Abstract
In contrast to traditional hierarchical models, most current models of visual perception suggest that distributed networks of regions across the visual processing stream underlie visual recognition. For example, multiple face patches, including the occipital face area (OFA) and the fusiform face area (FFA) likely work in concert to encode individual faces. However, direct evidence for distributed computation of individual faces does not exist because to date no methods exist to examine the information represented in neural interactions. Here we develop a novel pattern recognition method, called Multi-Connection Pattern Analysis (MCPA), to extract the discriminant information about cognitive conditions solely from the shared activity between two neural populations. In MCPA, functional connectivity models are built based on shared multivariate neural activity using canonical correlation analysis for each condition. Then using these models the activity in one area is predicted solely based on the activity in the other area for each condition. Classification is achieved by comparing the predicted activity with the true activity, revealing the representational structure of the shared neural activity (e.g. the information represented in the functional interaction). MCPA was applied to analyze intracranial EEG (iEEG) data recorded simultaneously from OFA and FFA in a human subject. Our results support the hypothesis that individual-level face information is not only encoded by the population activity within certain brain populations, but also represented through recurrent interactions between multiple distributed populations at the network level. In addition, the critical time window for face individuation based on MCPA was approximately 200 – 500 ms after stimulus onset, which is consistent with our previous study based on iEEG recording from FFA only. This suggests the involvement of FFA in the face individuation process is a result of temporally synchronized, recurrent interactions between FFA and other nodes in the face-processing network, including the OFA.
Meeting abstract presented at VSS 2016