Problem of vertical microstate topography in simultaneous EEG/fMRI data
Poster No:
1632
Submission Type:
Abstract Submission
Authors:
Tomáš Jordánek1,2, Radek Marecek2, Martin Lamos2
Institutions:
1Brno, Masaryk University, Faculty of Medicine, Brno, South Moravia, 2Brain and Mind Research Program, CEITEC Masaryk University, Brno, South Moravia
First Author:
Tomáš Jordánek
Brno, Masaryk University, Faculty of Medicine|Brain and Mind Research Program, CEITEC Masaryk University
Brno, South Moravia|Brno, South Moravia
Brno, Masaryk University, Faculty of Medicine|Brain and Mind Research Program, CEITEC Masaryk University
Brno, South Moravia|Brno, South Moravia
Co-Author(s):
Introduction:
Simultaneous recording of electroencephalography (EEG) and functional magnetic resonance (fMRI) is connected with strong artifacts mainly in EEG. Analysis of EEG microstates (MS) (Lehmann et al., 1987) as a method for spatio-temporal analysis of the EEG is very sensitive to residual artifacts in the data and its results can be affected.
Analyzing simultaneous EEG/fMRI data with MS analysis reveals a specific left-right topography with strong vertical line in the middle – we call it vertical topography (VT). It can be found in papers analyzing EEG/fMRI data (Agrawal et al., 2022; Bréchet et al., 2019), as well as analyzing EEG recorded in a shielded room (Custo et al., 2017).
Here we show that the VT cannot be considered as a manifestation of neural activity. Rather, we propose that it represents artificial variability in a data introduced mainly by movements of sensitive parts of EEG system during data acquisition.
Analyzing simultaneous EEG/fMRI data with MS analysis reveals a specific left-right topography with strong vertical line in the middle – we call it vertical topography (VT). It can be found in papers analyzing EEG/fMRI data (Agrawal et al., 2022; Bréchet et al., 2019), as well as analyzing EEG recorded in a shielded room (Custo et al., 2017).
Here we show that the VT cannot be considered as a manifestation of neural activity. Rather, we propose that it represents artificial variability in a data introduced mainly by movements of sensitive parts of EEG system during data acquisition.
Methods:
We collected resting state EEG data from 77 healthy controls acquired during fMRI (Dataset A) and in shielded cabin (Dataset B).
Further, we performed several testing measurements with spherical MR phantom covered with a wet towel and an EEG cap attached:
EEG/fMRI with same protocol as used for humans (Dataset C) and cabin EEG data (Dataset D).
All EEG recordings were done with MR-compatible HD-EEG cap (EGI 256-channel), sampling frequency 1 kHz, reference at Cz. After standard gradient artifacts and cardiobalistogram (CB) removal (Allen et al., 2000), the EEG data were revised and marked for large artifacts, filtered to 1–40 Hz, and processed by ICA to remove eye artifacts (and ECG in cabin data). Finally, bad channels were interpolated. Correction of CB and ICA was omitted in phantom data.
For each individual subject and phantom dataset, we extracted topomaps at global field power (GFP) peaks that were 100 times clustered with modified k-means algorithm to reveal most stable topographies. Resulting maps were 200 times clustered (k-means) in group analyses and optimal number of MS was based on metacriterion (Bréchet et al., 2019). Resulting maps were back fitted to the data.
Further, we performed several testing measurements with spherical MR phantom covered with a wet towel and an EEG cap attached:
EEG/fMRI with same protocol as used for humans (Dataset C) and cabin EEG data (Dataset D).
All EEG recordings were done with MR-compatible HD-EEG cap (EGI 256-channel), sampling frequency 1 kHz, reference at Cz. After standard gradient artifacts and cardiobalistogram (CB) removal (Allen et al., 2000), the EEG data were revised and marked for large artifacts, filtered to 1–40 Hz, and processed by ICA to remove eye artifacts (and ECG in cabin data). Finally, bad channels were interpolated. Correction of CB and ICA was omitted in phantom data.
