Community Detection in Multi-frequency EEG Networks

TL;DR

This paper introduces a multilayer network approach for community detection in multi-frequency EEG data, revealing task-related cross-frequency connectivity patterns that are not observable with single-frequency methods.

Contribution

It extends community detection algorithms to multilayer networks representing different EEG frequency bands, capturing full-spectrum neuronal connectivity.

Findings

Error responses show increased cross-frequency communities, especially between theta and gamma bands.

Community structures are more consistent across subjects during error responses.

Multi-frequency models reveal connectivity changes across tasks and response types.

Abstract

Objective: In recent years, the functional connectivity of the human brain has been studied with graph theoretical tools. One such approach is community detection which is fundamental for uncovering the localized networks. Existing methods focus on networks constructed from a single frequency band while ignoring multi-frequency nature of functional connectivity. Therefore, there is a need to study multi-frequency functional connectivity to be able to capture the full view of neuronal connectivity. Methods: In this paper, we use multilayer networks to model multi-frequency functional connectivity. In the proposed model, each layer corresponds to a different frequency band. We then extend the definition of modularity to multilayer networks to develop a new community detection algorithm. Results} The proposed approach is applied to electroencephalogram data collected during a study of error monitoring in the human brain. The differences between the community structures within and across different frequency bands for two response types, i.e. error and correct, are studied. Conclusion: The results indicate that following an error response, the brain organizes itself to form communities across frequencies, in particular between theta and gamma bands while a similar cross-frequency community formation is not observed for the correct response. Moreover, the community structures detected for the error response were more consistent across subjects compared to the community structures for correct response. Significance: The multi-frequency functional connectivity network models combined with multilayer community detection algorithms can reveal changes in cross-frequency functional connectivity network formation across different tasks and response types.