Characteristics of EEG microstates in stroke patients with cognitive impairment after basal ganglia injury.

Abstract

Objectives: To explore changes in Electroencephalography (EEG) microstates in patients with cognitive impairment following basal ganglia stroke to understand the neural mechanisms of cognitive deficits better.

Methods: Thirty post-stroke cognitive impairment (PSCI, MoCA < 26, age: 60.07 ± 7.57, male/female: 22/8) patients, 23 post-stroke patients without cognitive impairment (PSN, MoCA ≥ 26, age: 59.57 ± 8.65, male/female: 17/6), and 27 healthy controls (HC, MoCA ≥ 26, age: 62.26 ± 6.65, male/female: 17/10) underwent cognitive tests and EEG recordings. EEG data were preprocessed to analyze microstate parameters, with variance testing performed across groups. Following preprocessing of the raw EEG data, global field power (GFP) was computed to identify periods of maximal topographic stability. Four prototypical microstate classes were derived using K-means clustering, after which three key temporal characteristics were quantified for each participant: (1) microstate mean duration, (2) Mean Frequency of Occurrence, and (3) Mean Coverage. Correlation analyses were conducted between microstate parameters and cognitive scores in the PSCI group. The cut-off value, sensitivity, and specificity of metrics related to overall cognitive function were calculated with the receiver operating characteristic curve.

Results: Cognitive assessments revealed significantly poorer performance in all domains for the PSCI group than the PSN and HC groups (p < 0.001). The PSCI group exhibited a longer mean media duration (MMD) and lower incidence mean frequency of occurrence (MFO) of EEG microstates compared to other groups (p < 0.01). The mean duration of microstates A, and D negatively correlated with MoCA total scores (microstates A: r = -0.491, microstates D: r = -0.372), particularly in attention and orientation domains. Furthermore, receiver operating characteristic (ROC) curve analysis indicated that the mean duration of microstate A can potentially serve as a diagnostic biomarker for PSCI. The optimal cut-off values for A-MMD were 45.41 points. The area under the curve was 0.82, sensitivity was 80 %, and specificity was 69.6 %.

Conclusion: Basal ganglia injury is associated with abnormal EEG microstate dynamics, characterized by prolonged microstate duration and reduced incidence rate, contributing to cognitive network dysfunction. These findings suggest EEG microstates as potential biomarkers for diagnosis.