Visual 2 Pack

Question 1

How is the general treatement of visual information in the whole brain organized?

Answer 1

In a hierarchical manner:
RGC → LGN → V1 → many other areas that are all interconnected
- Each area has its own organization, retonotopic map

*So a lesion in V1 or under impact all other areas
*Neurons get more and more selective and specialized

Question 2

What are the ventral and dorsal pathways related to?

Answer 2

Ventral (temporal) Pathway: V1 → V2 → V4 → IT → temporal cortex (→ hippocampus) :
- recognition of shapes, colors, objects, face recognition

Dorsal (parietal) pathway: V1 → (V2) → MT → MST → Parietal Cortex
- Primary input via magnocellular pathway from LGN → V1
- Motion perception, control fo eye movements,self-motion, position in space, orientation
- Most neurons respond to moving stimuli, little dependance on the shape, color, texture, etc.
*Where pathway

Question 3

What are the 3 main roles of the extrastriate visual areas?

Answer 3

1) They are more directly involved than V1 in guiding visual perception and behaviour (experience bias, thoughts bias)
2) They are useful for measuring quantities that can’t be meaured with small receptive fields → the higher up, the larger the RF of the neuron is
3) They interact more than V1 with the animal’s cognitive state (goal/motivation driven)

Question 4

How is V2 organized?

Answer 4

*Functional organization
Blob cells (V1) → Thin stripe (V2) → V4 (Ventral)

Non-Blob (V1) → Pale stripes (V2) → V4 (Ventral)
Non-Blob (V1) → Thick stripes (V2) → MT (Dorsal)

Also have direct connections from V1 → MT (Dorsal)

Thin stripes → Colour
Pale stripes → Orientation
Thick stripes → Orientation, Direction

Question 5

What is special in the thin stripes of V2 when visualising it with cytochrome oxidase?

Answer 5

Blob cells in V1 took up the dye and sent it specifically to thin stripes

Question 6

Which cells are responsible for the Illusory Contour illusion?

Answer 6

In this stimulus, the edge that are defined by light intensity are all horizontal → see illusory vertical line
V2 responsible for this processing, V1 cells respond to real edges not illusory ones

*Tested by exposing V1 and V2 cells to single line oriented vertically vs horizontally and the illusion → V2 cells specific to vertical lines would fire for the illusion

Question 7

Why is it useful for V2 cells to process illusory contours in life?
Why is useful for them to process Angles?

Answer 7

Illusory contours → allows for depth perception
Angles → perception of shape and depth
*Angle selectivity obtained by summing up inputs of different orientation selective cells in V1 → allows for more efficient firing (compact coding → less metabolic activity because edges/angles contain more information than line)

Question 8

What specificities/processing characteristics do V2, V3 and MT cells have?

Answer 8

V2:
- Illusory contours processing
- Angle processing

V3:
- Origanization by binocular disparity

MT:
- Motion direction

Question 9

What is the effect of a lesion in V2?

Answer 9

• No problem in detecting color, motion, direction

Monkey has no ability to discriminate textures composed of multiple orientations (can’t see the change in orientation)

Question 10

How is V3 organized?

Answer 10

Contains columns for processing of retinal disparity (binocular disparity)
As move along the columns in V3, gradual change in prefered disparity

Question 11

How is variable direction motion tested for MT selectivity?

Answer 11

With stimulation by variable coherence dot field: dots moving 100% in one direction, 50% in one direction/50% random, 10% coherence, 0% coherence (all random)

Ask the monkey to indicate the direction of dot motion by making eye movement to right or left

Then stimulate (with current through electrode) area MT at some neurons that prefer some specific orientation and redo the test → response is biased to the direction prefered by the artificially stimulated cell

Question 12

What is the effect of a lesion in MT?

Answer 12

Coherence threshold without lesion ~ 95% (noise and 5% coherence)

With lesion → need 100% correlation

Conclusion: MT is required to disciminate motion direction when the stimulus contains noise

Question 13

How does the size of the receptive field change when going higher in the hierarchy of visual neurons in the brain?

Answer 13

The receptive field gets bigger and bigger as neuron integrate information from multiple inputs of lower levels → allow more big picture processing
Ex: MT neurons will be able to resolve the aperture problem, V1 cells can’t resolve (V1 neurons only see motion perpendicular to orientation of an edge)
→ RF of MT are 10x bigger than those of V1

Question 14

How is the MT response to the aperture problem?

