by Benedetta Mariani (University of Padova, Padova Neuroscience Center).
The critical brain hypothesis has been extensively studied since the first experimental signature [1], because criticality optimizes important quantities for information processing. However, the hypothesis remains debated. On the one hand, the actual foundations of the hypothesis – the self-organization process and how it manifests during sensory input – have been investigated only in few works. On the other hand, most of the evidence for criticality comes from scale-free distributions of neuronal avalanches; however, they can be explained also with alternative mechanisms, and their experimental definition present some issues. Recently, an alternative method to measure brain criticality has been proposed. It is based on coarse graining the neural activity through a procedure that exploits Principal Component Analysis and studying the convergence of the distribution of the coarse-grained variables as far as the coarse graining increases: in the case of criticality, a convergence of the joint distribution to a non-Gaussian fixed point is expected.
In the present work, criticality is studied in the barrel cortex of anesthetized rats, both at rest and during sensory input (controlled stimulation of the whisker), and both on high frequency data (spikes from 27 electrodes array), and coarse sampled ones (Local Field Potentials from 220 electrodes array). To fit power laws on avalanches sizes and durations, a state of the art maximum likelihood statistical method, based on undersampling the data in order to decorrelate them, is applied. In both kinds of data, the analyzed neuronal avalanches result power law distributed and, together with power-law scaling, bumps of activity are present when the resting state is perturbed, marking synchronized events of activity that involve all the electrodes of the array. The resulting exponents, in particular as regards spikes data, are widespread along a dynamical scaling line, in analogy with recent results.
Then, going beyond neuronal avalanche analysis, the test based on Principal Component Analysis is applied to LFPs, which reveals the convergence to a non-gaussian fixed form, thereby showing the presence of strongly dependent variables which exhibit critical-like features.
Spotlight talk presented on October 7th 2020, at the Brain Criticality Virtual Conference 2020 (Plenz D., Chialvo D., de Arcangelis L. & Battaglia D. organizers)