A Transient Model for Neuronal Oscillations

Background: Classical signal processing methods for neuronal oscillations, such as EEG and LFP, usually rely on the strong restrictive assumptions of stationarity and ergodicity. However, electrical potentials from the brain are well known for displaying statistical properties that vary over time giving rise to transient patterns [1]. Moreover, decomposition methods in the literature not only utilize these erroneous assumptions, but they also introduce processing windows that smear the temporal information encoded in the structure of the signal. We propose a transient model that poses a single EEG trace as the result of the noisy addition of reoccurring, transient events over time and frequency bands (Fig. 1). This enables to preserve the superior temporal resolution of EEG while incorporating additional features to the model, such as amplitude, frequency, duration, and modulation-based measures related to relevant phasic events. These measures can, consequently, be utilized to interpret behavior and stimuli encoding in the brain at a macroscopic level beyond single unit paradigms.