Unpacking Transient Event Dynamics in Electrophysiological Power Spectra

Abstract

Electrophysiological recordings of neuronal activity show spontaneous and task-dependent changes in their frequency-domain power spectra. These changes are conventionally interpreted as modulations in the amplitude of underlying oscillations. However, this overlooks the possibility of underlying transient spectral ‘bursts’ or events whose dynamics can map to changes in trial-average spectral power in numerous ways. Under this emerging perspective, a key challenge is to perform burst detection, i.e. to characterise single-trial transient spectral events, in a principled manner. Here, we describe how transient spectral events can be operationalised and estimated using Hidden Markov Models (HMMs). The HMM overcomes a number of the limitations of the standard amplitude-thresholding approach to burst detection; in that it is able to concurrently detect different types of bursts, each with distinct spectral content, without the need to predefine frequency bands of interest, and does so with less dependence on a priori threshold specification. We describe how the HMM can be used for burst detection and illustrate its benefits on simulated data. Finally, we apply this method to empirical data to detect multiple burst types in a task-MEG dataset, and illustrate how we can compute burst metrics, such as the task-evoked timecourse of burst duration.

Dissecting Transient Burst Events

“…The spatial domain is commonly reduced by extracting the time series from a single spatial location, or summarising the time series of several spatial locations through their average or a linear combination [e.g., first principle component (PC)]. Similarly, the spectral domain is often reduced by selecting a single peak frequency, averaging the signal or amplitude envelope within a specified frequency band, or by fitting a state-wise frequency profile from a time-delay embedded or autoregressive-based Hidden Markov Model (HMM) [ 8 ].…”

Oscillatory Bursting as a Mechanism for Temporal Coupling and Information Coding

“…They show rhythmicity peaks detected in ongoing sensorimotor signals that are not visible using conventional power analysis, suggesting that rhythmicity measures are more suitable for identifying neuronal oscillations. Another approach to the detection and characterization of neuronal rhythms uses Hidden Markov Models (HMMs) to overcome some of the limitations of the amplitude-threshold approaches by avoiding a direct amplitude envelope threshold (Quinn et al, 2019 ). The HMM represents the signals as a system that moves through a set of discrete states, with each state having a probability of being “on” at each time point.…”

“…Thus, the thresholding procedure is applied to the probabilities rather than the signals themselves. In addition, using temporal regularization, HMM can avoid state transitions due to small changes in the envelope close to the threshold [see Figure 2 in Quinn et al ( 2019 )]. One of the downsides of this method is that a fixed number of states must be defined in advance.…”

The hidden state-dependency of transcranial alternating current stimulation (tACS)

“…1 a,b). However, it is also possible that changes in spectral power arise from a more frequent occurrence of brain-states with comparably higher oscillatory power, without necessarily changing the actual amplitude of oscillations within each state 30 (Fig. 1c).…”