NEURAL COMPUTATIONS IN VISUAL CORTEX

点击翻译按钮获取中文摘要

We plan to study visual information processing in primary visual cortex of the macaque monkey at the level of the local field potential and the single unit. In the initial funding period, we have developed a set of techniques to examine and model spatial and temporal interactions in the visual cortex at an advanced level of detail and sophistication. Results obtained with these techniques have provided evidence of the involvement of VI in the extraction of spatial features, the linking of these features, segmentation, and temporal coding. The planned research will build on these findings, and will be organized along two lines: the functional role of a class of lateral interactions, and the functional role of the influence of visual inputs on the temporal correlation structure of neural activity. The study of lateral interactions will determine how they depend on contrast, whether they carry information on spatial scale, chromatic content, or orientation, whether they selectively modulate only some aspects of cortical form processing, and whether they vary across laminae. Answers to these questions will help determine whether these interactions are important for feature extraction, segmentation, "filling-in", and other processes. The study of how visual stimuli modulate the ongoing correlation structure of neural activity will begin with the use of phase-locked spectral analysis to test dynamical models for this process. We will then examine the spatial extent of this process, the degree to which it carries specific information concerning the visual stimulus, and its laminar organization. These findings will help determine to what extent VI transforms visual information into a temporal code, and how such temporal codes are generated. A central theme in this research is that nonlinearities are of great functional importance in both spatial and temporal processing. A successful study of complex nonlinear systems requires tailoring of the experimental approach to the biological questions of interest. Therefore, an explicit part of this research program is the continued development of new experimental and analytic techniques to probe nonlinear neural systems, and a tight interaction between theory and experiment. Visual information processing is considered to be a model for sensory information processing in general. Analysis of the roles played by lateral interactions and temporal patterns of ongoing neural activity in the processing of form and color allow us to study feature extraction, cross- modality interactions, and neural coding in a context in which specific hypotheses may be proposed and models may be tested.