Minimum Overlap Component Analysis (MOCA) of EEG/MEG data for more than two sources

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Analysis of electroencephalography (EEG) and magnetoencephalography (MEG) data often yields subspaces of sensor space potentials from multiple sources.
• Existing methods may struggle with decomposing these complex subspaces into distinct source contributions.

Purpose of the Study:

• To propose a general, model-free method for decomposing subspaces of EEG/MEG data into contributions from distinct sources.
• To generalize the Minimum Overlap Component Analysis (MOCA) for analyzing more than two sources.

Main Methods:

• A linear inverse method is used to find the subspace in source space.
• A linear transformation is applied to achieve mutually orthogonal source distributions with minimal overlap.
• The algorithm is a generalization of MOCA, designed for computational efficiency and robustness against local minima.

Main Results:

• The proposed method successfully decomposes subspaces into distinct source contributions.
• The algorithm demonstrates negligible computational cost and avoids local minima.
• Illustrative results for alpha rhythm analysis are presented, showcasing the method's effectiveness.

Conclusions:

• The developed method offers a robust and efficient approach for source separation in EEG/MEG data.
• This technique provides a unique decomposition of source contributions within sensor space subspaces.
• The generalized MOCA algorithm is a valuable tool for analyzing complex brain activity patterns.