Mutual Information Reveals Temporal Dynamics of Functional Connectivity in Awake Auditory Cortex

Sensory inputs are encoded in neuronal populations operating as an interconnected network. Sequential impulses of acoustic stimulus are encoded by cortical neurons in the primary auditory cortex (A1) with a temporal precision of 20~50 msec. Therefore, it can be speculated that the functional connectivity (FC) of A1 neurons can be altered at a similar time resolution, which gives rise to an intracortical network undergoing dynamic changes. In this study, we quantified FC as a function of time using mutual information (MI), which is known for its strength in better representing non-linear characteristics in neurophysiological data. We hypothesized that acoustic stimuli have a short-latency impact on FC and neuronal networks in A1, which can be characterized with cross-channel MI.

We recorded local field potentials (LFPs) in left A1 from 32-channel electrode matrices. Relationships between the LFP activities from any two different recording channels was quantified by an MI value as a measure of FC. Collectively, pairwise cross-channel MI values served as a characterization of the neuronal networks in the recorded cortical area. Neuronal activities were driven by acoustic stimuli, which consisted of multiple clicks in each recording trial. In addition to periodic and aperiodic clicks, we also presented tone pips, conspecific vocalizations, and light flashes in some recording sessions as well. At varying stimulus lags, MI values were compared, for both individual channel pairs and as in a matrix, to a baseline regardless of stimulus lags.

A preliminary analysis in 4 electrode matrices showed consistently that click stimuli disrupted FC at stimulus lags less than 50 msec. For these specific lags, individual pairs of channels demonstrated drastic changes in the joint distribution of LFPs from the baseline, as indicated by elevated Chi-square statistics. Similarly, pairwise MI matrices were re-ordered as compared to the baseline, which was revealed by a decrease in spearman correlations at similar stimulus lags.

Our current results suggested that aperiodic click stimulations disrupt baseline FC in A1, which may in turn impact neuronal networks and, consequently, cortical desynchronizations. Next, phase information derived from narrowband LFPs can be incorporated to examine FC at a different time scale. The effect of acoustic periodicity, carrying frequency, naturalistic familiarity, and cross-modal modulations will also be investigated. Taking advantage of information theory, this study will provide deeper insight into the temporal dynamics of cortical connectivity in A1 and expand our understanding of the auditory cortex in humans, with normal and abnormal hearing, through non-invasive electrophysiological techniques (such as MEG and ECoG).