Implication Of The Primary Visual Cortex Neural Activity From Top-down Face Processing

Face processing plays an important role in people’s daily communication:people started to accept training about face processing since childhood automatically. They have accumulated a large amount of specialized face processing experience. With the help of this expert knowledge, we not only have the ability to identify hundreds of thousands of faces without any difficulty, but also can recognize incomplete or extremely fuzzy faces with the help of visual and memory information. Top-down face processing is the combination of primary visual information processing and advanced cognition processing (e.g. extraction of memory information, transfer of attention and acquisition of prior knowledge).With the development of functional magnetic resonance imaging (fMRI), more and more researchers have applied this technology to the study of face processing and neural mechanisms in recent years. Nowadays, the research of top-down face processing mainly focuses on the influence of the signal from the senior cognitive prefrontal cortex to the fusiform face area, which is located in the ventral temporal. However, we still know very little about how this signal modulates primary visual cortex neural responses.In order to have an in-depth discussion on this issue, we will study the neural mechanism of illusory face perception. Compared with previous research methods, this study has made improvements in these aspects:first, on the selection of stimuli, we use Chinese characters as the control condition of faces. The advantage of this improvement is that our subjects have rich expert knowledge in both face processing and Chinese character processing. Faces and Chinese characters are recognized at the individual level; second, we adopt the "fall-related" paradigm to study the neural mechanism of illusory face detection. The design of this experiment requires subjects to detect faces and Chinese characters in pure noise images individually with the existing expectations of faces and Chinese characters. The advantage of this improvement is that little information about faces or Chinese characters is contained in pure noise images. Consequently, little information coining from bottom-up face processing and Chinese characters exists when we compare the activation level of face processing with the activation level of Chinese characters. Third, in the design of algorithms, we use multivariate pattern analysis (MVPA) to analyze the relationships between activation patterns of the subjects’ brains and the cognitive tasks. We improve an algorithm to test the weight of each input dimension of classifier to generate sensitivity maps. The final hope is to determine how important each voxel is in the classification of the current task. The advantage of this improvement is that we can use it to analyze the reaction patterns in visual cortex caused by different stimuli, rather than to compare the activated level of the independent region of interest as the traditional method of general linear model.Through these methods, we have the following findings:(1) The information coming from top-down face processing not only can modulate the stage of the activation of the advanced visual cortex, but also place implementation on primary visual cortex; the modulation rises with the increasing level of primary visual cortex.(2) With the level of primary visual cortex increases, reaction level of primary visual cortex coming from expectation of face versus expectation of Chinese characters increases gradually.In its final words, the paper discusses the development trend of this research area.