Visual perception II: Higher-level vision

Question 1

What is the (main) visual pathway?

Answer 1

The visual pathway refers to the anatomical structures responsible for the conversion of light energy into electrical action potentials that can be interpreted by the brain. It begins at the retina and terminates at the primary visual cortex.

Question 2

What does the main visual pathway consist of?

Answer 2

The visual pathway consists of the retina, optic nerves, optic chiasm, optic tracts, lateral geniculate bodies, optic radiations, and visual cortex.

Question 3

What are two visual pathways?

Answer 3

The visual pathway is segregated into two distinct streams—ventral and dorsal.
The ventral stream (what) processes information about object identity, sampling high-resolution/focal spaces. The dorsal stream (where) either processes information about object location or executes movements under visual control, sampling nearly all of space with reduced foveal bias.

Question 4

How can retina be divided?

Answer 4

The retina can be divided into four quadrants: the upper and lower nasal retina and the upper and lower temporal retina. The nasal retina of the left/right eye and the temporal retina of the right/left eye receives visual input from the left/right visual field. The upper parts of the retina receive visual stimuli from the inferior visual field, the lower part of the retina is receive visual stimuli from the upper visual field.

Question 5

What is lateral geniculate nucleus?

Answer 5

The thalamic nucleus that is the main target of axons of the optic tract.

Question 6

What does LGN do?

Answer 6

The LGN receives visual information transmitted from the retina through the optic nerve in the retinogeniculate pathway, and then sends information on to the primary visual cortex (V1) via the geniculocortical pathway.
Its functions include attentional modulation, temporal decorrelation, and binocular facilitation or suppression via monocular gain modulation.

Question 7

What happens in the optic chiasm?

Answer 7

The nasal fibers of each eye cross the midline to join the temporal fibers of the contralateral eye. The visual input from the right half of the visual field will travel to the left hemisphere of the brain via the left optic tract, while all left visual field information to the right hemisphere of the brain via the right optic tract.

Question 8

How many layers does LGN have? What are they?

Answer 8

LGN has six layers that are distinguished on the response properties (magnocellular vs. parvocellular and contralateral vs. ipsilateral eye)

Question 9

What are the characteristics of magnocells?

Answer 9

Magnocells code for achromatic vision (bright vs. dark), they respond linearly increasing to the intensity of the light, the code for motion vision.

Question 10

What are the characteristics of parvocells?

Answer 10

Parvocells code for chromatic vision (colour contrasts), they respond to more stimulus, not so quick, they code for form vision.

Question 11

What are koniocells (“sand” cells)?

Answer 11

Koniocells are dispersed between the magno- and parvolayers of LGN and code for blue-yellow pathway.

Question 12

What are the principles of opponent colour system?

Answer 12

Colors are determined by the proportional excitation of three cone types (short - blue, medium - green, long - red).
The levels of excitation of each cone type define color space.
To calculate the opponent process tristimulus values from the LMS color space, the cone excitations are compared:
- the luminous opponent channel (L + M, sometimes + rods)
- the red–green opponent channel (L - M)
- the blue–yellow opponent channel (S - (L + M))

Question 13

What types of cells are there in V1?

Answer 13

Simple cells, complex cells, and end-stop cells.

Question 14

What are the characteristics of simple cells?

Answer 14

Simple cells are one-dimensional — they only respond to orientation. They have clearly defined inhibitory and excitatory areas. Their firing rate is directly proportional to the intensity of the stimulus (respond maximally to stimuli that match the preferred frequency).

Question 15

What is Gabor filter?

Answer 15

Gabor filter is a model of simple cell response in V1, a sinusoid multiplied with a Gaussian distribution curve that can be oriented at different angles.
The Gabor filter model is based on the idea that these cells respond selectively to specific orientations and spatial frequencies of visual stimuli. A Gabor filter extracts these specific features from an image and gives a set of complex values (the amplitude and phase of the filtered image) as an output. These squared complex values represent the energy or magnitude of the filtered image — the response of the simple cell to a visual stimulus. This process can be applied to an image multiple times, each with a different orientation and spatial frequency, to simulate the selectivity of simple cells to different visual features.

Question 16

What are the characteristics of complex cells?

Answer 16

Complex cells have no definied inhibitory and excitatory areas and respond to a feature (orientation, movement in particular direction) no matter where in the receptive field it is.

Question 17

What are the characteristics of end-cells?

Answer 17

End-cells (hypercomplex cells) are more of a subclass of simple and complex cells and have properties of both.
They respond to a line of a certain length or to a corner of a larger stimulus, and show reduced or absent response when the line or corner doesn’t reach a certain point or is extended beyond it (=> the outside of the receptive field affects the firing of a cell). They play crucial role in detecting borders of the luminance.

Question 18

What is classical receptive field?

Answer 18

A classical receptive field (CRF) is a small, localized area within area of the visual field where a change in the stimulus will result in a change in the firing rate of the visual cortex neuron.

Question 19

What is non-classical receptive field?

Answer 19

A non-classical receptive field (non-CRF) describes the situation when a neuron in the visual cortex responds to a stimulus that is outside its CRF. The reason that we get these effects is
- the long-range horizontal (lateral) connections between neurons in the V1, and
- top-down connections from higher-order visual areas projecting back to V1

Question 20

What types of non-CRF are there?

Answer 20

The non-CRF can be either suppressive or facilitatory
- A suppressive non-CRF: a stimulus in one location reduces the firing rate of a neuron when a stimulus is presented in another location.
- A facilitatory non-CRF: a stimulus in one location increases the firing rate of a neuron when a stimulus is presented in another location.

cf. lateral inhibition — the activity of a neuron in the visual cortex is influenced by the activity of neurons in its surroundings. It helps to enhance contrast and improve the ability of the visual system to detect edges in the visual scene.

