Eyes-open resting-state EEG microstate analysis using ratio-normalized parameters across four- and seven-class solutions in healthy controls, mild cognitive impairment, and Alzheimer's disease
Chanda Simfukwe, Seong Soo A. An, Young Chul YounBackground
Resting-state EEG microstate analysis indexes large-scale network dynamics, yet the value of ratio-normalized temporal parameters under the eyes-open resting (EOR) condition remains unexplored across the cognitive decline continuum.
Objective
To examine whether ratio-normalized microstate parameters from four- and seven-class clustering solutions under EOR differentiate healthy controls (HC), mild cognitive impairment (MCI), and Alzheimer's disease (AD).
Methods
One hundred and eighty participants (HC, MCI, and AD; n = 60 each) were recruited from a hospital-based cohort in an elderly Korean population. Resting-state EEG recorded under EOR was analyzed using MICROSTATELAB in EEGLAB. Absolute temporal parameters (duration, occurrence, coverage) and ratio-normalized forms (relative duration, relative occurrence) were compared using ANCOVA with age as covariate, false discovery rate (FDR) correction, and Bonferroni-corrected post-hoc comparisons.
Results
In the four-class solution, three parameters reached nominal significance, none surviving FDR correction across the full set of 20 tests: coverage of Microstate B (visual network; F(2176) = 3.561, p = 0.030), relative occurrence of Microstate B (F(2176) = 5.281, p = 0.006), and relative occurrence of Microstate E (executive network; F(2176) = 3.816, p = 0.024). MCI showed higher visual and lower executive microstate engagement than AD (p ≤ 0.023); no contrast involving HC was significant. Absolute occurrence parameters were non-significant whereas their ratio-normalized counterparts were significant. The seven-class solution yielded no corrected effects.
Conclusions
Ratio-normalized EOR microstate parameters distinguished MCI from AD through visual and executive network engagement, with small-to-moderate effects and HC intermediate. The four-class solution outperformed the seven-class solution, contrasting with eyes-closed reports. As these analyses were exploratory and did not survive correction, confirmation in independent cohorts is required.