Towards an objective evaluation of EEG/MEG source estimation methods - The linear approach.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Spatial resolution in EEG/MEG source estimation is limited by source leakage.
• Existing software offers many methods but lacks tools for objective evaluation and comparison.
• Experimental researchers need intuitive tools to assess the performance of source estimation techniques.

Purpose of the Study:

• To present a framework and tools for objective resolution analysis of EEG/MEG source estimation methods.
• To evaluate widely used methods like L2-MNE and LCMV beamformers using defined resolution metrics.
• To provide a tutorial for experimental researchers on linear EEG/MEG source estimation and resolution analysis.

Main Methods:

• Developed a framework based on linear systems analysis for resolution analysis.
• Defined key resolution metrics: Peak Localization Error (PLE) and Spatial Deviation (SD).
• Utilized the resolution matrix, including point-spread (PSF) and cross-talk (CTF) functions, to characterize source leakage.

Main Results:

• Some methods achieved low or zero PLE for PSFs, indicating good localization accuracy in ideal cases.
• All evaluated methods showed suboptimal Spatial Deviation (SD) and performance on cross-talk functions (CTFs).
• Performance was particularly limited for deep cortical areas across all methods.

Conclusions:

• The developed framework and tools enable objective comparison of EEG/MEG source estimation methods.
• Current linear methods have inherent limitations in spatial resolution and leakage, especially for deeper sources.
• Encourages wider adoption of resolution analysis to improve the reliability of EEG/MEG source estimation in research.