Nonlinear Dynamics of the Primary Visual Cortex

New York University, New York NY

Investigators

Abstract

McLaughlin 0211655 The investigator and colleagues have been developing a biologically constrained large-scale computational model of the "front-end" of the cortical visual system -- the primary visual cortex (V1). To date, their work has focused upon local properties of individual cells within the large-scale network -- properties such as orientation selectivity and simple vs complex cellular dynamics. In this project, they scale up to a more global model of V1, reaching scales large enough to study some elementary optical illusions of psychology and psychophysics. This involves several square millimeters of lateral cortical area, together with a multi-layered architecture -- with emphasis given to cortical dynamics. First, a "coarse-grained mixed representation" is derived mathematically and tested numerically -- a representation that combines spatially coarse-grained (local) mean firing rates, representing local background cortical operating points, with an idealized representation of a sub-network of individual point neurons embedded within this background and retaining the detailed firing patterns of individual neurons. These two components interact with each other -- with the coarse-grained local operating points influencing the responses of the individual neurons, and vice versa. Second, the global mixed representation is used to study specific dynamical phenomena in visual cortical processing: (i) the layer-specific "dynamics of orientation selectivity" (as measured by reverse time correlation methods); and later (ii) "bistability in figure-ground assignments" (as detected in psychology and psychophysics experiments). In both cases, the phenomena involve extensive lateral regions of V1, its layered structure, and (likely in the case of figure-ground assignment) other cortical regions. And in both cases the work involves close interaction with the experimental work of neural scientist Robert Shapley. Today, through new biological experiments combined with the power of modern scientific computation, scientists and applied mathematicians are making significant strides toward understanding the human brain. Visual perception and the cortical processing of visual information provide important starting points. By focusing upon the "front end" of the cortical visual system, McLaughlin and his colleagues develop computational models of the visual cortex that are strongly constrained by biological experiments. In this project the investigators develop computationally efficient numerical methods that permit scale-up of the models to global representations of the primary visual cortex -- reaching cortical scales large enough to study some elementary optical illusions of psychology. This work requires the direct interaction of applied mathematicians, computational scientists, and neural scientists -- and involves theoretical, computational, and experimental components.