate cortex is because it has this very dark line here form D to D called
stria, which is a very thick bundle of fibres that is present in certain layers unique to V1.
V1 has a particular organization known as retinotopic layout, which means it is map-like, so
what is present in V1 reflect what is present in retina.
So, let's say you're looking right at the centre of this circle here. We can segment the different
parts of the visual world into different segments and then we can see which parts of V1
process each of these segments.
The upper part of the world is processed by the lower part of V1 and the lower part of the
world is processed by the upper part of V1, so it is basically mirror reversed in your brain
from the top to the bottom.
It does have map-like properties, so the parts that are next to each other are also next to each
other in the visual cortex. Adjacent things in the real world are processed by adjacent neurons
in the brain.
The other property that V1 has is that tinny segments in real world (1 and 2) seem to have big areas, therefore high amounts of
neurons dedicated to process them, and bigger areas in real world seem to be processed by fewer neurons.
This process where what you're looking at is processed in more detail than everything else in the periphery is known as cortical
magnification. The cortex dedicates a lot of neurones to whatever you're looking at a certain time.
Looking at the centre of this feature of this person's face, what V1 does?
Left half of the picture is processed by the right
hemisphere and the other half by the left hemisphere.
Both are mirrored reverse in the visual cortex.
Also there are a lot more neurones processing the face
here compared to the other things. It's not that the
images seem distorted by the brain. It's just many
more neurones are processing face part of the picture
compared to other parts of the picture. This is just an
illustration of what it might look like.
4.1-Primary visual cortex function
It is designed to process the basic building blocks the world is constructed of, which are known as features. Some examples
of the basic features are orientation, colour, spatial frequency, depth and motion. Orientation will be used to understand how
V1 is organised and how different neurons process orientation.
Until 1950-60´s and the work by Hubel and Wiesel how orientation was processed was unknown. Their findings were so
impactful and important that even though there is no prize for psychology, they got a Nobel prize through the field of medicine
and physiology. They revolutionised our understanding of how is it that neurones represent things and brain processes
information.
The researchers listened in to individual nerve cells firing at the anaesthetised cat as they presented it with different visual
images. Initial experiments didn't go well because at the beginning they couldn't make the cells fire at all. They were expecting
responses because that's the kind of stimulate that neurones in the retina and in LGN like. So, they thought the same kind of
stimulus might also activate neurones, but they didn't.
Rather by accident, one day they were shining small spots, either white spots or black spots onto the screen and found a big
black dot seemed to be working in a way that they couldn't understand. It was the process of slipping the piece of glass into the
projector which swept a very faint, precise, narrow line across the retina, and every time they did that, they'd get a response.
But then they did see a long series of experience to characterise how is it that V1 and later regions actually process orientation
addition, which is why they got the Nobel prise.
