A mathematical model for activity-dependent self-organization of functional maps in the visual

Masanobu Miyashita

The visual cortex contains regular arrangements of neurons corresponding to specific types of visual stimulation, such as ocular dominance, orientation, and direction. These cortical structures can be considered to constitute the functional architecture for early visual information processing (Fig. 1). Recently, ocular dominance and orientation maps were observed in cat primary visual cortex of a cat raised without visual experiences. These experimental results suggest that organized visual input is not necessary for formation of cortical maps. On the other hand, it has been found that map formation is affected by extraordinary visual experiences during critical period. For example, in monocularly deprived animals, it is well known that the number of cells that respond to visual inputs to the deprived eye decreases compared with a normally reared animal and that the width of deprived eye domains become narrow. With respect to orientation maps, there is still some controversy as to whether or not the number of cells that have selectivity to the presented orientation increases, when the animal is presented an oriented contour as it is reared.

Fig.1

We proposed a mathematical model for self-organization of neural networks based on Hebbian learning rule, and show that cortical receptive fields and their related columns (ocular dominance, orientation and direction map) were formed simply following exposure to unorganized light inputs. In this project, we attempt to construct an activity-dependent self-organization model by incorporating recent experimental results from neurophysiology and anatomy, aiming at understanding the role of visual experience in the functional map formation. Now, we conducted computer simulations for two conditions: (1) activities of LGN cells were evoked only by unorganized random visual input to the retina (Fig. 2a); and (2) LGN neurons were activated by the presentation of moving grating patterns in various directions of motion (Fig. 2b). The results suggest that visual experience can promote the maturation of orientation maps.

Fig.2