Answer 14

*Timing-dependent:
- Early MT cell response → bar orientation (preference) is perpendicular to motion direction (65 - 85ms) → not yet resolved aperture problem
-FINISH

Question 15

Where does MT receive input from?

Answer 15

From layer 4B of V1 and thick stirpes of V2

Question 16

How is MT organized?

Answer 16

Columns for directions selectivity + poorly-defined clustering based on binocular disparity

Question 17

What is the receptive field of MST relative to the one of LGN, V1, MT

Answer 17

LGN < V1 &laquo_space;MT &laquo_space;MST
- At MT and before, the receptive fields are congied to contralateral field only
- MST has a bias for the contralateral side, but its receptive field extends all the way to the ipsilateral side
- Not as much retonotopic organization as lover levels, more functional organization

Question 18

What is the anatomy/organization of MST like?

Answer 18

• MST receives input form MT
• Neurons are clustered based on preference for complex motion stimuli → combination of expansion, contraction, rotation, translation
• Overall organization of MST is columnar, as in V1 and MT, but the clustering is based on selectivity for complex motion stimuli

*Responsible for optice flow
*Useful during navigation
ex: if movement is equal all direction and very rapid, we are about to run into that object

Question 19

What feature are MST neurons tuned for?

Answer 19

Tuned for translation → motion in 1 direction
Tuned for optice flow (which is the combination of translation of rotation and expansion)
*Patterns are bigger than MT receptive field

1 neuron would respond to motion in all directions at the same time, but not to motion in 1 of these direction only → for when take a step forward

Question 20

How would MST neurons be useful during navigation?

Answer 20

• Some neurons specific for equal movement in all direction (straight in front)
• Some for more movement in left direction than in right, but movement in both → heading left (these would be silenced in opposit case when moving towards the right)
• Some to inverse → heading right

Question 21

What is the effect of stimulating artificially neurons from MST? (with electrode)

Answer 21

Microstimulating a column that prefers leftward motion → bias towards leftward heading motion

*Decreases probability of the subject to perceive rightward movment

Question 22

In the ventral pathway, how does receptive field size change when going to higher cortical area?

Answer 22

LGN < V1 &laquo_space;V4 (10x RF of V1) &laquo_space;IT

Question 23

What do V4 neurons respond to?

Answer 23

V4 neurons respond to complex shapes:
- Some are a bit more orientation selective (selective for stripes and their orientation)
- Some are selective for a saddle-ish shape
- Some for circles
- Respond to complex 3D shapes difined by curvature and/or slant
*Neuron response are modulated by attention → to boost signal that is at the visual threshold (if very present, attention won’t change much and if very hard to see, attention doesn’t to much either, more in the middle of the range) → Attentio increases the amplitude of the tunning curve (firing) without change shape (what they respond to)

Question 24

What role of the extrastriated visual areas is the ventral pathway most important in?

Answer 24

They interact more than B1 does with animal’s cognitive state (attention modulated)
*V4 way more than V1

Question 25

What is the effect of a lesion in V4?
What experiment was done to assess this?

Answer 25

V4 impairs the animal’s ability to use attention to filter out irrelevant stimuli

Had a circle with striped grid in the middle with 3 circles (noise) around
Normal → Could see 5˚ incline in the line with noise
V4 lesion → needed 20˚ incline to notice with noise
V4 + TEO → needed 90˚change to notice some change (with noise)
*But without noise, did much better job

Question 26

How does the scaling of attention for neuronal response change from V1 → higher cortical areas?

Answer 26

The attentional enhancement increases as on ascends the cortical hierarchy in both the dorsal and ventral visual pathways

Question 27

What is the main role if IT in visual information processing?
*IT = Inferotemporal cortex

Answer 27

Useful for measuring quantities that can’t be measured with small receptive fields → Face or object complex recognition
- Respond to conjunction of features (need circle for eyes + line for mouth + circle contour to fire for face)
*Test with Jennifer Aniston test (when add Brad → no response)

Question 28

How did the harvard lab try to understand what made IT neurons fire best?

Answer 28

Used AI:
Record from 1 neuron → generate random images → when the image caused higher firing → take information + previous info and try to make an image to would make firing even more intense

Question 29

How is IT organized?

Answer 29

Clustering for compelx shape selectivity in IT

Question 30

How was artificial stimulation used to confirm the selectivity of IT cortex?