For each individual subject and phantom dataset, we extracted topomaps at global field power (GFP) peaks that were 100 times clustered with modified k-means algorithm to reveal most stable topographies. Resulting maps were 200 times clustered (k-means) in group analyses and optimal number of MS was based on metacriterion (Bréchet et al., 2019). Resulting maps were back fitted to the data.
Results:
Dataset A: 6 MS as optimum, VT is present (1A) and is explaining 13,5 % of variance and covering 18,1 % of signal. We have revealed significant correlation (Pearson's) between amount of movement of subject (framewise displacement) and parameters of map 1A – GEV (r= 0.355; p=0.0015), Mean duration (r=0.293; p=0.0097), Time coverage (r=0.340; p=0.0077), Occurrence (r=0.302; p=0,0025).
Dataset B: 4 MS as optimum – so-called canonical MS, VT is not present (VT is present after segmentation to 6 clusters explaining 3,3 % of GEV and covering 8,0 % of EEG.)
Dataset C: 8 MS as an optimum, four of them (map 2C, 4C, 6C and 8C) are VTs and map 6C is the most dominant, explaining 60,1 % of GEV and covering 61,7 % of the signal.
Dataset D: 7 MS as an optimum, no VT is present even after segmentation up to ten clusters. 99 % of the EEG is unlabeled, data have no structure and represent noise only.
Dataset B: 4 MS as optimum – so-called canonical MS, VT is not present (VT is present after segmentation to 6 clusters explaining 3,3 % of GEV and covering 8,0 % of EEG.)
Dataset C: 8 MS as an optimum, four of them (map 2C, 4C, 6C and 8C) are VTs and map 6C is the most dominant, explaining 60,1 % of GEV and covering 61,7 % of the signal.
Dataset D: 7 MS as an optimum, no VT is present even after segmentation up to ten clusters. 99 % of the EEG is unlabeled, data have no structure and represent noise only.
·Resulting microstate topographies for each dataset
Conclusions:
Vertical topography affects analysis of MS with its very presence and also changes spatio-temporal parameters of canonical MS A and B (during fitting part VT can be falsely marked as A or B and conversely). Based on our analyses we suppose that VT is connected with movement of EEG cap and wires:
- In MR environment the VT is present in both human and phantom data – movement of subject, vibrations, heart activity; effect is increased by strong magnetic field
- Significant correlation between parameters of VT and amount of movement
- Presence of VT in human cabin EEG – subject's movement, heart activity; effect is lower without strong magnetic field
- VT is not present in cabin phantom EEG – phantom lies stationary
VT is very likely artificial topography, does not represent physiological activity - it is present in phantom data and it is absolutely dominant topography there.
- In MR environment the VT is present in both human and phantom data – movement of subject, vibrations, heart activity; effect is increased by strong magnetic field
- Significant correlation between parameters of VT and amount of movement
- Presence of VT in human cabin EEG – subject's movement, heart activity; effect is lower without strong magnetic field
- VT is not present in cabin phantom EEG – phantom lies stationary
VT is very likely artificial topography, does not represent physiological activity - it is present in phantom data and it is absolutely dominant topography there.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 1
Task-Independent and Resting-State Analysis 2
Keywords:
Electroencephaolography (EEG)
FUNCTIONAL MRI
Other - artifacts; microstates; simultaneous EEG/fMRI; phantom
1|2Indicates the priority used for review
Provide references using author date format
Allen, P.J., 2000. A Method for Removing Imaging Artifact from Continuous EEG Recorded during Functional MRI. NeuroImage 12, 230–239. https://doi.org/10.1006/nimg.2000.0599
Bréchet, L., 2019. Capturing the spatiotemporal dynamics of self-generated, task-initiated thoughts with EEG and fMRI. NeuroImage 194, 82–92. https://doi.org/10.1016/j.neuroimage.2019.03.029
Custo, A., 2017. Electroencephalographic Resting-State Networks: Source Localization of Microstates. Brain Connectivity 7, 671–682. https://doi.org/10.1089/brain.2016.0476
Lehmann, D., 1987. EEG alpha map series: brain micro-states by space-oriented adaptive segmentation. Electroencephalography and Clinical Neurophysiology 67, 271–288. https://doi.org/10.1016/0013-4694(87)90025-3