Cognitive Science > Action and Cognition > Lecture/2

Jump to: navigation, search

Contents

[edit] Visual pathway

[edit] Retina, LGN

[edit] Ocularity

  • Organization of afferents
  • Regular alternation of the eye of origin of the information
  • Left eye and right eye information is separated in the LGN, but blended at the level of the cortex
  • "ocular dominance" columns in the visual cortex

[edit] Primary visual cortex (V1)

  • Located at the occipital pole of the cerebral cortex
  • Macaque monkey: On the outer surface, easily accessible
  • Human: Most of it deep down in calcarine sulcus

[edit] Orientation columns

  • Response properties over depth of cortex are remarkably constant -> cortex organized in columns
  • Cells of similar orientation selectivity form "columns" in the cortex
  • Orientation selectivity is mapped in a continuos fashion across the cortex

[edit] Simple cells & complex cells (Repetition)

  • Simple cells
    • Explanation: Convergence of LGN cells with on-center, off-surround property where centers are aligned in a line
    • ~15% of synapses that serve as input to simple cells are from LGN -> Input can only give a bias
    • Orientation selectivity: a) Property of bottom-up wiring | b) Property of cortex
    • Circuitry is a sufficient explanation of simple cell effects, but not the only one
  • Complex cells
    • Convergence of simple cells which each have the same orientation selectivity
    • "Superposition of simple cell receptive fields"
    • Again circuitry explains effects, but is not the only explanation
    • Problem: Complex cells do also get primary input, not only from simple cells
  • Pattern of lines moved across receptive field
    • Reaction of simple cells: Sinusoidal oscillation of membrane potential matching rhythm of bars passing
    • Complex cells: Steady depolarization, action potentials triggered at seemingly random times (adaptation after some time)
    • Invariance means throwing away information

[edit] Putting it together: The Ice Cube Model

  • z: orientation | x: ipsi vs contralateral eye | y: 6 layers
  • hyper column: Roughly 2x2mm^2, building blocks of ice cube model
  • One processing unit devoted to one region of space

[edit] Blobs of Color

  • Color info from LGN cells in input to layers other than layer IV, within columns

[edit] Split brain patients

  • Cut at optic chasm
  • Experiment: "I see nothing" if in wrong part of visual field
  • Left hemisphere confabulates explanation for things the other hemisphere was doing
  • Movie: Split brain (right hemisphere "just a relic", planning etc in left)

[edit] Retinotopy and receptive fields

  • Large part of visual cortex reserved for fovea -> "Fish-eye view"
  • Overrepresentation of central part of visual field
  • Maps uneven distribution of ganglion cells in retina, emphasizes it even more
  • Need for attention
  • Impression that we see everything with similar resolution

[edit] Selectivity for different features

  • Direction selectivity
    • Dark bar moving across receptive field
    • Different neurons in V1 respond differently to different orientations and (new) different to forward-backward
  • Disparity selectivity
    • Distance of image from fovea changes to the same degree in both eyes, independent of whether ##
    • From disparity, brain can infer distance of objects
    • Response strength of a particular neuron as a function of disparity
    • Distance of stimulus vs. distance of fixation point
  • Selectivity for spatial frequency
    • Images can be described by combinations of different spatial frequencies
    • Neuronal activity often dependent on spatial frequency
    • Image with high freq filtered out -> blurred
    • Low freq filtered out -> Only edges, no brightness

[edit] A modern view on cortical maps

  • Optical imaging
    • Direct video camera on brain, record blood supply increase in cortical region, oxygen saturation
    • Oxygenated haemoglobin looks slightly different from deoxygenated h.
    • Present tilted bars, hundreds of trials, compute difference in images, look for orientation of bar where each pixel is most active, color it according to orientation of bar
    • Orientation map encodes for each neuron the preferred orientation
  • Pinwheels
    • Regions in cortex where all orientation-sensitive neurons meet
    • -> without analysis not completely clear, where hypercolumns are, discontinuities
  • Analysis of cortical maps
    • Combine orientation map with black-white map which signifies ocular dominance
    • In regions where orientation changes fast, ocular dominance changes slowly
    • General scheme: When one selectivity changes quickly, the other changes slowly
    • Quantitative analysis
      • Compute, whether for each combination of features we find a neuron which is interested in it
      • Whole stimulus space should be covered a smoothly as possible
      • The relative arrangement of of feature maps optimizes coverage of stimulus space
    • By shifting one map, one creates feature combinations which cannot be detected (-> optimality)

[edit] Non-classical receptive fields and natural stimuli

  • Camera mounted on cat, electrodes implanted
  • Capture actual visual stimuli to understand processing in the visual system
  • Properties of cortex correlate with statistical properties of natural simuli
  • In V1: Neurons with similar receptive field properties are preferentially connected, even over larger distances
  • General phenomenon: Neurons with similar properties wire up together
  • Property of visual world found in brain
  • Hypothesis: Properties of visual world learned from environment
  • "If there is a lot of vertical contour in one place, the chance to find another vertical contour nearby is very high" - similar likelihood of connection

[edit] Experiment: Glasses which show world upside down

Source: http://wiki.stuhlmueller.info/Cognitive_Science/Action_and_Cognition/Lecture/2