Forward-Projected Cortical Eigenmodes Provide an Efficient Sensor-Space Representation of Resting-State EEG.

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Computational Neuroscience

Background:

• Standard electroencephalography (EEG) analyses often neglect cortical geometry, hindering the link between scalp patterns and brain anatomy.
• Comparing EEG patterns across individuals is challenging due to the lack of anatomical encoding in traditional sensor-space methods.

Purpose of the Study:

• To introduce a novel sensor-space basis dictionary that explicitly integrates cortical geometry for EEG analysis.
• To improve the anatomical interpretability and cross-participant comparability of EEG data.

Main Methods:

• Computed Laplace-Beltrami (LB) eigenmodes on a standard cortical template (fsaverage).
• Mapped LB eigenmodes to sensor space using a lead-field matrix from a boundary-element (BEM) head model, creating cortex-anchored sensor-space harmonics.
• Assessed representational efficiency and between-condition consistency of the new basis using resting-state EEG data and comparing it with spherical harmonics (SPH), principal components (PCA), and independent components (ICA).

Main Results:

• The cortex-anchored LB basis demonstrated superior representational efficiency, explaining more variance with fewer modes compared to SPH, PCA, and ICA.
• The method achieved higher early-K variance explained (e.g., LB 0.54–0.59 vs. SPH 0.42–0.46 vs. PCA 0.07–0.09 for 59-channel eyes-closed data).
• Mode-wise coefficient consistency between eyes-open and eyes-closed conditions was comparable between the LB basis and SPH.

Conclusions:

• The proposed cortex-anchored basis provides a geometry-aligned and interpretable representation of sensor-space EEG.
• This approach offers superior fidelity-complexity trade-offs for low-dimensional EEG analysis.
• It establishes a principled scaffold for analyzing EEG data with enhanced anatomical relevance.