Classifying Migraine Phases using Fuzzy Entropy and Microstate Features in Resting-state EEG

This study proposed a migraine phase detection system based on resting-state EEG, utilizing features derived from both fuzzy entropy and microstate parameters. EEG data were recorded using an 8-channel wearable EEG device developed by NYCU Neural Engineering Laboratory, and the extracted features were used to classify different migraine phases and investigate the dynamic EEG changes throughout these phases. The experiment results revealed consistent patterns in fuzzy entropy and microstate parameters corresponding to distinct migraine phases. Notably, EEG entropy significantly decreased during the preictal phase (p <.05) in the C3 and C4 channels under eyes-open conditions. Additionally, the mean Global Field Power (GFP) of microstate A during the interictal phase was significantly higher than that during the preictal phase (p <.01). Using a random forest classifier, the proposed classification system could achieve an accuracy of 88% in identifying different migraine phases. These findings demonstrate that integrating fuzzy entropy and microstate parameters offers an innovative approach to migraine phase detection. This system holds promise for providing early warnings, enabling timely interventions, supporting personalized treatment, reducing healthcare costs, and ultimately improving the quality of life for individuals with migraines.