Collections > Electronic Theses and Dissertations > A Computational Model to Investigate the Influence of V1 Cell Properties and Topographic Organization on V2 Response to Illusory Contours, with Applications in the Study of Cortical Injuries in the Primary Visual Cortex

I present a model capable of illusory contour detection. Unlike previous models, this model uses a realistic topographic organization of the orientation preferences of cells in the primary visual cortex. I show that using a feed-forward mechanism, this model can accomplish illusory contour detection at the level of V2 even with a non-uniform distribution of orientation preferences amongst simple and complex cells. The model is applied to the study of the properties of V2 cells that respond to illusory contours. I show that 1) inducer spacing preference depends directly on the receptive field width of simple cells in the primary visual cortex, 2) the shift in the orientation tuning peak as a function of inducer angle relative to illusory contour orientation is determined by the distribution of end-stopped cell orientation preferences in the presynaptic input to the V2 cell, and 3) the contrast response function of V2 cells increases more rapidly for real contours than for illusory contours. I also use the model to study the consequences of a primary visual cortex lesion on visual function. I show that for small lesions, response degradation for neurons downstream from the injured area increases linearly with the size of the damage. Using an additional layer of classifier units as a proxy for neural correlates of higher visual functions, I characterize the extent to which recovery of visual function is possible following cortical injury. I show that while both spontaneous and training-induced recovery can lead to restoration of visual function, spontaneous recovery is more effective, and under certain conditions can restore visual function to pre-lesion levels.