Decoding Perceptual States from Local Field Potentials in Visual Cortex during Generalized Flash Suppression

Generalized flash suppression (GFS) [Wilke et al., 2003] in which a salient visual stimulus can be rendered invisible despite continuous retinal input has provided a powerful means to study neural processes directly related to perception. However, the mechanisms underlying such perceptual suppression remain poorly understood. Here we ask how reliably the population responses can determine the perceptual states and what the roles of different areas play in visual cortex during suppression. Three monkeys were trained to perform the GFS task in which multi-channel neural activities in visual cortical areas V1, V2 and V4 were simultaneously recorded. Linear regression analysis on the time-frequency distributions of local field potentials (LFP) recordings showed that significant power difference between the two perceptual states (visible vs. invisible) was mainly focused on the beta frequency band of 10-30 Hz. Linear discriminant analysis (LDA) was performed to decipher perceptual states. The results showed that LFP recordings always provide significantly better prediction than the chance level and the beta band provides better prediction than the broad band in all the three visual areas. In addition, multi-channel recordings are more informative than single electrode recordings in V1, V2 and V4, while in V1 and V4 the benefit of decoding population activity is significant. Furthermore, we found that the channel-wise correlations limit the benefit of population activity. These results suggest that considerable information is retained in the beta band of the local field potentials and that population activity is more informative than that of single channel in isolation.