Combining electro- and magnetoencephalography data using directional archetypal analysis.

Area of Science:

• Neuroscience
• Data Analysis
• Signal Processing

Background:

• Metastable microstates in EEG/MEG are typically found using k-means, which assumes discrete brain state transitions.
• Existing methods often fail to capture the continuous nature of brain microstate dynamics.

Purpose of the Study:

• To develop a novel method for analyzing electroencephalography (EEG) and magnetoencephalography (MEG) microstates.
• To model brain microstates as continuous trajectories rather than discrete states.
• To account for intersubject and intermodal variability in neural data.

Main Methods:

• Introduced multimodal, multisubject directional archetypal analysis (DAA).
• Utilized a loss function based on the Watson distribution for scale and polarity invariance.
• Modeled EEG/MEG microstates as subject- and modality-specific archetypes representing continuous topographic maps.

Main Results:

• Successfully modeled scale and polarity invariant data, including EEG/MEG microstates.
• Demonstrated the ability to account for intersubject and intermodal variability.
• Showcased model performance on synthetic and real-world face perception data.

Conclusions:

• Directional archetypal analysis offers an advantageous approach to modeling brain microstate trajectories over discrete assignments.
• The method effectively captures continuous state transitions and variability in neural signals.
• The model is adaptable for other modalities, ensuring component correspondence and elucidating spatiotemporal signal variability.