Eeg Microstates and Balance Parameters for Stroke Discrimination: A Machine Learning Approach

Area of Science:

• Neuroscience
• Biomarkers
• Machine Learning

Background:

• Electroencephalography microstates (EEG-MS) are increasingly recognized for their potential as neurobiological biomarkers.
• Stroke diagnosis and management can benefit from objective neurophysiological markers.

Purpose of the Study:

• To identify reliable biomarkers for discriminating stroke patients from healthy controls.
• To utilize machine learning to analyze EEG-MS and clinical data for stroke classification.

Main Methods:

• Recruited 54 participants (27 stroke patients, 27 controls).
• Recorded 32-channel EEG-MS under eyes-closed and eyes-open conditions, analyzing classical microstate maps (A, B, C, D).
• Employed k-means clustering with clinical and EEG-MS features, evaluating clustering quality using the Silhouette score.

Main Results:

• Machine learning identified balance scores, duration/coverage of microstate A, and duration/coverage/occurrence of microstate C as key discriminators.
• Global variance explained was also a significant parameter.
• The Silhouette score indicated overlapping clusters (approx. 0.20), suggesting further refinement is needed.

Conclusions:

• The study demonstrates the potential of EEG-MS combined with clinical data for stroke patient classification.
• These findings support the development of novel therapeutic strategies and improved clinical management for stroke.
• This approach may lead to reduced healthcare costs associated with stroke.