An enhanced microstate clustering algorithm based on canopy, K-means, and genetic simulated annealing.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Electroencephalogram (EEG) microstate analysis reveals transient brain activity patterns crucial for motor and cognitive functions.
• Limitations in traditional microstate algorithms hinder a comprehensive understanding of neural mechanisms in complex conditions.

Purpose of the Study:

• To introduce a novel Canopy-KM-GSA algorithm for automated microstate determination and sequence refinement.
• To enhance the accuracy and depth of EEG microstate analysis.

Main Methods:

• Developed the Canopy-KM-GSA algorithm, integrating Canopy clustering, K-means, and a genetic simulated annealing framework.
• Applied the algorithm to analyze EEG data from motor tasks, auditory oddball paradigms, and epileptic patients.
• Benchmarked Canopy-KM-GSA against seven traditional and advanced microstate analysis algorithms.

Main Results:

• Canopy-KM-GSA demonstrated superior performance across all tested datasets, significantly outperforming baseline algorithms.
• Achieved high Global Explained Variance (GEV), Calinski-Harabasz Index (CHI), and favorable Davies-Bouldin Index (DBI) across pedaling, auditory, and epileptic patient datasets.
• Specific metrics include average GEV of 94.43% (pedaling), 94.46% (auditory), and 58.40% (epileptic patients).

Conclusions:

• The proposed Canopy-KM-GSA algorithm provides a more accurate and robust tool for EEG microstate analysis.
• This advancement facilitates deeper insights into brain function and dysfunction.
• The algorithm's effectiveness is validated across diverse neurological and cognitive tasks.