Studies of Local Functional Network Connectivity in Cat Visual Cortex

Abstract

In the mammalian brain, network structures are apparent on multiple temporal and spatial scales. While substantial knowledge is available about the action potential on the level of the individual cell, as well as about large-scale networks across brain areas, the intermediate level of local functional neuronal networks within one brain area has not been analyzed to the same extent. The advancement of multi-electrode recordings in the brain enables the analysis of functional connectivity between simultaneously recorded groups of neurons. With this, there is an increasing need for analysis approaches to study the characteristics of the recorded local functional networks. In this context, the present study had two main goals: 1) the extension and development of analysis methods, and 2) the application of these methods to electrophysiological data in order to contribute to the knowledge in the field of local functional connectivity. The data which were analyzed in this study had been collected as part of a cycle of studies on visual hemineglect. The study of neglect is closely linked to the study of feedback signals between areas in the visual system: electrophysiological signals were recorded from the primary visual area 18 in the anaesthetized cat, while a hierarchically higher area, the posterior middle suprasylvian (pMS) sulcus, was thermally deactivated. This deactivation silences the feedback signals from pMS to area 18. The deactivation was performed unilaterally – on the same (ipsilateral) and on the opposite (contralateral) hemisphere to the recording – as well as bilaterally. The effect of the pMS deactivation was probed on multiple temporal and spatial scales: on the singlecell level, spike rates and tunings were assessed for both multi- and single-unit data. On the level of local functional network connectivity, multiple approaches were explored to extract different aspects of the deactivation effects. Parallel Factor Analysis (PARAFAC) was applied as a means to scan the effect of deactivation of pMS on cross-correlations between multi-units in area 18 over all recording sessions. PARAFAC proved to be a suitable approach to extract the deactivation effects. To the author's knowledge, this is the first application of PARAFAC to electrophysiological spike data. Functional networks were created based on joint-spike events by application of and effective connectivity was computed based on the application of a Generalized Linear Model (GLM) to the spike data. Network connectivity was then assessed with a variety of graph theoretical measures. After excluding randomness (in the sense of Erdos-Renyi graphs), connection density was compared for short and long-range connections, and units with a similar or differing orientation or direction preference, respectively. Entire graphs were examined for similarity between different experimental conditions, based on a test using the Hamming distance. Effects of the pMS deactivation on the local network structure in area 18 could be observed on all probed spatial and temporal scales, namely LFP and spike rates, as well as millisecond-precise synchrony based on NeuroXidence, and directed connectivity based on the GLM. Overall, the results were in line with previous findings in the context of studies on the visual hemineglect in the anaesthetized cat: ipsi- and bilateral pMS deactivation led to lower activity and connectivity levels as compared to the warm condition, while contralateral pMS deactivation left rates and coordinated activity largely unaltered. However, a large variability of the deactivation effects was observed. In order to find a possible correlate for this variability, the data were divided into three subsets based on the level of activity in the LFP prior to thermal deactivation of pMS. The group with the highest initial power was observed to be the most dynamic and showed the largest effects during deactivation, both for rates and correlated activity. This raised the notion that the presence of low or medium initial activity could be an indication that the global network was partially disconnected even before the deactivation of pMS, and for this reason did not exhibit the same extent of deactivation effects as compared to the high initial gamma group. Thus it is concluded that global brain state bears a significant relevance for local networks and should be taken into account for any study of local connectivity in the brain.