Microstate Detection in Naturalistic Electroencephalography Data: A Systematic Comparison of Topographical Clustering Strategies on an Emotional Database

Abstract

Electroencephalography (EEG) microstate analysis is a powerful tool to study the spatial and temporal dynamics of human brain activity, through analyzing the quasi-stable states in EEG signals. However, current studies mainly focus on rest-state EEG recordings, microstate analysis for the recording of EEG signals during naturalistic tasks is limited. It remains an open question whether current topographical clustering strategies for rest-state microstate analysis could be directly applied to task-state EEG data under the natural and dynamic conditions and whether stable and reliable results could still be achieved. It is necessary to answer the question and explore whether the topographical clustering strategies would affect the performance of microstate detection in task-state EEG microstate analysis. If it exists differences in microstate detection performance when different topographical clustering strategies are adopted, then we want to know how the alternations of the topographical clustering strategies are associated with the naturalistic task. To answer these questions, we work on a public emotion database using naturalistic and dynamic music videos as the stimulation to evaluate the effects of different topographical clustering strategies for task-state EEG microstate analysis. The performance results are systematically examined and compared in terms of microstate quality, task efficacy, and computational efficiency, and the impact of topographical clustering strategies on microstate analysis for naturalistic task data is discussed. The results reveal that a single-trial-based bottom-up topographical clustering strategy (bottom-up) achieves comparable results with the task-driven-based top-down topographical clustering (top-down). It suggests that, when task information is unknown, the single-trial-based topographical clustering could be a good choice for microstate analysis and neural activity study on naturalistic EEG data.

Emotion recognition based on microstate analysis from temporal and spatial patterns of electroencephalogram

“…Furthermore, to be consistent with previous literatures ( Shen et al, 2020 ; Hu et al, 2022a , b ) for comparison, we repeated the microstate analysis on DEAP and SEED datasets and designated the number of microstates as 4 a priori . We also repeated the same statistical analysis on the DEAP dataset when the number of microstates was 4, as described in Section 2.3 (Statistical Analysis of Microstate Features).…”

“…In order to further analyze the changing rules of microstates and parameters under different emotional states, and have a better understanding of the neurophysiological significance of microstates during the cognitive process of emotion, we summarize the current studies that also employ the microstate analysis method to emotional EEG signals ( Shen et al, 2020 ; Hu et al, 2022a , b ). These three studies conducted the microstate and statistical analysis on the DEAP dataset with one accord.…”

Bayesian Optimization of Machine Learning Classification of Resting-State EEG Microstates in Schizophrenia: A Proof-of-Concept Preliminary Study Based on Secondary Analysis

“…We identified four microstates (i.e., A, B, C, and D, Figure 2 ), which reflected the activity of the EEG channels with almost 70% of the total topographic variance, and labeled the topography at each GFP peak as one of these microstates. Five categories of features were computed for these microstates ( Figure 3 and Figure S1 ): (1) the average number of times per second that each microstate occurred during the EEG recording (occurrence); (2) the average amount of time each microstate lasted after it occurred (duration); (3) the percentage of total recording time in which each microstate was dominant (coverage); [ 19 ] (4) the average GFP during microstate dominance (mean GFP) [ 28 ]; and (5) the average correlation between each labeled GFP peak map (i.e., A) and the corresponding microstate template (microstate map correlations) [ 34 ]. All microstate features are summarized in Table 1 .…”

“…In contrast, more complex statistical models, such as those used in machine learning approaches, can be utilized to evaluate the combination of several predictors for the classification of SZ concurrently [ 31 ]. Although previous studies investigated SZ classification with machine-learning-based clustering algorithms [ 32 , 33 , 34 ], these studies have mostly used univariate or conventional machine learning methods, rather than optimized multivariate analyses [ 18 ]. Choosing optimized SZ classification is challenging and comprises multiple hyperparameters, which are set before the training process and define how the model can best fit the data [ 31 , 35 ].…”

Feature extraction based on microstate sequences for EEG–based emotion recognition

“…Shen et al (2020) explored EEG microstates for emotional experiences during music video watching. Hu et al (2022) systematically examined and compared the microstates for task-state EEG analysis during naturalistic music videos. In the existing studies, most studies acknowledge four standard microstate maps that can explain up to 65-85% of the EEG signal's global variance.…”