Maximizing Sensory Dynamic Range by Tuning the Cortical State to Criticality

Abstract

Modulation of interactions among neurons can manifest as dramatic changes in the state of population dynamics in cerebral cortex. How such transitions in cortical state impact the information processing performed by cortical circuits is not clear. Here we performed experiments and computational modeling to determine how somatosensory dynamic range depends on cortical state. We used microelectrode arrays to record ongoing and whisker stimulus-evoked population spiking activity in somatosensory cortex of urethane anesthetized rats. We observed a continuum of different cortical states; at one extreme population activity exhibited small scale variability and was weakly correlated, the other extreme had large scale fluctuations and strong correlations. In experiments, shifts along the continuum often occurred naturally, without direct manipulation. In addition, in both the experiment and the model we directly tuned the cortical state by manipulating inhibitory synaptic interactions. Our principal finding was that somatosensory dynamic range was maximized in a specific cortical state, called criticality, near the tipping point midway between the ends of the continuum. The optimal cortical state was uniquely characterized by scale-free ongoing population dynamics and moderate correlations, in line with theoretical predictions about criticality. However, to reproduce our experimental findings, we found that existing theory required modifications which account for activity-dependent depression. In conclusion, our experiments indicate that in vivo sensory dynamic range is maximized near criticality and our model revealed an unanticipated role for activity-dependent depression in this basic principle of cortical function.

Spontaneous cortical activity is transiently poised close to criticality

“…In contrast, we found largely subcritical distributions during the awake state during the eyes open condition, which were replaced by more critical distributions during the closed eye condition. In addition, correlations during anesthesia were not stationary in our study despite having used the same anesthetic agent, similar to other studies [38,44,64]. These discrepancies may originate in different levels of anesthesia (0.8% isoflurane concentration in this study, 1–2% in [22] potentially influencing the appearance of state fluctuations, temporal resolution of spike recordings (fast electrophysiological recordings vs. slow calcium imaging) or recording site (primary visual cortex vs. frontal cortex).…”

“…In contrast, desynchronized activity is associated with active behavior [41,42], responses to visual stimuli [43] and suggested to be a neuronal correlate of attention [35]. In line with this hypothesis, putative transitions between critical, supercritical and subcritical dynamics have been observed during anesthesia [20,44], shifts between anesthesia and the physiological sleep-wake cycle [21,22,45–48], prolonged waking [49], and in a rest versus task setting [50]. Here, we tested whether the signature of criticality at the level of collective spiking activity remains robust across different cortical states or fluctuates together with changing cortical dynamics recorded in the primary visual cortex of the anesthetized cat and awake monkey using classical criticality markers and data modeling.…”

Landau–Ginzburg theory of cortex dynamics: Scale-free avalanches emerge at the edge of synchronization

“…However, avalanches do not usually appear (or are not searched for) in such modeling approaches (see, however, (18,45,46)). …”

Strong long‐range temporal correlations of beta/gamma oscillations are associated with poor sustained visual attention performance

“…A system operating near criticality is poised between ordered (sub-critical) and disordered (super-critical) states, optimally combining the ability to form patterns while also responding swiftly to input. Indeed, critical-state dynamics have been related to optimal information processing (Shew & Plenz, 2013) and capacity (Shew et al, 2009), high robustness against perturbations (Hahn et al, 2010), and largest dynamic range in sensory processing (Gautam et al, 2015;Kinouchi & Copelli, 2006). In a supercritical state, the brain would be in a state of hyperarousal, whereas in a subcritical state propagation of neuronal activation is suppressed.…”