Data-driven visualization and group analysis of multichannel EEG coherence with functional units

Area of Science:

• Neuroscience
• Data Visualization
• Signal Processing

Background:

• Multichannel electroencephalography (EEG) coherence visualization often suffers from visual clutter using traditional graph layouts.
• Functional units (FUs) are defined as data-driven regions of interest (ROIs) to mitigate this clutter.
• Previous methods for detecting FUs include maximal clique based (MCB) and watershed based (WB).

Purpose of the Study:

• To introduce an improved watershed based (IWB) method for detecting FUs in EEG coherence, aiming to reduce over-segmentation issues.
• To develop novel group maps for data-driven group analysis of EEG coherence.
• To evaluate the performance of the IWB method and new group maps using a case study.

Main Methods:

• Developed the improved watershed based (IWB) method, which merges basins if they are spatially connected and their union forms a clique.
• Implemented MCB, WB, and IWB methods for detecting FUs and compared their performance.
• Introduced two group maps: group mean coherence map and group FU size map for data-driven group analysis.

Main Results:

• The IWB method demonstrated a smaller difference compared to the MCB method (gold standard) than the WB method.
• Both WB and IWB methods are significantly faster (up to 100,000x) than the MCB method, enabling interactive visualization.
• Case study results indicated differences in EEG coherence between younger and older adults, validating the IWB method and group maps.

Conclusions:

• The IWB method provides a more accurate and computationally efficient approach for detecting FUs in multichannel EEG coherence.
• The novel group maps facilitate data-driven group analysis, potentially extending hypothesis-driven ROI selection.
• This approach enhances the visualization and analysis of complex EEG coherence patterns for neuroscience research.