A general theory of coherence between brain areas

Abstract

What does neuronal coherence tell us about neuronal communication? Does coherence between field potentials (e.g. LFP, EEG, MEG) reflect spiking entrainment or coupling between oscillators? Is it a mechanism for communication between brain areas, or a byproduct of interareal connectivity? We hypothesized that interareal coherence is explained by the fact that outputs from one cortical area give rise to synaptic inputs in the same brain area, and correlated synaptic inputs in another area. Our mathematical analysis demonstrates that coherence between a sending and receiving area is precisely predicted from only two parameters: Interareal connectivity and oscillatory synchronization in the sending area. This model predicts narrow-band coherence even in case of a flat transfer function and in the absence of spiking entrainment in a downstream area, and reproduces frequency-specific Granger-causality patterns between brain areas (gamma feedforward, beta feedback). In general, we find that Granger-causality between field potentials is dominated by oscillatory synchronization in the sending area, whereas spiking entrainment is determined by the resonant properties of the receiver. Our model accurately reproduces LFP-LFP beta-coherence between macaque areas 7B and F5 in the absence of beta phase-locking within area F5. Together, these findings uncover a precise mechanistic model of interareal coherence as a (by)product of connectivity and power.

Ultra-sparse Connectivity within the Lateral Hypothalamus

“…Because the LH does not contain a gamma generator ( Figure 4 ) needed to receive and send gamma-rhythmic communications, an information transfer protocol based on gamma coherence [ 28 , 43 ] is unlikely to be utilized by the LH. Gamma oscillations seen in vivo with local field potential recordings in the LH [ 44 ] could arise from the many passing axon bundles such as the fornix or incoming synapses [ 45 ]. Dense connectivity in the neocortex is based on inhibitory interneurons that migrate into the cortical scaffold from the ganglionic eminences during development [ 46 ].…”

Predictive coding of natural images by V1 activity revealed by self-supervised deep neural networks

“…A number of groups have hypothesized that γ-(30-80Hz) and β-synchronized (13-30Hz) firing reflects prediction error and prediction signaling, respectively (30; 31; 32; 33). Indirect evidence for this hypothesis comes from the finding that feedforward and feedback V1-V2/V4 communication are particularly strong in the γand β-frequency band, respectively (34; 35), (but see (36)). In contrast, it has been hypothesized that V1 γ-synchrony reflects the extent to which RF inputs can be predicted from the surround (37).…”

Symmetric and Asymmetric Causal Neural Firing Interactions across Striatum, Anterior Cingulate and Prefrontal Cortex during Cognitive Control

“…While the causality measures derived above may suggest the direction and possible content of information transmitted between two regions, they do not indicate how neuronal firing changed during the time period of significant directionality. One possibility is that firing increases in one region drive synaptic activity in the target region and increases the likelihood of neural firing in the target population (Schneider et al, 2020). However, synaptic drive of the target area could also target inhibitory cell populations and hence decrease overall firing probability (Medalla et al, 2009;.…”

“…Alternatively, the directionality of firing rate changes could be due to a third brain area that modulates firing rate increases or decreases of each of the recorded brain areas independently (Qao et al, 2020). These scenarios could underlie the Granger causal modulations albeit with gross variations in overall firing rate modulations during the periods of directional interactions (Schneider et al, 2020). The averaged activity time course for units associated with significant causality clusters was correlated against causality measures calculated between regions at cue (left panels) and feedback (right panels) periods (positively correlated units shown above).…”

“…However, local field potentials reflect not only how synaptic events from a remote, afferent source modulates a local area but inevitably also reflect local sources from neighboring neurons. Such a synaptic mixture of local and remote synaptic influences makes it ambiguous to interpret phase specific directional transfer and gives rise to spurious directional indices in the absence of true synaptic drive of a sending to a receiving area (Schneider et al, 2020). To prevent the difficulties interpreting field potential effects it is therefore desired to quantify directional exchange of information directly from neuronal spiketrain interactions.…”