Skip to main content

Advertisement

Detecting network states in white noise

Article metrics

  • 360 Accesses

Nonstationarity of neural dynamics is a ubiquitous property that is crucial to understanding many key phenomena of both healthy and diseased brain function, including circadian rhythms, dynamics of epileptic activity as well as cognitive processing. Detecting switching of brain states has recently become of growing interest in the human brain neuroimaging community. However, from the data analysis/modelling perspective the task is quite challenging, and competing approaches exist [1]. One widely adopted approach is the use of clustering methods in the temporal domain to detect temporally contiguous clusters of time points with a similar structure of some instantaneous property - e.g. neural activity or functional connectivity profile. While this approach may in principle help to explore the switching structure of brain dynamics, it comes with technical challenges related the presence of noise in both the dynamics and measurements. In particular, as we documented in a recent study [2], comparison of the results with an appropriate null hypothesis is necessary to avoid spurious detection of nonstationarity markers such as switching of neural network states.

We document this danger by applying an example analysis pipeline used in [3] to simulated EEG datasets. The simulated data are generated as realizations of temporally white noise process (either spatially uncorrelated or spatially correlated in a pattern corresponding to real EEG data). In each case, one hundred realizations of a 5 seconds long epoch of N = 20 'electrodes' (each of 2500 time points corresponding to 2ms sampling rate). A k-means clustering algorithm with k = 2 to 10 is applied to cluster the instantaneous synchronization likelihood matrix estimates with parameters as in [3]. The key observation stable across all setting is that the applied typical network switching analysis pipeline leads to spurious discovery of a multitude of network states in the stationary process realizations, with dominant state duration timescales of several tens to hundreds milliseconds qualitatively similar to the original results reported in [3]. These results suggest that observations of network switching should be always cautiously interpreted and tested against appropriate null models.

References

  1. 1.

    Hutchison RM, Womelsdorf T, Allen EA, Bandettini PAD, Calhoun V, Corbetta M, Duyn JH, Glover GH, Gonzalez-Castillo J, et al: Dynamic functional connectivity: Promise, issues, and interpretations. NeuroImage. 2013, 80: 360-378.

  2. 2.

    Hlinka J, Hadrava M: On the danger of detecting network states in white noise. Frontiers in Computational Neuroscience. 2015, 9: 11-

  3. 3.

    Betzel RF, Abell M, O'Donnell BF, Hetrick WP, Sporns O: Synchronization dynamics and evidence for a repertoire of network states in resting EEG. Frontiers in Computational Neuroscience. 2012, 6 (74): 1-13. doi:10.3389/fncom.2012.00074

Download references

Acknowledgements

We thank Martin Brunovský from Psychiatric Center Prague for providing sample EEG data. The research was supported by the Czech Science Foundation projects No. 13-23940S and No. 13-17187S.

Author information

Correspondence to Jaroslav Hlinka.

Rights and permissions

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Keywords

  • Network Switching
  • Synchronization Likelihood
  • Connectivity Profile
  • Neural Network State
  • Spurious Detection