Answer 30

Selectivity of IT for Complex Stimuli:
Used dots forming an image and gradually less coherent (Face → Flower) → to establish a continuum for stimuli by adding noise

By stimulating the neurons selective for faces, the probability of seeing a face even when the picture was actually a flow increases significantly (biase)

Question 31

What was discovered on face recognition by MRI of macaque cortex?

Answer 31

The is a large area devoted to face recognition → middle face patch

It can be in a bit different areas, but in the floor of the superior temporal sulcus in macaque

Question 32

What is the importance of STPa in visual information processing?

Answer 32

STP = some cells in anterior superior temporal polysensory area → respond to conjunction of form and motion

Humans are highly sensitive to biological motion → requires conjunction of form and motion (can be seen when defined by very few points)

*Much less response (spikes) to dots corresponding to someone walking backwards because not very physiological compared to movement forward

Question 33

Why are computational models important?

Answer 33

Experiements thell us what is happening in the brain
Computational models try to tell us why and how something happens
(Why do we have Orientation selectivity?)

Question 34

Why are Fourier transforms important? What are they?

Answer 34

Any signal s(t) can be written as the sum of sinusoidal functions of different frequencies + amplitudes

Different orientations = sin(x) vs sin(y)

Question 35

What is a power spectrum?

Answer 35

A power spectrum describes the frequency components of an image (not the phase)
- Distance from origin = spatial frequency (farther = lower frequency)
- Angle in the plance = Orientation of the sine wave
- Different spatial frequencies = different image features

*Can represent Mona Lisa as a Power Spectrum → has lots of low frequency sin waves

Question 36

In Mona Lisa’s face, what is encoded in higher vs lower frequencies? (when isolated from the power spectrum)

Answer 36

Low frequencies → see the smile very well (can also see in peripheral vision/magnocellular)

Higher frequencies → can’t see the smile as well, see the details in the eyes very well (in the fovea/parvocellular)

Question 37

What is Fourier synthesis?

Answer 37

It is the fact that was can build any image by starting with nothing add gradually adding higher and higher frequency wave grating

Question 38

How are Simple cells frequency analyzers?

Answer 38

V1 neurons are tuned to spatial frequency of sinusoidal grating (SPATIAL FILTER) → need to match Orientation + Frequency
- Larger Center/Surround = selective for lower frequency
- Thiner Center/Surround = selective for higher frequency

*Most cells (peak response) at 2cycles/degree frequency (degree = thumb at arm length)

Question 39

What is spatial correlation in images?
What does the power spectrum of natural images look like?

Answer 39

It is a measure of how similar neighbouring pixels are → shows how much redundancy is in the image

• Low spatial frequencies ~ strong local correlation
• High spatial frequencies ~ pixel values change very quickly
  *High redundancy in nature

Power Spectrum of natural images declines with spatial frequency (Less cells are specific for higher frequency)

Question 40

What is information theory? (From Who?)

Answer 40

Claude Shannon
Information Theory states that: The optimal code for sending telephone messages has minimal redundancy.

Reducing redundancy allows efficient transmission of complicated signal
*Phone transmission and optic nerve have the same problem of having to pass too much information → individual neurons can transmit limited amounts of information about input image → need to reduce redundancy to maximize efficiency
*Redundancy reduction

Question 41

What system in visual processing allows for redundancy reduction?

Answer 41

ON-center/OFF-surround RGC → signal strongly when there is a change between center and surround
- When they tried to represent an image with Center/Surround instead of just individual pixels → much less correlation (much less redundancy) + Loss of the lower frequencies on the power spectrum (less signaling for lower frequency)

Question 42

What was the conclusion to the question: What set of filters provide the most efficieny representation of natural images?

Answer 42

Try to represent an image with the least amount of filters (cells) → The Answer ressembled V1 simple cells receptive fields a lot! (brain is very efficient)

Question 43

How does temporal correlation show in natural images?

Answer 43

*Similar to spatial correlation:
Luminance values of individual pixels ususally remain constant over short time intervals (there is also redundancy)
ON/OFF LGN cells are also sensitive to temporal changes → weaker resposne to a long lasting stimuli

Question 44

What is the Troxler fading illusion?

Answer 44

Stare at a red dot in the middle, the grey circle in the periphery will eventually fade as the peripheral magnocellular cells fatigue

*Eye constantly moves a bit around to deal with this quick fatigue problem, but if you completely stop moving you eye, you won’t see anything after a few seconds