Research

  1. Interests
  2. Gallery
    1. Multiple single unit responses to natural scene movies in primary visual cortex
    2. Simulation of a cortical polymap for 8 binary sub-features
    3. Retinal wave model
    4. Computer simulation of the formation of ocular dominance and orientation maps in primary visual cortex
    5. Simulation of V1 development using the Kohonen algorithm.

Interests

I am interested in vision, and the development and organization of the primary visual cortex. This includes studying how the genes and environment interact in early post-natal development, how cellular mechanisms contribute to perceptual processing, and how disorders such as amblyopia and glaucoma may affect visual function. I use computer models to simulate developmental mechanisms, cats as experimental models for visual processing, and humans as subjects for psychophysical research. Present and past research projects include the following:

  • Application of neural net models to the formation of computational maps of ocular dominance and orientation columns in the visual cortex
  • Analysis of columnar organization in the cat visual cortex using optical recording of stimulus evoked neural activity
  • Multi-electrode recording methods for the comparison of receptive field properties in simultaneously recorded clusters of neurons
  • Quantitative analysis and modeling of spatial summation in simple and complex cell receptive fields in cat visual cortex
  • Physiological and psychophysical studies of the mechanisms of vernier hyperacuity
  • The development of clinically diagnostic psychophysical tests of visual function in glaucoma
  • Early detection of glaucoma using mathematical modelling of optic nerve head shape and neural network methods for classifying images as normal or glaucomatous.

Gallery

Multiple single unit responses to natural scene movies in primary visual cortex

We designed a variety of 16 and 54-channel silicon electrodes, formerly made by the University of Michigan’s Center for Neural Communication Technology, now made by NeuroNexus. These “polytrodes” comprise a 2D planar array of closely-spaced (46-70 µm) electrode sites, arranged into two or three columns (colinear and staggered).

We used natural scene movies (those shown here courtesy of the Peter Konig lab) that were captured by attaching a video camera to a cat’s head and allowing it to roam freely in the woods. We presented these movies while recording multiple units simultaneously with silicon polytrodes in anesthetised cat primary visual cortex. In the movies below, which were provided with sound tracks by Martin Spacek, spikes from different neurons are each assigned a different audible pitch. Neurons are also separated in stereo space. A pulse of a given pitch represents a spike from a specific neuron at that time. In this way, you can listen to the entire recorded population of neurons responding to the movie stimulus in real time. The movie shows responses to a repeated stimulus, which gives a sense of the reliability of the responses.

spiking responses to repeats of a natural scene movie
spiking responses to a continuous natural scene movie

Simulation of a cortical polymap for 8 binary sub-features

Each coloured domain represents a particular permutation of 8 binary sub-features. Since they are binary, there are 256 possible combinations, each represented here with a unique colour. Adjacent domains normally differ by only one sub-feature, so the map is continuous, while particular features (specific colours) have a roughly constant spacing across the map, i.e. good coverage. Individual sub-features form a pattern of irregular stripes and blobs resembling normal ocular dominance patterns. Note that in a real map, particular combinations of sub-features might or might not be present, depending on whether or not they had been experienced in early life. For a discussion based on this, and a model of how cortical maps could serve as content-addressable memories, see Swindale, 2008). For further details of how this map was calculated, see Swindale (2000).

Retinal Wave Model

Retinal waves play an important role in the very earliest stages of visual map formation. This is a simulation of spontaneous waves of activity in a network of amacrine cells in the developing retina based on the model of Godfrey and Swindale (2007)

Computer simulations of the development of ocular dominance and orientation maps in primary visual cortex

Ocular dominance map produced by the model of Swindale (1980).
Orientation map produced by the model of Swindale (1982).
Combined orientation and ocular dominance map with singularities (or pinwheels) shown as white asterisks and ocular dominance column boundaries as white lines, produced by the model of Swindale (1992).

Simulation of V1 development using the Kohonen algorithm

The Kohonen algorithm was used to map from a manifold in a 6-dimensional space onto a 2D surface (the cortex). For this implementation, ocular dominance and spatial frequency were binary parameters, so there is essentially no difference between them. Orientations were points on a circle. The retinotopic map shown represents the mapping of the cortex into retinal space (rather than the mapping of a parameter onto the cortex, as shown in the other panels). The lines between cortical points are coloured according to their ocular dominance value.