Information processing in visual systems
Author(s)
Saleem, Aman
Type
Thesis
Abstract
One of the goals of neuroscience is to understand how animals perceive sensory information.
This thesis focuses on visual systems, to unravel how neuronal structures
process aspects of the visual environment. To characterise the receptive field of a
neuron, we developed spike-triggered independent component analysis. Alongside
characterising the receptive field of a neuron, this method provides an insight into
its underlying network structure. When applied to recordings from the H1 neuron of
blowflies, it accurately recovered the sub-structure of the neuron. This sub-structure
was studied further by recording H1's response to plaid stimuli. Based on the response,
H1 can be classified as a component cell. We then fitted an anatomically
inspired model to the response, and found the critical component to explain H1's
response to be a sigmoid non-linearity at output of elementary movement detectors.
The simpler blowfly visual system can help us understand elementary sensory information
processing mechanisms. How does the more complex mammalian cortex
implement these principles in its network? To study this, we used multi-electrode
arrays to characterise the receptive field properties of neurons in the visual cortex of
anaesthetised mice. Based on these recordings, we estimated the cortical limits on
the performance of a visual task; the behavioural performance observed by Prusky
and Douglas (2004) is within these limits. Our recordings were carried out in anaesthetised
animals. During anaesthesia, cortical UP states are considered "fragments
of wakefulness" and from simultaneous whole-cell and extracellular recordings, we
found these states to be revealed in the phase of local field potentials. This finding
was used to develop a method of detecting cortical state based on extracellular
recordings, which allows us to explore information processing during different cortical
states. Across this thesis, we have developed, tested and applied methods that help
improve our understanding of information processing in visual systems.
This thesis focuses on visual systems, to unravel how neuronal structures
process aspects of the visual environment. To characterise the receptive field of a
neuron, we developed spike-triggered independent component analysis. Alongside
characterising the receptive field of a neuron, this method provides an insight into
its underlying network structure. When applied to recordings from the H1 neuron of
blowflies, it accurately recovered the sub-structure of the neuron. This sub-structure
was studied further by recording H1's response to plaid stimuli. Based on the response,
H1 can be classified as a component cell. We then fitted an anatomically
inspired model to the response, and found the critical component to explain H1's
response to be a sigmoid non-linearity at output of elementary movement detectors.
The simpler blowfly visual system can help us understand elementary sensory information
processing mechanisms. How does the more complex mammalian cortex
implement these principles in its network? To study this, we used multi-electrode
arrays to characterise the receptive field properties of neurons in the visual cortex of
anaesthetised mice. Based on these recordings, we estimated the cortical limits on
the performance of a visual task; the behavioural performance observed by Prusky
and Douglas (2004) is within these limits. Our recordings were carried out in anaesthetised
animals. During anaesthesia, cortical UP states are considered "fragments
of wakefulness" and from simultaneous whole-cell and extracellular recordings, we
found these states to be revealed in the phase of local field potentials. This finding
was used to develop a method of detecting cortical state based on extracellular
recordings, which allows us to explore information processing during different cortical
states. Across this thesis, we have developed, tested and applied methods that help
improve our understanding of information processing in visual systems.
Date Issued
2009-12
Date Awarded
2010-01
Copyright Statement
Attribution NoDerivatives 4.0 International Licence (CC BY-ND)
Advisor
Schulz, Simon
Sponsor
Gatsby Charitable Foundation
Creator
Saleem, Aman
Publisher Department
Bio-engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)