Recovering arrhythmic EEG transients from their stochastic interference

Traditionally, the electroencephalogram (EEG) has been understood as arising from rhythmic neuronal oscillators with varying degrees of synchronisation. Alternative insights, however, highlight the arrhythmic nature of the EEG, primarily inferred from broadband properties like the ubiquitous 1/f spectrum. From the analysis of EEG simulations based on stochastic pulse superposition, we identified mathematical relations between the statistical features of the superposition signal and the shape of the underlying pulse(s), allowing us to develop a new method for recovering EEG transient components from their stochastic interference. Applying this approach to spontaneous mouse EEG recordings sampled at 5 kHz during the sleep-wake cycle, we discovered unique patterns that unambiguously identified all major behavioural states. These patterns are composed of fast transients with temporal features comparable to those observed in Local Field Potentials, which may help us to unify our understanding of neuronal dynamics across spatiotemporal scales.