Abstract
I present a systematic evaluation of different types of metrics, for inferring magnitude, amplitude, or phase synchronization from the electroencephalogram (EEG) and the magnetoencephalogram (MEG). I used a biophysical model, generating EEG/MEG-like signals, together with a system of two coupled self-sustained chaotic oscillators, containing clear transitions from phase to amplitude synchronization solely modulated by coupling strength. Specifically, I compared metrics according to five benchmarks for assessing different types of reliability factors, including immunity to spatial leakage, test-retest reliability, and sensitivity to noise, coupling strength, and synchronization transition. My results delineate the heterogeneous reliability of widely used connectivity metrics, including two magnitude synchronization metrics [coherence (Coh) and imaginary part of coherence (ImCoh)], two amplitude synchronization metrics [amplitude envelope correlation (AEC) and corrected amplitude envelope correlation (AECc)], and three phase synchronization metrics [phase coherence (PCoh), phase lag index (PLI), and weighted PLI (wPLI)]. First, the Coh, AEC, and PCoh were prone to create spurious connections caused by spatial leakage. Therefore, they are not recommended to be applied to real EEG/MEG data. The ImCoh, AECc, PLI, and wPLI were less affected by spatial leakage. The PLI and wPLI showed the highest immunity to spatial leakage. Second, the PLI and wPLI showed higher test-retest reliability and higher sensitivity to coupling strength and synchronization transition than the ImCoh and AECc. Third, the AECc was less noisy than the ImCoh, PLI, and wPLI. In sum, my work shows that the choice of connectivity metric should be determined after a comprehensive consideration of the aforementioned five reliability factors.
Highlights
My results delineate the heterogeneous reliability of widely used connectivity metrics, including two magnitude synchronization metrics [coherence (Coh) and imaginary part of coherence (ImCoh)], two amplitude synchronization metrics [amplitude envelope correlation (AEC) and corrected amplitude envelope correlation (AECc)], and three phase synchronization metrics [phase coherence (PCoh), phase lag index (PLI), and weighted PLI]
Connectivity metrics (AECc, ImCoh, PLI, and weighted PLI (wPLI)) that are immune to spatial leakage effects showed relatively lower overlapping strength than those (AEC, Coh, and PCoh) affected by spatial leakage, in particular, the PLI and wPLI [Figs. 2(e) and 2(f)] and AECc [Fig. 2(e)]
I compared metrics according to a comprehensive evaluation framework including five benchmarks: (1) immunity to spatial leakage in both neural mass models (NMMs) and coupled Rössler attractors; (2) test–retest reliability in NMMs; (3) sensitivity to noise in coupled Rössler attractors; (4) sensitivity to the increase of coupling strength in both NMMs and coupled Rössler attractors; and (5) ability to trace the synchronization transition in both NMMs and coupled Rössler scitation.org/journal/cha attractors
Summary
EEG and MEG suffer from some inherent methodological challenges.[12,36,37,38,39,40,41,42] First, a single source produces a signal at multiple recording sites, known as spatial leakage, including volume conduction in the EEG recordings,[43] and field spread in the sensor space or (primary) signal leakage in source space MEG signals This effect can give rise to spurious estimates of functional connectivity.[12,44,45] This problem cannot be completely solved by projecting sensor space signals to source space using inverse modeling methods, such as beamformer[46] and minimum-norm source estimates.[47,48] Second, real EEG and MEG signals often contain large amounts of physiological, experimental, and external noise that can result in spurious or incorrect estimates of functional connectivity.[49–52] Third, a reliable connectivity metric should be able to provide repeatable connectivity estimates and, must have high test–retest reliability.[53–61]. Any synchronization metric that is believed to be able to accurately and reliably estimate the functional connectivity on the basis of EEG/MEG signals should be robust to the five confounding factors
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Chaos: An Interdisciplinary Journal of Nonlinear Science
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
