主題演講摘要
黃榮村教授 (Huang, Jong-Tsun)
　 Monocular and binocular behavior around the blind spot: Implications for possible cortical dynamics
The blind spot is a retinal region devoid of photoreceptors and the same area for one eye’s blind spot in space is covered by the corresponding area in the fellow eye. In human primary visual cortex, the blind spot is represented as a relatively large monocular region that receives direct input solely from the ipsilateral eye and not from the contralateral blind-spot eye. Monocular blind-spot filling-in is widely believed to exist and exert its influence on a variety of accompanying visual tasks. Recent observations, for example, have demonstrated its effects on the process of binocular rivalry. Through this experimental set-up, He & Davis (2001) proposed that the filling-in process occurred in the natural blind spot likely happens at least before the completion of the rivalry process. Tong & Engel (2001) also found that the filled information in one eye’s blind spot can contribute to binocular rivalry, although in a diminished way. They suggested that this influence may indicate an interocular competition in monocular V1 area. This functional demonstration of the filling-in effect in V1 is also consistent with the supposed existence of ODC (optic disk cortex) neurons in V1 (Gattas et al., 1992; Fiorani, et al., 1992). We report three observations in this study: (1). The filling-in mechanism of the blind spot can compute both 2D and 3D completion through the spreading effect of background or surrounding texture information. (2). While an artificial scotoma was created to the blind-spot corresponding area in the fellow eye, the completion is less efficiently obtained through the facilitation of surrounding texture than does in the natural blind spot. The fact suggests a possible intervention from the supposed function of ODC neurons that are only responsible for figural or depth completion around the blind spot. (3). Depth interpolation behavior of both binocular rivalry and blind spot was simultaneously tested in the same RDS experiments. While both of these tasks lose their clear disparity information (i.e., ambiguous), the depth representation in blind spot as well as in binocular rivalry seems to be strongly penetrated by the surrounding disparity information. Based on these observations, a neural network was then proposed and simulated on the 2D filling-in process around the blind spot by incorporating with the supposed function of ODC neurons that is both excitatory and symmetric in operation. The fit of the model is nicely matched with experimental results. The network should also be applicable for the simulation of 3D behavior around the blind spot. Implications will be discussed in the sequel.