An architectural model of the non-linear subunits of complex cells of visual cortex based on gated intracortical interactions

In this study we propose an architectural model to describe the non-linear effects of intracortical interactions over complex cell receptive fields (RFs). Afferent inputs from lateral geniculate nucleus (LGN) are recurrently combined to form non-linear computational subunits described as second order Volterra integral operators over RFs. Specifically, the feedback path of intracortical interactions is gated by additional LGN afferent projections. The resulting complex cell subunits incorporate summing mechanisms activated by the concomitant stimulation in both central and peripheral parts of the RF. The relation of this type of controlled coordination of sensory information in visual cortex to its signal processing capabilities are considered in perspective.