Introduction to Functional Connectivity

Functional connectivity refers to the statistical dependencies or temporal correlations between spatially separated brain regions. It reflects how different areas of the brain communicate and coordinate their activity, even when they are not directly connected by anatomical pathways.

Key Concepts

  • Not Anatomical: Functional connectivity does not imply direct physical connections (e.g., synapses or white matter tracts). It measures statistical relationships between neural signals.
  • Dynamic: Connectivity patterns change over time, depending on cognitive states, tasks, or pathological conditions.
  • Directionality: Functional connectivity can be unidirectional (A → B) or bidirectional (A ↔︎ B).
  • Frequency-Specific: Connectivity can occur in specific frequency bands (e.g., alpha, beta, gamma), each associated with different cognitive processes.

Functional connectivity provides insights into:

  • Brain Organization: How different regions work together to perform cognitive functions.
  • Cognitive Processes: How attention, memory, and perception emerge from distributed neural networks.
  • Neurological and Psychiatric Disorders: Disruptions in connectivity are linked to conditions like epilepsy, schizophrenia, Alzheimer’s disease, and autism.
  • Brain-Computer Interfaces (BCIs): Understanding connectivity can improve the design of BCIs for communication and control.

Phase-Based Measures

Phase synchronization is a key mechanism for functional connectivity, as it reflects temporal alignment of oscillations between regions.

  1. Inter-Site Phase Clustering (ISPC)
  2. Inter-Trial Phase Clustering (ITPC)
  3. Phase Locking Value (PLV)
  4. Phase Lag Index (PLI)
  5. Weighted Phase Lag Index (wPLI)
  6. Coherence

Amplitude-Based Measures

Amplitude-based measures capture correlations in power between brain regions, which may reflect co-activation rather than phase synchronization.

  1. Amplitude Envelope Correlation (AEC)
  2. Synchronization Likelihood (SL)

Graph-Theoretic Measures

Graph theory provides tools to analyze large-scale brain networks derived from connectivity matrices.

  • Nodes: Represent brain regions (e.g., EEG channels or anatomical regions).
  • Edges: Represent connectivity between nodes (e.g., ISPC, PLV, AEC values).

Key Graph Metrics

Metric Description Use Case
Degree Number of connections per node. Identifying hubs in the network.
Clustering Coefficient Measures how well nodes are connected to their neighbors. Studying local connectivity.
Path Length Average number of steps between any two nodes. Studying global efficiency.
Modularity Measures the strength of division into modules (communities). Identifying functional modules.
Small-Worldness Combines high clustering and short path length. Studying efficient brain networks.

Practical Considerations for EEG Functional Connectivity

Preprocessing

  • Referencing: Choose an appropriate reference (e.g., average, linked mastoids, or Laplacian).
  • Filtering: Apply band-pass filters to focus on specific frequency bands (e.g., alpha: 8–12 Hz, beta: 13–30 Hz).
  • Artifact Removal: Use ICA (Independent Component Analysis) or regression to remove artifacts (e.g., eye blinks, muscle activity).
  • Epoch Selection: Segment data into epochs (e.g., stimulus-locked or response-locked).

Volume Conduction

Problem: Nearby electrodes may show artificially high connectivity due to shared signals.

Solutions:

  • Use Laplacian referencing to reduce volume conduction.
  • Apply PLI, wPLI, or imaginary coherence to ignore zero-lag synchronization.
  • Use source localization (e.g., sLORETA, beamforming) to estimate connectivity between brain regions.

Statistical Testing

  • Permutation Testing: Compare observed connectivity values to a null distribution generated by shuffling data.
  • Multiple Comparisons: Correct for false positives using Bonferroni correction or FDR (False Discovery Rate).
  • Effect Size: Use Cohen’s d or Hedges’ g to quantify the strength of connectivity differences.

Applications

  1. Cognitive Neuroscience
  • Attention: Alpha-band connectivity between frontal and parietal regions during attention tasks.
  • Memory: Theta-band connectivity in the hippocampus and neocortex during memory encoding.
  • Language: Gamma-band connectivity between Broca’s and Wernicke’s areas during language processing.
  1. Clinical Neuroscience
  • Epilepsy: Disrupted connectivity in default mode network (DMN) and epileptic networks.
  • Schizophrenia: Reduced theta-band connectivity in the frontal-temporal network.
  • Alzheimer’s Disease: Decreased alpha-band connectivity in the parietal-occipital network.
  1. Brain-Computer Interfaces (BCIs)
  • Motor Imagery: Beta-band desynchronization in the motor cortex during imagined movement.
  • Error Processing: Theta-band connectivity between ACC (anterior cingulate cortex) and frontal regions during error detection.

Limitations and Challenges

Challenge Description Solution
Volume Conduction Nearby electrodes show artificial connectivity. Use PLI, wPLI, or source localization.
Non-Stationarity Connectivity patterns change over time. Use time-varying connectivity methods (e.g., sliding window).
Reference Choice Different references (e.g., average, mastoids) affect connectivity estimates. Use Laplacian referencing or source space analysis.
Noise and Artifacts Muscle activity, eye blinks, and line noise contaminate signals. Apply ICA or regression to remove artifacts.
Frequency Leakage Power in one band may leak into another. Use narrowband filtering or wavelet methods.
Directionality Phase-based measures do not indicate directionality. Use Granger causality or transfer entropy.

Summary Table

Method Type Frequency-Specific Directional Volume Conduction Resistant Use Case
ISPC Phase-based Yes No No Functional connectivity
ITPC Phase-based Yes No No Event-related synchronization
PLV Phase-based Yes No No Functional connectivity
PLI Phase-based Yes Yes Yes Directional connectivity
wPLI Phase-based Yes Yes Yes Robust connectivity
Imaginary Coherence Phase-based Yes Yes Yes Frequency-specific connectivity
AEC Amplitude-based Yes No No Co-activation
Synchronization Likelihood Non-linear No No No Non-linear dependencies
Graph Theory Network-based No Yes No Large-scale brain networks