Rejuvenating classical brain electrophysiology source localization methods with spatial graph Fourier filters for...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography/magnetoencephalography source imaging (ESI) seeks to identify brain activity origins from scalp recordings.
• Classical ESI methods face limitations in accurately estimating the extent of brain activation, often producing overly diffuse or sparse results.
• Accurate estimation of both source location and spatial extent is crucial for clinical decision-making.

Purpose of the Study:

• To develop an improved ESI method that accurately estimates extended brain activation regions.
• To leverage graph structures and the spatial graph Fourier transform (GFT) to enhance source extent estimation.
• To integrate GFT into classical ESI algorithms for improved performance.

Main Methods:

• Utilized 3D brain mesh graph structures and spatial graph Fourier transform (GFT).
• Decomposed spatial graph structures into low-, medium-, and high-frequency basis.
• Employed low-frequency GFT basis to approximate extended brain activation areas and embedded GFT into classical ESI methods.
• Validated the proposed GFT-enhanced ESI methods using synthetic and real EEG/MEG data.

Main Results:

• The proposed GFT-based method effectively reconstructed focal source patterns.
• Significant performance improvements were observed compared to classical ESI algorithms in both synthetic and real data.
• The method demonstrated enhanced accuracy in estimating the extent of brain activation regions.

Conclusions:

• The integration of spatial graph Fourier transform offers a promising approach to improve EEG/MEG source imaging.
• The GFT-based method enhances the accurate localization and extent estimation of brain sources.
• This advancement holds potential for more precise clinical decision-making in neurological assessments.