Purchase this article with an account.
Melanie Bernard, Zhiyi Zhou, A. B. Bonds; A Synchrony-based sparse code in Cat visual cortex signals complex contours in natural images. Journal of Vision 2007;7(9):389. doi: https://doi.org/10.1167/7.9.389.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The mechanisms by which salient stimulus features are represented in neuronal responses remain unresolved. We have previously shown that synchrony between cell pairs can represent co-circular contours, suggesting that synchrony within larger assemblies may be involved in encoding more complex contours. To investigate the role of synchrony as a contour-encoding mechanism in natural vision, we measured the synchronous responses of large neural assemblies to a sequence of 3024 natural images. Using a 10×10 microelectrode array, we recorded from 75 complex cells in the primary visual cortex of two paralyzed and anesthetized cats. We randomly identified 4500 neural assemblies, ranging in size from 2 to 10 cells with 500 assemblies per size group. Using a novel measure to quantify synchrony within assemblies of arbitrary size, we found that the receptive fields of cells within an assembly tend to be aligned on a complex contour in the image with the highest synchronized response. Furthermore, spline and receptive field analyses reveal that each complex contour is relatively conserved across images with highly synchronized responses. In contrast, contours are not conserved in the images generating the highest average firing rate across an assembly. We also measured the synchrony and average firing rate response distributions for each assembly over the set of images and used six different selectivity metrics to compute population and lifetime sparseness values. All metrics indicate that synchrony response distributions were sparser than firing rate response distributions and sparseness increased with assembly size. We propose that higher-order features found in natural images (e.g. complex contours) are responsible for the high selectivity of synchrony compared to average firing rate because adequate descriptions of high-order spatial correlations require the coordinated response of multiple cells. This process is cumulative, in that more complex structures require larger neural assemblies for accurate description.
This PDF is available to Subscribers Only