1. Cross-Frequency Synchrony Correlation between PS and CFS hubs
Large-scale synchrony within and across frequencies in resting-state SEEG data
Felix Siebenhühner1, Gabriele Arnulfo2, Lino Nobili3, Matias Palva1, Satu Palva1
1: Neuroscience Center, University of Helsinki; 2: Universita degli Studi di Genova; 3: Claudio Munari Epilepsy Surgery Centre, Niguarda Hospital, Milan
Phase Synchrony
Connection Density
Strength
1:2 Connection Density
1:3 Connection Density
Significant CFS was observed at 1:2 between q and a and between a and b bands and 1:3 between a and b/g
Significant CFS was reduced, but still showed same patterns with very conservative controlling for spurious CFS (right)
Connection Density
without spurious control
with spurious control
Strength
Cross-Frequency Synchrony
In 1:2 CFS, strength and connection density followed similar patterns for all functional subsystems
In 1:3 CFS, strength is similar across subsystems
1:2 Strength
1:3 Strength
Introduction
Long-range within-frequency (1:1) phase synchrony (PS) of neuronal oscillations regulates neuronal communication among human brain areas [Fries 2015].
Large-scale PS networks are crucial for cognitive functions [Palva 2012, Siegel 2012, Petersen 2015].
Neuronal activity during task performance is also characterized by n:m cross-frequency phase synchrony (CFS), which may coordinate processing across distinct frequencies [Palva 2005, Siebenhühner 2016].
Yet, concerns have been raised over the possibility of CFS observations being spurious [Gerber 2016, Scheffer-Tezeira 2016]
Observations of long-range CFS in human resting-state data have been sparse [Palva 2005], and it has not been investigated whether it connects PS networks
Hypothesis:
Long-range CFS connects networks of cortical oscillations in resting state
Frequency [Hz] K
Frequency [Hz]
PLV
Frequency [Hz]
PLV
Frequency [Hz] K
Frequency [Hz] K
1:1 phase synchrony f1
n:m-phase synchrony (here 1:3) f2
Frequency [Hz]
PLV
Frequency [Hz] K
Frequency [Hz]
PLV
Frequency [Hz] K
Materials and Methods
Stereotactical EEG (SEEG) data were recorded from epileptic patients undergoing pre-surgical clinical assessment
Data was recorded from monopolar (with shared white matter reference) local field potentials (LFPs) from brain tissue with platinum–iridium, multi-lead electrodes
Closest White Matter (CW) scheme was used for referencing [Arnulfo 2015]
38 subjects with 116±17 channels each, recording time 10 min
Only cortico-cortical connections between non-epileptic channels were analyzed
Data were filtered into 31 center frequencies from 3.3 – 100 Hz using Morlet wavelets
1:1 phase synchrony (PS) and cross-frequency phase synchrony (CFS, for ratios 1:2 – 1:6) were computed with the phase-locking value (PLV) over the whole time period (time windows with spiky/epileptic activity excluded)
Surrogate values were constructed by shifting time series by a random number of samples
The connection density K was computed as the fraction of connections above 2.42*mean(surrogates) over the total number of possible connections
Electrode locations were assigned to the functional subsystems defined by Yeo et al. [Yeo 2011]
Frequency [Hz]
PLV
Frequency [Hz] K
Both strength and connection density K of phase synchrony peak in a band
a peak can be observed in all functional subsystems
Subsytems:
Vis = Visual, SM = Somatomotor,
DA = Dorsal Attention, VA = Ventral Attention,
Lim = Limbic, FP = Frontoparietal, DM = Default Mode
Significant Cross-Frequency synchrony (CFS) is observed in human SEEG data in resting state
Observed CFS can not be explained by possibly spurious CFS
Strength and Connection density of both PS and CFS tend to be similar across functional subsystems
Hubs of the low-frequency PS networks and 1:2 CFS networks are correlated
Conclusions
Correlation between PS and CFS hubs
Controlling for possible spurious CFS
Theoretically, spurious CFS might arise as a result of PS and local CFS and of harmonic or subharmonic components of non-sinusoidal signals
Importantly, both sinusoidal and non-sinusoidal signals may be accompanied by true CFS.
We developed a conservative approach for detecting true CFS. We posit that spurious CFS must always be accompanied with 1:1 PS. In our model scheme of this approach all full “triangle motifs” of PS, local CFS & long-range CFS are discarded; this rules out all spurious CFS observations and also may discard a number of true CFS connections.
Degree
Betweenness Centrality
References
Arnulfo G, et al. (2015): ‘Phase and amplitude correlations in resting-state activity in humanstereotactical EEG recordings’, NeuroImage, 112, pp
Fries P (2015): ‘Rhythms for Cognition: Communication through coherence’, Neuron, 88, pp
Gerber EM, et al. (2016): ‘Non-Sinusoidal Activity Can Produce Cross-Frequency Coupling in Cortical Signal in Absence of Functional Interaction between Neural Sources’, PLoS One, e
Palva JM, et al. (2005): ‘Phase synchrony among neuronal oscillations in the human cortex’, Journal of Neuroscience, 25/15, pp
Palva S, Palva JM (2012): ’Discovering oscillatory interaction networks with M/EEG: challenges and breakthroughs’, Trends in Cognitive Sciences, 16/4, pp
Petersen SE, Sporns O (2015): ‘Brain Networks and Cognitive Architectures’, Neuron, 88, pp
Scheffer-Teixerira R, Tort ABL (2016): ‘On cross-frequency phase-phase coupling between theta and gamma oscillations in the hippocampus’, eLife 5:e20515
Siebenhühner F, et al. (2016): ‘Cross-frequency synchronization connects networks of fast and slow oscillations during visual working memory maintenance’, eLife 5:e13451
Siegel M, et al. (2012): ‘Spectral fingerprints of large-scale neuronal interactions’, Nature Reviews Neuroscience, 13, pp
Yeo BT, et al. (2011): ‘The organization of the human cerebral cortex estimated by intrinsic functional connectivity’, Journal of Neurophysiology 106, pp. 1125–1165.
This work was supported by: Academy of Finland, Helsinki University Research Funds, Doctoral Program of Brain & Mind of the University of Helsinki.
Frequency [Hz]
Pearson’s r
True CFS
Spurious CFS A B A B A B A B A B A B
Regular b signal
Non-zero-mean,
amplitude-
modulated b
b signal
a signal
m signal
(1)
(2)
(3)
(1) A B A B A B A B A B A B A B
(2)
Hubs of PS networks, as identified by node degree and betweenness centrality, of q, a and b bands and corresponding 1:2 and 1:3 CFS networks are significantly correlated across subjects (Pearson correlation, p=0.05, Benjamini-Hochberg corrected)
Significant correlations were also found in most individual subjects for both metrics A B A B A B A B
True CFS/PS
Spurious CFS/PS
(3)
Sinusoidal a
Non-sinusoidal a
Sinusoidal b
Harmonic b
Non-sinusoidal b
Subharmonic a
True positives
False negatives
True negatives