Reducing the Computational Complexity of EEG Source Localization With Cortical Patch Decomposition and Optimal...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Real-time electroencephalography (EEG) source localization is crucial for clinical diagnosis and brain-computer interfaces.
• Current methods face challenges in power efficiency, low complexity, and real-time implementation due to extensive computational iterations.
• The subspace-based MUltiple SIgnal Classification (MUSIC) algorithm, while effective, requires significant computational resources.

Purpose of the Study:

• To introduce novel techniques for reducing the computational burden of the MUSIC algorithm for EEG source localization.
• To enable power-efficient, low-complexity, and real-time EEG source localization.
• To improve the feasibility of long-run and mobile monitoring of cortical activity.

Main Methods:

• The cortex is divided into regions, and a dictionary learning-based nomination procedure identifies regions containing active sources, reducing the search space.
• A new electrode selection algorithm, utilizing the Cramer-Rao bound, optimizes the selection of electrode subsets for improved accuracy and efficiency.
• These methods aim to decrease the exhaustive search inherent in the standard MUSIC algorithm.

Main Results:

• The proposed techniques demonstrated a reduction in computational complexity by up to 90% in simulations.
• Performance was validated using simulated EEG signals across varying signal-to-noise ratios, electrode numbers, and nominated regions.
• The methods were successfully tested on real EEG data from an auditory oddball experiment.

Conclusions:

• The developed techniques achieve significant computational complexity reduction for EEG source localization.
• A strong agreement was found between the proposed method and the standard MUSIC algorithm in terms of topography and localization errors.
• These advancements facilitate real-time, long-term, and mobile cortical activity monitoring for clinical and research applications.