What do we know about the role of primary visual cortex in vision?
The cerebral cortex is a very thin (2mm thick), highly specialised sheet
of tissue wrapped around the brain. Despite this, cortical volume is
very large due to the many folds or convolutions. In humans the surface
area is about 2000cm2. This area is larger still in dolphins but smaller
in lower mammals such as rodents and minimal in amphibians. The posterior
part of the cortex is called the occipital cortex. Within this region
lie several differentiable visual areas including the primary visual
cortex, which can also be known as visual area 1 (V1).
Light bleaching the photo pigment in rods and cones results in the creation
of a receptor potential which in turn results in a reduction in the
release of the neurotransmitter glutamate. Due to this reduction in
neurotransmitter, the membrane of a bipolar cell depolarises, this new
potential, in turn depolarises the membrane of a ganglion cell, which
increases its rate of firing. The axons of ganglion cells converge to
form the optic nerve and pass visual information to the lateral geniculate
nucleus (LGN). The optic nerves from the two eyes join together at the
base of the brain and form the optic chiasm. At this point axons from
the inner halves of the retina cross over and go up to the LGN in the
opposite hemisphere. This means that both the LGN and the primary visual
cortex receive information from the contra lateral half of the visual
field. The LGN consists of six main layers, each receiving information
from one eye. The inner two layers contain larger cell bodies and are
known as the magnocellular layers. The outer four layers contain smaller
cell bodies and are known as the parvocellular layers. The koniocellular
sub layer is located ventrally to the other layers. This layer is responsible
for transmitting information from short wavelength ('blue') cones to
the primary visual cortex. In turn the cell bodies of the LGN send their
axons through a pathway known as the optic radiations to V1.
V1 consists of six main layers, arranged in parallel to the surface.
This cortical area is also known as the striate area as the layers consist
of the nuclei of cell bodies and dendritic trees which show up as dark
and light bands in stained sections of tissue. The middle layer (4C)
receives information from the magnocellular and parvocellular layers
of the LGN. The information is then passed to different layers to be
analysed. Layers 2 and 3 receive information from the koniocellular
layers, and are also responsible for passing information onto visual
association areas.
Hubel and Weisel carried out pioneering work into cortical cell function,
firstly on neurons in the cortexes of cats and monkeys (60s&70s).
They found that small spots of light, so effective at stimulating the
circular receptive fields of cells in the retina and LGN are much less
effective as stimuli for cortical receptive fields. Cortical cells respond
to more than just light spots, they respond selectively to features.
Cortical cells are capable of analysing more abstract information than
just light intensity and hue. Information in the cortex is combined
in such ways that features can be detected that are larger than the
receptive field of single retinal cells. Most neurons in V1 are sensitive
to the orientation of a bar for example. Hubel and Weisel classified
these cells into three different types. Simple cells have elongated,
opponent receptive fields. For example such a cell could have an excitatory
(ON) central region with inhibitory regions on either side. Each simple
cell is responsive to a very small (10') range of orientations. The
location of the stimulus is as important as it's orientation. Complex
cells are also orientation sensitive but don't have excitatory and inhibitory
regions. These cells also increase their rate of firing if the line
moves perpendicular to the original position. The size of their receptive
fields is larger so they can also detect movement. They also respond
to light bars on dark backgrounds and equally vice versa. Hypercomplex
cells are again orientation detectors but also demonstrate end inhibition;
the cells increase their firing when they detect the end of a line at
a particular orientation.
It appears that the cells in the cortex are somewhat hierarchical. A
complex cell surveys the activity of a group of simple cells that in
turn survey the cells that first enter layer 4C (stellate cells) and
so on down to the retinal ganglion cells. It is these convergent connections
that are thought of as the initial steps in perception. At each successive
level the cells are equipped to analyse more abstract information. The
cells of the primary visual cortex respond to lines and boundaries that
comprise the visual fields of several lower cells.
The cells have been described as feature detectors; each cell will respond
to a slightly different orientation, and will have different excitatory/inhibitory
regions. For example simple or complex cells 'seeing' the sides and
hypercomplex cells seeing the corners would detect a triangle. This
seems intuitively possible but problems are encountered when the idea
is extended; are their feature detectors for every object in the world?
There are not enough cortical cells for them to be this specific, what
if a neuron died, would we lose the ability to recognise a certain object?
Each cortical cell responds to a variety of stimuli for example, bars,
edges, waves, checkerboard patterns, and so on.
Albrecht et al. (1978) found that cortical cells are also 'spatial frequency
analysers'. They discovered that neurons in V1 responded better to sine
wave gratings than to lines or edges. A sine wave grating is designated
by its spatial frequency (its variation in brightness measured in cycles
per degree of visual angle). Different neurons appear to detect different
spatial frequencies (at different orientations). It appears that most
receptive fields are large enough to include between 1.5 and 3.5 cycles
of the grating. Spatial frequency is a valid way of analysing information
as, for example, small objects and large objects with sharp edges provide
high spatial frequency signals whilst low frequencies are signalled
from large areas of light and dark. The most important visual information
comes from low spatial frequencies, as when such information is removed,
shapes are very hard to perceive. On the other hand removing high spatial
frequency information simply reduces the detail, or visual acuity. Harmon
and Julesz (73) constructed a picture of Abraham Lincoln on a computer
using small squares representing the average brightness in specific
portions of the picture. A second picture was created by removing the
high spatial frequencies from the original image, i.e. all the sharp
lines (visual noise). When the visual association cortex receives such
noisy information (the first picture) it finds it hard to perceive the
underlying image.
