Anatomical harmonics basis based brain source localization with application to epilepsy.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain Source Localization (BSL) via Electroencephalogram (EEG) is crucial for diagnosing epilepsy and brain disorders.
• Current BSL methods face limitations in computational cost and accuracy due to head shape assumptions and basis functions.
• Existing techniques struggle with high dimensionality and computational complexity.

Purpose of the Study:

• To present an anatomical harmonics basis (Spherical Harmonics (SH) and Head Harmonics (H2)) for improved BSL.
• To reduce computational cost and enhance localization accuracy in EEG-based BSL.
• To validate the proposed H2 MUSIC method for clinical epileptogenic zone localization.

Main Methods:

• Formulated a spatio-temporal four-shell head model in the SH and H2 domains.
• Developed anatomical harmonics-based BSL using SH and H2.
• Compared spatial subspace methods (MUSIC, RAP-MUSIC) with SH and H2 counterparts on simulated data.
• Validated H2 MUSIC on clinical EEG data for epileptogenic zone localization.

Main Results:

• Anatomical harmonics domain formulation reduced dimensionality and increased source eigenvalue contribution.
• Achieved decreased computation and increased accuracy in BSL.
• SH and H2 domain processing resolved high computational cost without sacrificing localization accuracy.
• H2 MUSIC demonstrated effectiveness in clinical epileptogenic zone localization.

Conclusions:

• The proposed SH and H2 domain formulation offers an effective BSL framework.
• This method significantly decreases computational load while maintaining or improving localization accuracy.
• The H2 MUSIC approach provides a clinically viable solution for automated, time-efficient seizure localization.