Abstract
The topological architecture of brain connectivity has been well-characterized by graph theory based analysis. However, previous studies have primarily built brain graphs based on a single modality of brain imaging data. Here we develop a framework to construct multi-modal brain graphs using concurrent EEG-fMRI data which are simultaneously collected during eyes open (EO) and eyes closed (EC) resting states. FMRI data are decomposed into independent components with associated time courses by group independent component analysis (ICA). EEG time series are segmented, and then spectral power time courses are computed and averaged within 5 frequency bands (delta; theta; alpha; beta; low gamma). EEG-fMRI brain graphs, with EEG electrodes and fMRI brain components serving as nodes, are built by computing correlations within and between fMRI ICA time courses and EEG spectral power time courses. Dynamic EEG-fMRI graphs are built using a sliding window method, versus static ones treating the entire time course as stationary. In global level, static graph measures and properties of dynamic graph measures are different across frequency bands and are mainly showing higher values in eyes closed than eyes open. Nodal level graph measures of a few brain components are also showing higher values during eyes closed in specific frequency bands. Overall, these findings incorporate fMRI spatial localization and EEG frequency information which could not be obtained by examining only one modality. This work provides a new approach to examine EEG-fMRI associations within a graph theoretic framework with potential application to many topics.
Introduction
In modern neuroscience, brain graphs serve as the standard mathematical framework for mapping how distinct regions communicate. However, researchers typically face a trade-off between modalities: fMRI provides high spatial precision but low temporal resolution, while EEG captures millisecond-level electrical oscillations but remains spatially blurry. Because these data streams are often analyzed in isolation, the complex "cross-talk" between electrical frequencies and spatial blood-oxygen patterns is frequently missed. This study introduces a novel framework to fuse concurrent EEG and fMRI data into a unified multi-modal brain graph, allowing for a more comprehensive view of brain architecture across different behavioral states.
Methods and Materials
Dataset
The study utilized concurrent EEG-fMRI data from 25 healthy subjects (average age 29). Data were collected during 8.5-minute resting-state sessions for both Eyes Open (EO) and Eyes Closed (EC) conditions using a 1.5T Siemens Sonata scanner and a 32-channel MR-compatible EEG system.
Analytical Pipeline
The construction of the multi-modal graph followed a multi-step pipeline:
- Node Definition: 54 "intrinsic connectivity networks" were identified via Group Independent Component Analysis (ICA) of fMRI data, combined with 30 EEG electrode locations to form an 84-node graph.
- Signal Alignment: EEG spectral power was calculated across five bands (delta, theta, alpha, beta, and low gamma) and convolved with a Hemodynamic Response Function (HRF) to align with the slower fMRI signals.
- Graph Construction: Edges were defined by Pearson correlations between the fMRI ICA time courses and the processed EEG spectral power time courses.
- Static vs. Dynamic Analysis: Static graphs utilized the full recording duration, while dynamic graphs employed a 40-second sliding window method to observe temporal fluctuations in connectivity.
Key Results
The "Eyes Closed" Efficiency Boost
A primary finding was that graph metrics, including global efficiency and clustering coefficients, were significantly higher during the Eyes Closed (EC) state compared to Eyes Open (EO). This suggests that the brain enters a more integrated and efficient communication state when visual input is removed, particularly within the alpha and beta frequency bands.
Frequency-Specific Connectivity
The study demonstrated that EEG-fMRI associations are highly frequency-dependent. For instance, the beta band showed significantly higher clustering coefficients ($P < 0.01$) during EC in positive connection graphs. In negative connection graphs (anticorrelations), global efficiency and connectivity strength were notably higher in the alpha and beta bands during EC.
Dynamic Connectivity Variance
Dynamic analysis revealed that multi-modal connectivity is non-stationary. The variance and amplitude of low-frequency fluctuations in graph metrics were significantly higher for delta and theta frequencies, indicating that lower-frequency rhythms drive much of the brain's dynamic network reconfiguration.
Discussion and Future Directions
While this framework successfully integrates spatial and spectral data, several limitations remain. There is an inherent asymmetry between the 30 EEG nodes and 54 fMRI nodes due to differing signal-to-noise ratios and physical origins. Additionally, the use of Pearson correlation may include redundant information that more advanced methods, such as partial correlation, could refine.
Future research should aim to scale this framework to high-density EEG (128+ channels) and high-field fMRI (7T) to provide more granular insights. Nonetheless, this work establishes a foundational proof-of-concept for moving toward truly integrated, multi-dimensional maps of the human brain.
Summary updated