Visual system 3

Visual system 3

1. Lateral geniculate nucleus

   • Thalamic relay station
     • Ganglion cell axons make 1:1 connections with LGN projection neurons
       • Receptive fields of LGN neurons are similar to the receptive fields of retinal ganglion cells
     • But:
       • 60% of synaptic input from cortex (“back propagation”)
       • Local interneurons in LGN
       • The human and monkey results indicate two visual pathways
         • Neurons in the V1 cortex form retinotopic maps
           • Occular dominance columns in the visual cortex
             • Each column receives input from either ispilateral or contralateral eye

2. Simple and complex cells

   • Simple cells
     • Localised in layers 4 and 6
     • Probable circuit mechanism of receptive field
       • See bears neuroscience p328
   • Complex cells
     • Receptive field
       • i.e. respond to a bar in certain orientation presented anywhere in the receptive field
       • circuit basis
         • Complex cells collect information from many simple cells with similar orientation
     • Localised in layers 2,3 and 5
     • 'Hypercomplex' = end stopped complex cell

3. Receptive fields downstream of L1

   • Increase complexity
   • Difficult to study
     • 2 approaches
       • Guess
       • “Spike triggered average” and the like
       • Both approaches have limitations

4. V4 neurons respond to presentation of simple shapes

   • Response of the same neuron to presentation of different shapes
   • Complexity taken into extreme "grandmother cell" or "jennifer anniston neuron"
     • Recording from humans find that 1 or a few neurons recognise a particular object

5. Object recognition

   • Orientation invariance
   • Scale invariance
   • Lesions in IT cortex decrease ability to recognise objects
     • Mainly studied by modelling
   • Hierarchical models of object recognition inspired by cortical processing
     • IT and higher areas perform categorisation
       • Shape selective neurons converge to form object selective neurons
         • Feature selective neurons converge to form shape selective neurons (V2-V4?)
           • Each small part of the retina contains a set of feature selective neurons
             • Features = edges?
     • Problems
       • Still poor in scale- and orientation invariance
       • Do not take into account feedbacks from higher cortical areas (“top-down regulation”)
       • Need experimental validation
         • Difficult because would need to record pre- and post-synaptic responses
           • Discussed in lecture 5

Bilder und Dateien der Mindmap