Discovering recurring patterns in electrophysiological recordings.

Background: Fourier-based techniques are used abundantly in the analysis of electrophysiological data. However, these techniques are of limited value when the signal of interest is non-sinusoidal or non-periodic.

New Method: We present sliding window matching (SWM): a new data-driven method for discovering recurring temporal patterns in electrophysiological data. SWM is effective in detecting recurring but unknown patterns even when they appear non-periodically.

Results: To demonstrate this, we used SWM on oscillations in local field potential (LFP) recordings from the rat hippocampus and monkey V1. The application of SWM yielded two interesting findings. We could show that rat hippocampal theta and monkey V1 gamma oscillations were both skewed (i.e. asymmetric in time), rather than being sinusoidal. Furthermore, gamma oscillations in monkey V1 were skewed differently in the superficial compared to the deeper cortical layers. Second, we used SWM to analyze responses evoked by stimuli or microsaccades even when the onset timing of stimulus or microsaccades was unknown.

Comparison With Existing Methods: We first validated the method on simulated datasets, and we checked that for recordings with a sufficiently low noise level the SWM results were consistent with results from the widely used phase alignment (PA) method.

Conclusions: We conclude that the proposed method has wide applicability in the exploration of noisy time series data where the onset times of particular events are unknown by the experimenter such as in resting state and sleep recordings.