Abstract
Long-range lateral interconnections between cells in the striated cortex are characteristic of the adult visual brain. These connections are not present at birth and begin development after the first 2.5 months, a period of rapid synaptogenesis. Improvements in various visual functions such as contour integration have been attributed to this maturation process. We implemented a computational model to explore the role of these developing lateral interconnections in contour integration. Our model is based on the known cortical structures in the primary visual cortex. The model represents the retinotopic spatial layout in the cortex using two dimensions to create a plane. The columns of orientation-selective cells are represented in a third dimension. The model used a facilitation rule that is a function of the Cartesian distance between two cells. In effect, a cell of a particular orientation preference in a particular spatial location sends a sphere of excitatory connection to cells of other orientations on other spatial locations. The goal of the computational model is to locate contours consisting of oriented elements in a background of noise consisting of randomly oriented elements. The preliminary results showed that the model preferred contours whose elements followed “good continuation”. Co-aligned and co-axial elements produced greater levels of activation despite the lack of an explicit co-linearity facilitation rule. Physiological recordings of firing activity in the primary visual cortex have revealed facilitatory interaction of cells separated by several millimeters. These long-range lateral connections may serve to integrate information across multiple receptor fields. These connections are not present at birth and develop after the first two months. The next step will be to manipulate the cortical extent of these lateral facilitatory connections in the model to simulate the effect of the maturation of the primary visual cortex on contour integration.