In 1992 von der Heydt et al. found a new type of cortical cells whilst
studying monkey striate cortex. These neurons appear to respond to 'periodic
patterns'. They don't respond to lines or bars, rather to sine wave
gratings of particular spatial frequency, orientation and frequency.
Von der Heydt claimed that there are approximately 4 million cells of
this type serving just the central 4 degrees of vision in monkey striate
cortex. This type of cell is clearly capable of analysing more complex
information than cells lower in the visual pathway, such as ganglion
cells that can detect sudden changes in brightness.
A further difference in LGN and cortical cells is that the cortex contains
binocular cells. They receive information from both eyes. V1 is the
first point in the visual pathway to combine information from both eyes.
A function of the primary visual cortex is therefore depth perception
involving retinal disparities. Many of these binocular cells especially
in layer 4C (from the magnocellular layers) have response patterns that
seem to aid the perception of depth. These cells respond most strongly
when the same object is seen in slightly different locations by the
two eyes. That is they respond to retinal disparity (images falling
on slightly different parts of the retina), which is used in stereoscopic
vision, indicating differences in the distances of objects. It is suggested
that it is the horizontal disparities of these binocular neurons that
is the neural basis for stereo depth. It is possible that having a collection
of neurons each responding to a single disparity creates this binocular
response. The disparity could be estimated by identifying the neuron
with the largest response. Richards (1971) suggested an alternative;
disparity might be measured by creating a few pools of neurons with
coarse disparity tuning. The pools might respond to objects in slightly
different positions relative to the horopter (arc swept about the viewer
where any object on the arc will fall on corresponding points on the
retinas of the two eyes). To estimate depth the relative responses from
the pools could be compared.
One last main function of the primary visual cortex is colour perception.
Colour sensitive ganglion cells pass information through the parvocellular
and koniocellular layers of the LGN to the primary visual cortex. This
information is transmitted to special cells in cytochrome oxidase (CO)
blobs. These specialised areas of the striate cortex were discovered
by Wong-Riley in 1978 when he found that a stain for cytochrome oxidase
(an enzyme in mitochondria) showed a patchy distribution in the cortex.
Later research with the same stain (e.g. Horton and Hubel 1980) showed
a polka dot pattern of dark columns through layers 2 and 3 and (to a
lesser extent) layers 5 and 6. It has recently been discovered that
the parvocellular system receives and transmits information about the
long and middle wavelengths ('red' and 'green' cones) whilst the koniocellular
system does the same for short wavelengths ('blue' cones). The blob
cells are double opponent, meaning a single cell can receive input from
more than one type of cone cell. For example some cells have centres
that respond in and excitatory way to 'red' cones and a surround that
responds in an inhibitory way to the same cones. In the same cells 'green'
cones have the opposite effect; inhibit in the centre and excite in
the surround. Cells in the blobs have circular receptive fields and
are therefore not responsive to orientation. From V1 the parvocellular-blob
system projects to thin strips in V2 and then onto V4 where there are
colour selective cells. In this way the colour of a stimulus is processed
separately from form and movement until the visual association cortex.
It is mostly agreed that the cortex is organised into modules. Each
module consists of two segments each surrounding a CO blob. Blob neurons
are sensitive to colour and low spatial frequencies but relatively insensitive
to other features for example orientation. Neurons outside the blobs
seem to be the opposite, they are mostly insensitive to colour but do
show sensitivity to orientation, movement, spatial frequency, texture
and binocular disparity. The two segments receive information from different
eyes but the module then combines the information.
The receptive fields of neurons in the same module overlap, demonstrating
that they receive information from the same area of the visual field.
By inserting a microelectrode straight into an interblob region, it
has been found that both simple and complex cells are present but they
respond to the same orientation and share the same ocular dominance
(same percentage of input from the two eyes). Orientation sensitivity
and ocular dominance vary systematically around the interblob region
and are arranged at right angles to each other. Cells are arranged in
columns about 0.5mm across according to their 'orientation preference'.
The connection between simple and complex cells in the same module allow
for the first deconstruction of the visual world into lines of various
orientations, this is a necessary requirement for the visual analysis
of form.
Blasdel (1992) invented an inspired way of mapping the modules in the
striate cortex. Monkeys were operated on to have part of their skull
removed and replaced with a glass window. This window was made so that
it allowed voltage sensitive dye to be injected into the cortex. The
dye was injected and the monkey's shown different visual stimuli designed
to cause responses in neurons sensitive to certain features.
An interesting phenomenon found when the visual cortex is damaged is
blindsight. Damage to earlier stages in the visual pathway results in
blindness in all or part of the visual field. People with blindsight
do not see anything in the blind part of their visual field and yet
they can point at or reach for objects located in this region and can
discriminate their size and orientation. This behaviour appears to occur
with out any conscious awareness of the objects.
The primary visual cortex is an important stage in visual perception
but is not responsible for all the processing. Cortical neurons can
only analyse parts of the visual field at any one time. It is not until
the visual association cortex that these parts are combined. The modules
in the striate cortex respond selectively to colour, orientation, movement,
spatial frequency, texture and retinal disparity to build up the image
of the visual field. This information is then passed to the visual association
cortex, for example colour processing occurs in the CO blobs (after
the P and K systems) and is then transmitted to areas V4 and V8 of the
extrastriate cortex. Damage to these areas abolishes colour constancy
and causes loss of colour vision respectively. The primary visual cortex
is responsible for further analysis of form, movement and colour in
larger areas of the visual field than have been looked at in the retina
and LGN. Form and movement appear to follow a different pathway to the
analysis of colour. The two pathways were also separated in the LGN
in the different layers. This information is combined in the visual
association cortex.
|