Connectivity-based Meta-Bands: A new approach for automatic frequency band identification in connectivity analyses

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Current electroencephalographic (EEG) and magnetoencephalographic (MEG) analyses often use fixed "canonical" frequency bands.
• This approach lacks individual adaptation, potentially limiting the accuracy of neural signal analysis.
• Existing methods may overlook subject-specific neural idiosyncrasies.

Purpose of the Study:

• To develop and validate a novel data-driven method for automatic identification of personalized neural signal frequency ranges.
• To overcome the limitations of traditional fixed frequency band segmentation in EEG and MEG studies.
• To enhance the robustness and personalization of brain connectivity analyses.

Main Methods:

• Introduced the Connectivity-based Meta-Bands (CMB) algorithm, a data-driven approach for unsupervised band segmentation.
• Analyzed resting-state MEG and EEG data from 195 healthy subjects using narrow-band filtering (1-70 Hz).
• Estimated frequency-dependent functional neural networks via orthogonalized amplitude envelope correlation and community detection.

Main Results:

• The CMB algorithm successfully identified subject-specific frequency ranges based on functional neural network topology.
• Traditional canonical bands showed partial alignment with group-level MEG network topology but missed individual patterns.
• EEG signals exhibited limited sensitivity to detailed frequency-dependent network structures, showing simpler parcellations.

Conclusions:

• The CMB algorithm offers a robust, personalized approach to analyzing brain connectivity by accounting for individual neural idiosyncrasies.
• This methodology moves beyond fixed frequency bands, enabling more accurate and tailored neurophysiological studies.
• The findings pave the way for exploring the nuanced frequency-dependent structure of brain connectivity in diverse populations.