Arousal-informed neuronal avalanches operate near criticality

Cortical networks are believed to perform optimally when operating near a critical point, although testing this hypothesis at the single-cell level across extensive brain regions has proven challenging. Traditional methods using narrow fields of view often miss overlapping neural avalanches, leading to underestimations of subcritical conditions.

To address this, researchers employed mesoscopic two-photon calcium imaging in the awake mouse visual cortex to record thousands of neurons. By applying convolutional non-negative matrix factorization, they decomposed population activity into distinct, overlapping spatiotemporal avalanches. Contrary to expectations of a single dynamical state, they discovered that the cortex comprises multiple overlapping subnetworks distributed across a phase diagram at varying distances from the critical point.

Notably, arousal-linked avalanches propagate extensively and operate closest to the critical point, exhibiting an optimized dynamic range, minimized algorithmic complexity, and maximized temporal entropy. These findings reveal how the awake brain strategically orchestrates parallel near-critical subnetworks to facilitate efficient sensory computation.