Question 21

What is contextual modulation?

Answer 21

Contextual modulation is the phenomenon where the perceived properties of a stimulus, such as its color, shape, or size, are influenced by the surrounding context or background. So, the same stimulus can be perceived differently depending on the context in which it is presented. It can be either excitatory or inhibitory and can occur both is early stages (V1) and higher stages (V2, V4).

Question 22

What topographical models of the visual cortex are there?

Answer 22

Retinotopic mapping: the visual cortex is organized in a way that maps onto the layout of the retina, with adjacent regions of the visual cortex corresponding to adjacent regions of the retina.

Feature maps: the visual cortex is organized into “maps” (coding for color, orientation, spatial frequency, ocular dominance, space)

Hierarchical models: the visual cortex is organized into a hierarchy of processing stages, with each stage being responsible for a different level of visual processing (from early feature detection to complex object recognition)

The Hubel & Wiesel ice-cube model: the V1 is organized into columns that are sensitive to specific features such as orientation, spatial frequency and direction of motion.

Question 23

What are the higher stages of visual cortex?

Answer 23

The visual cortex is organized into several hierarchical stages, with each stage receiving input from lower stages and processing increasingly complex visual information.

V3: orientation, also motion and depth

MT (Middle Temporal Area)/V5: perception of motion (mind that there can be a difference in [rapid vs. slow motion] perception)

V8 (formerly V4): coloured (vs. non-coloured) vision

LOC (Lateral object recognition complex): recognizing objects and faces.

Question 24

How are effects of lesions in V1 and higher stages different?

Answer 24

Lesions in V1 lead to visual field defects, such as blind spots or blindness in the corresponding area of the visual field that the damaged portion of V1 represents. Lesions in higher stages do not lead to blindness, but rather to losing certain features (eg. perceiving of colour, motion)

Question 25

What do lesions in MT+ lead to?

Answer 25

Lesions of MT+ can lead to cerebral akinetopsia —the inability to perceive (fast) movement (a person sees spared perception of static images instead).

Question 26

What do lesions in V8/V4 lead to?

Answer 26

Lesions in V8/V4 (the color area) can lead to hemi-achromatopsia — the inability to experience color (or perceive it muted) in one of the visual hemifields (not to be confused with color-blindness — abnormalities in the photoreceptor system)

Question 27

What do lesions in LOC lead to?

Answer 27

Lesions in LOC can lead to prosopagnosia (face blindness) or object agnosia.

Question 28

What is the concept of linearity?

Answer 28

We don’t know what the neuron in the visual system has the strongest response to. Most findings were accidental (cf. in the beginning Hubel & Wiesel were using simple stimuli—small spots of light or a black dot on a glass slide projected onto a screen, but once, they accidentally moved the glass slide a little too far, bringing its faint edge into view, and the same neurons showed strong firing). And systematic research of the easiest image would take thousands of years.
=>
The Linear decomposition is a simplification, modeling the neural response to a visual stimulus.
R = w1i1 + w2i2 + … + wnin,
where R is response, w is filter weight (specific feature) at location n, i is intensity (firing rate) at location n.

It works for the simple cells, but not for the complex cells.

Question 29

Why do complex cells violate the assumption of linearity?

Answer 29

Complex cells have a more complex receptive field than simple cells, rather combing the responses of multiple simple cells to extract more complex features from the visual input than responding to a specific spot in the visual field. They have a non-linear response to the visual input, meaning that the output of the cell is not a sum of multiple inputs and is not directly proportional to the input.

This non-linearity allows the visual system to extract more complex features from the visual input, such as edges, contours, and movement, and to perform more sophisticated visual processing tasks such as object recognition.

Question 30

What is the ice-cube model?

Answer 30

It was first proposed by Hubel & Wiesel (Nobel Prize in physiology in 1981, “for their discoveries concerning information processing in the visual system”).
The ice-cube model of the cortex is the topographical model that illustrates how the cortex is divided into two kinds of slabs, one set of ocular dominance (left and right) and one set for orientation.

Question 31

What are the characteristics of retinotopic representation in V1?

Answer 31

• the contralateral visual field is processed in each hemisphere
• retinotopic representation is contralateral in both hemispheres: up is down, down is up, right is left, left is right
• fovea is overrepresented (0.01% of retina, but 8-10% of V1)

Question 32

What are the basics of retinotopic mapping?

Answer 32

The basics of retinotopic mapping are mirror symmetry and polar angle.

Question 33

What is mirror symmetry?

Answer 33

V1 is sandwiched between V2 layers, V2 represents a quarter of a visual field, with upper and lower V2 representing different quarters. V3 (dorsal V3) and VP (ventral posterior) representing lower and upper parts of the visual field, respectively. V1 is normally oriented, V2 is inverted, V3 is normal again etc.

Question 34

What idea is behind polar angle?

Answer 34

• fovea is overpresented in the V1
• the visual field is flipped and separated between two hemispheres
• when you move from the fovea, eccentricity (how close to the center of the visual field a stimulus is) is increasing
• the more eccentric the area is, the less it is presented in the V1

Question 35

What is simple feed-forward model of simple cell selectivity

Answer 35

The feed-forward model assumes that this selectivity arises simply from the arrangement of thalamic inputs to a simple cell. A stimulus oriented along a specfic axis activates all LGN cells simultaneously, generating a large synchronous spiking response.