The neural computation for perceptual filling-in

点击翻译按钮获取中文摘要

Filling-in is the perceptual tendency of an observer to perceive a continuous visual pattern despite the presence of an intermittent blank region. Filling-in occurs at the blindspot but also in scotomatous regions in patients with diseases of the visual pathways. Since filling-in occurs in a blank region there is no physical stimulus to elicit a response in any visual mechanisms at that retinotopic location. Observers are usually requested on a given trial to report when they experience filling-in. Neural responses to stimuli with and without filling-in can be compared to determine whether these responses differ on these two trial types. Machine learning can be applied to determine if a classifier can distinguish between these two types of events. With univariate analysis, we found no difference between BOLD activation in the filling-in and no filling-in trials. However, using a leave-one-out training procedure and a support vector machine, it was possible to classify the percepts by the activation pattern differences in early visual cortex (Lin, Greenlee & Chen, 2020). To resolve these issues, we have developed a new paradigm to assess the presence or absence of perceptual filling-in. Observers will be presented periodic patterns with intermittent blank regions, which serve as artificial scotoma. By presenting a target in the blank region after filling-in occurs we can measure the neural response to the target and determine whether this response is affected by the presence of filling-in. The target will be a stimulus that can elicit a large enough neural response to allow for a reliable measurement of neural activity. We will vary the contrast of the target to determine the contrast response function in the presence of the inducer. This allows us to determine the contrast response function to the target. Variations in the physical properties of the inducers and targets will be conducted to separate response components related to the inducer and target stimuli.In a series of three studies, we will parametrically measure the response functions to the target with functional magnetic resonance imaging (fMRI), event related potentials (ERP) and psychophysics experiments. The latter will be conducted to establish the operating range of the basic phenomena. The fMRI experiments will precisely identify the brain areas for filling-in. Such precision is required as the candidate area for filling-in, V2, is small. The ERP experiments will determine the temporal dynamics of the target response with respect to the onset of filling-in. The filling-in phenomenon will be related to other illusions involving border contrast.The results of these studies will be simulated using a computational model of early visual processing. In these models, lateral inhibition and excitation influence the neural mechanisms that respond selectively to target stimuli. In this way, we will be able to develop a new theory of perceptual filling-in using computational modelling.