Back to the abstracts page
Back to the NCCD 2015 home page
Local inhibitory plasticity tunes critical brain dynamics and enables emergence of functional brain networks
Peter John Hellyer1,2, Barbara Jachs1, Claudia Clopath3 and Robert Leech1
1Computational, Cognitive and Clinical Neuroimaging Laboratory, Imperial College London 2Center for Neuroimaging Sciences, King's College London 3Department of Bioengineering, Imperial College London

Activity in the human brain is frequently characterized in terms of spontaneous neural fluctuations that are present even in the absence of an explicit task. Evidence for such ceaseless dynamics have been reported across spatial and temporal scales and contain both spatial and temporal structure. A number of theoretical frameworks have been developed to describe these highly flexible dynamics, [1-5]. One remarkable feature of spontaneous neural dynamics is that they are present in all but the most extreme of brain states e.g., deep anaesthesia [6]. Typically, these dynamics only occur within a narrow window of parameters. Outside this, models will often fall into a pathological state, in contrast to the actual brain that maintains dynamics in the face of changing external environment as well as with structural changes. A theoretical account of how spontaneous dynamics emerge from the brain must therefore include mechanisms that tune the dynamics. E/I mediated homeostasis has been shown to induce critical dynamics within a simple model with mean-field approximations of coupled excitatory and inhibitory neurons [7, 8], suggesting that inhibitory homeostatic plasticity may provide a mechanism to stabilize brain dynamics at the macroscopic level, and may be relevant for understanding macroscopic patterns of brain activity.

We use a mean field model of macroscopic brain activity, adapted from the Wilson-Cowan model [9]. The model is based on an empirically defined network topology derived from white matter connectivity between 66 cortical regions [10]. Each node of the model contains a pool of inhibitory and a pool of excitatory neurons; nodes receive input from other excitatory nodes weighted by the strength of long-distance white matter tracts. We adapt the model by adding a simple local learning rule that adjusts the inhibitory weight within a node such that excitation of the node equals a target value. We then evaluate the effect of adding inhibitory plasticity into the computational model, both by estimating measures of complex dynamics such as 'neural avalanches' [11]. Furthermore, we directly compare the ability of the model to generate patterns of intrinsic functional connectivity, by evaluating time-dependent functional connectivity of the model against empirical resting state activity of the brain estimated using fMRI [12].

We demonstrate that the addition of inhibitory plasticity leads to the following results: 1) it regulates the E/I balance, both locally and globally; 2) it enhances dynamics (resulting in dynamics consistent with criticality); 3) the dynamics do not fall into obviously pathological states; 4) the activity patterns reflect the underlying structural information in a way which maximally matches empirically-measured functional connectivity from resting state fMRI in humans. Therefore, this local inhibitory learning rule provides a mechanism whereby the underlying brain network topology can shine through while also ensuring rich, spontaneous dynamics.

This work presents a simple, local, biologically plausible inhibitory mechanism that allows stable dynamics to emerge in the brain and be maintained over time, which facilitates the formation of functional connectivity networks.

[1] Bressler, S.L. and J.A. Kelso, Cortical coordination dynamics and cognition. Trends Cogn Sci 5(1):26-36 (2001).
[2] Kelso, J.A.S., Multistability and metastability: understanding dynamic coordination in the brain. Phil. Trans. R. Soc. B 367:906-918 (2012).
[3] Tognoli, E. and J.A.S. Kelso, The Metastable Brain. Neuron 81(1):35-48 (2014).
[4] Shew, W.L. and D. Plenz, The functional benefits of criticality in the cortex. NRO 19(1):88-100 (2013).
[5] Friston, K.J., Transients, metastability, and neuronal dynamics. NeuroImage 5:164-171 (1997).
[6] Scott, G., et al., Voltage imaging of waking mouse cortex reveals emergence of critical neuronal dynamics. J. Neurosci 34(50):16611-20 (2014)
[7] Cowan, J.D., et al., Self-organized criticality in a network of interacting neurons. Journal of Statistical Mechanics: Theory and Experiment 2013(04):P04030 (2013)
[8] Magnasco, M.O., O. Piro, and G.A. Cecchi, Self-tuned critical anti-Hebbian networks. Phys Rev Lett 102(25):258102 (2009).
[9] Wilson, H.R. and J.D. Cowan, Excitatory and inhibitory interactions in localized populations of model neurons. Biophys J 12(1):1-24. (1972)
[10] Hagmann, P., et al., Mapping the structural core of human cerebral cortex. PLoS biology 6(7):159 (2008)
[11] Beggs, J.M. and D. Plenz, Neuronal avalanches are diverse and precise activity patterns that are stable for many hours in cortical slice cultures. J. Neuroscience 24(22):5216-5229 (2004)
[12] Honey, C.J., et al., Predicting human resting-state functional connectivity from structural connectivity. Proceedings of the National Academy of Sciences 106(6):2035--2040 (2009)