Vision and spatial cognition : Perception & visual categorisation Flashcards
Central & peripheral retina
Central retina (macula) covers 18-20° around the fovea. It is surrounded by the peripheral retina.
The macula is composed of 3 regions:
- The fovea (2°)
- The parafovéa (2-6°)
- The perifovéa (6-18°)
Fovea composed of 100 000 to 350 000 cones responsible for high spatial resolution.
2 types of photoreceptors
Cones for daily vision
Rods for twilight
5-6 millions cones/retina
100-130 millions rods/retina
Ganglion cells
More than 100 million photoreceptors converge to 1 million ganglion cells.
Alpha cells
Large receptive field.
Sensitive to rapid variations of contrast but do not respond to color.£
→ Magnocellular pathway
Beta cells (midget)
Small receptive field. Activated by a few cones.
Sensitive to color and high spatial resolution.
→ Parvocellular pathway
M pathway
Achromatic
signals rapid changes in luminance in visual field
Phasic, transcient
P pathway
sensitive to color
responds to static stimuli
Tonic, sustained
Retinal projections
Axons of ganglions cells project to
90% LGN
9% superior colliculus
The LGN
In primates the central visual field occupies more than 50% of the LGN
6 layers : 2 layers M pw, 4 layers P pw
Parvo cells sensitive to :
Color
High spatial frequencies (contours)
Low temporal frequencies (slow motion)
Predominant in central vision
Magno cells
Have a low spatial resolution (low acuity).
Do not respond to isoluminant color
Respond strongly to :
high temporal frequencies (motion)
low spatial frequencies
low contrasts (<8%)
LGN : Lesions
Lesions of Parvo layers in the LGN
loss in color vision, in shape perception and in depth perception.
Lesions of Magno layers in the LGN
loss in motion perception.
The visual cortex
More than 30 areas deal with vision in the brain
V1
Over-representation of central vision: more than 50% of V1 for central vision.
Layer 4C in V1 is the input of retinal information
4Cα gets the M neurons of the LGN
4Cβ gets the P neurons of the LGN
2-3 gets the K neurons
5 projects to the Sup Colliculus
6 feed-back to the LGN
Dorsal and ventral streams
2 main channels in V1
[1] predominant Magno: 4Cα inV1 - thick bandwidth in V2 – MT (V5)
→Dorsal pathway
[2] predominant Parvo: 4Cβ in V1 – fine bandwidth in V2 - V4 (formes, couleurs)
→Ventral pathway
M pw arrives 20ms before P pw
→ Preactivation based on low spatial freq - high tempo freq
Spatial properties of the ventral pathway
20-40ms : retina
30-50ms : LGN
40-60ms : V1
orientation/continous
50-70ms : V2
contour neurons, complete missing parts, end-stop cells (end of stimuli)
60-80ms : V4
colors, textures, simple shapes
LOC - 70-90ms : lat occ cx - fusiform gyrus
textures - colors - silouhette
LOC - 80-100ms : Ext tempo cx
categorization / faces
> 100ms Perirhinal
object specific representation (my car, my dog)
Functional properties : Ventral stream
Object, face, word identification and conscious recognition
Lesions:
Object Agnosia
Prosopagnosia
Alexia
Ventral Simultan-agnosia
Functional properties : Dorsal stream
Visuo-motor interactions
Motion perception
Ocular saccades
Spatial attention
Non conscious representations
Lesions
Hemineglect
Optic Ataxia (guide the hand toward an object using visual info)
Apraxia (planificatiuon of mouvement)
Dorsal Simultan-agnosia
Tests magno/parvo
≠ properties ⇒ ≠ stim to check ≠ pathways
Superposition d’image low SF et high SF
Dev and disease
The M/Psegregation exists before the 47th day of gestation
M and P neurons exist only in human and non human primates.
The dorsal stream has a slower development are is more vulnerable to pathology (deficits in M pathway in 4-10% of dyslexia, in schizophrénia, in Alzheimer, in glaucoma, and in multiple sclerosis).
Objects perception
Occipital part of LOC
Tools > motor parietal cx
lesions
LOC: Patient (DF) cannot identify objects (visual agnosia) but she can categorize scenes (Steeves et al 2004 J Cog Neuroscience).
(overlap with FFA)
Face perception
FFA : Fusiform Face Area
inferior temporal cortex & LOC
**OFA : Occ Face Area** physical features (e.g. oritenations, shadows, spatial freq..) \> sends useful info to FFA (invariance to physical properties) \> variable aspects (lips motion, gaze direction) processed in STS
Facial expressions in amygdala
Anterior teporal lobe : categorization
Lesions
loss of object & face recognition
Space perception
PPA : Parahippocampal Place Area (Epstein et al, 98)
Cerebral organization
2 models
The modular representation model, derived from neuroimaging studies, proposes that IT cortex contains discrete patches specialized for individual visual categories, such as faces, body parts, places or nanimate objects.
