Brain-wide spontaneous neural avalanches: Definition, functional dynamics and cognitive relevance

Although spontaneous activity is ubiquitous across multiple spatiotemporal scales, its functional organization and cognitive relevance remain poorly understood. Following the classic neuronal avalanche framework, a spontaneous avalanche is defined as consecutively active frames separated by inactive time bins. Hence, multiple distinct avalanches may be considered as one avalanche, thereby ignoring their spatial and temporal distinguishability. Furthermore, group-level power-law fitting of such neural avalanches is often performed to evaluate brain criticality (referring to a system perched between order and disorder) due to the limited recording length of macroscale neuroimaging (such as functional magnetic resonance imaging), and the functional representation of brain-wide neural avalanches is largely unexplored. To address these issues, we proposed large-scale neural avalanches as a single, spatially consecutive cascade pattern and further investigated their functional dynamics, network propagation, and association with task-evoked activity. Compared with the conventional inactive-bin definition, our current approach is more favorable to power-law fitting of avalanche size and duration distributions at the individual level. We also demonstrated that participants whose brain activities were close to the critical point tend to have higher cognitive abilities. Notably, the ratio of neural avalanches that evolved from primary sensory to association networks negatively correlated with cognitive abilities. Moreover, the geometric distance between low-dimensional representations of task-evoked activity and spontaneous avalanches was associated with behavioral performance. This study not only provides a promising avenue for measuring avalanche criticality based on human whole-brain neuroimaging, but also suggests that spontaneous neural avalanches and their low-dimensional representations contribute to human cognitive abilities.