The distributed representation model, derived primarily from physiological studies in animals, proposes that the visual system is organized to extract generic visual features necessary for object recognition, and that objects are represented by combinations of these features.
More recent studies indicate that object representations in IT cortex are likely organized according to some hierarchical model, incorporating both modular and distributed elements
Visual scene perception
Dissociation between regions involved in place perception (PPA) and object perception (LOC)
BUT
Mc Evoy & Epstein (2011)
LOC also involved in some aspect of scene perception provided that the exposure time is sufficient to activate a mental representation of a scene.
2 main regions for scenes :
PPA & RSC Retro spinal cortex
PPA
Parahippocampal Place Area
(located between the posterior part of the hippocampal and the anterior part of the lingual gyrus)
Activated by the geometry of the known or unknown scene, even layout, wireframes (like FFA for face)
Photograph of places (rather than objects, textures and faces)
Not involved in spatial navigation
Responds more to HSF
Lesions
cannot identify building, houses shops ..
Patient (GN) cannot recognize familiar places but he can recognize his own objects (Mendez et Cherrier (2003) Neuropsychologia)
RSC
Retrosplenial cortex
Occipito-temporo-parietal junction
Connected to PPA, lateral parietal cortex, area 7a and enthorinal cortex. (involved in spatial working memory).
>Activated by the active or passive viewing of a scene,
> by the mental imagery of a scene (fMRI - reach mentally some places)
> by navigation in familiar or unfamiliar, real or virtual environments
>In contrast to PPA it is more activated by familiar than by unfamiliar environments.
Lesions
difficulty to navigate (can disappear after few month if unlateral)
The context effect
The association between an object and its setting
Congruency btw context and object improves object recogntion
(priest in chuarch VS priest on gamefield)
> spatail attention : background is used to direct eyes mouvement to the probable location of target
fMRI - context network
RSC - PPA - SOS (superior orbital sulcus)
Ophtalmology : AMD
Age related Macular Degeneration
Dry (75%) / wet (cured with laser)
Destruction of pigmental eputhelium > no food for receptors > death
20% of > 70 y.o.
All functions of central vision affected :
accuracy
contrast sensitivity
High spatial F
color vision
Functional consequences on :
Reading, writing
Watching TV
Daily living (cooking, shopping, using public transports…)
Impaired object and face recognition
Mobility
Social life
Risk of fall X4
AMD - summary
People with a central scotoma are able to
Categorize scenes (>75% correct)
Facilitation colored vs gray level scenes and objects
Facilitation for isolated objects than objects in scenes (crowding effect)
A partial isolation improves performance
Benefit from context information implicitely (but not explicitly)
(citer une methodo pour chaque)
Scene exploration
Perf 30% lower than control
PRL Prefered Retina Locus
Daily life
Longer to observe / saccades ++ (try to put object in PRL)
→ Familiar task OK, impaired for unfamiliar
Ophtalmology : Glaucoma
Painless, ppl don’t realize they’re affected
Quiet comon : 1M France - 10M Europe
Symptoms
not aware
blured patches
tunnel
Cause
Occular hypertension because of obstruation of evacuation channel
magnocells axons impaired by hypertension
Early diagnosis : contrast test (cheap & robust)
Perf ↘ when eccentricity ↗
Future
Relation btw glaucoma and M pw
Explore other natural actions : shopping virtual with eye movement recording
Neurological pathology : Agnosia
Benson syndrome : temp-pariet lesion (variant of Alzheimer but ≠ location)
hemianopsia (blind for left visual field)
Naming task (cat & exact)
0-20 % exact
18-50% cat
Categorization (saccades)
Perf idem as control on right visual field
Psychiatric pathology : schizophrenia
Visual exploration
NO exploration of faces, object, scenes, meaningless texture
> without any instruction
With task, e.g. object naming : normal exploration (normalized pattern of exploration)
Reduced influence of context (cognitive dysfunction)
Social problems of SZ could come from an impairement in (emotional > face analysis) context integration
BEE
Boundary extension effect
> distortion of visuospatial memory (only for objects in scenes)
> occurs in PPA & RSC
We do not memorize details but meaning of scenes
The current view of a scene is rapidly incorporated to a mental schema which includes the memory of similar scenes encountered before together with the subject’s expectations of what should be in that scene.
Even if we are not aware of it scene recognition is an active processing which combines perception, memory and past experience to create an internal reconstruction of the visual world.
