Microstate analysis

EEG microstates were analysed as described by Koenig et al.^12,24,28, one of the most widely used methods. With this method, we obtained four archetype microstates with functional significance found in fMRI studies using the EEGlab toolbox. This measures global electrical activity in instant of time. First, GFP was derived from EEG recordings as follows:

where Vit and Vmean(t) denote the instantaneous and mean potentials across the N electrodes at time t.

Then, successive microstates, which are discrete states of the EEG analysed based on local maxima of the GFP, were derived. Using modified k-means clustering, all microstates were assigned to four archetype microstates as follows: left–right orientation (type A), right-left orientation (type B), anterior–posterior orientation (type C), and a fronto-central maximum (type D)^12,24.

Once the number of microstates was identified (four archetype microstates), we have to label them into a sequence by using modified K-means clustering algorithm and Global Explain Variance (GEV) criteria⁷⁵. The setting parameters for K-means algorithms are re-run and iterations as explained following. In principle, by re-running the stochastic k-means algorithm multiple times (in this analysis, we set re-run parameter to 20 times), we are able to test multiple segmentations on the same dataset and select the best re-run based on the GEV criteria⁷⁵. GEV is a measure of how similar each EEG sample is to the microstate prototype it has been assigned to. The higher the GEV the better⁶⁴. More importantly, we are able to reach the global minimum among 20 local minimums (20 re-runs). After 20 re-runs, the one that maximises the GEV is selected. However, the number of re-run is a trade-off between computation time and how likely we are to converge on the same optimal solution. In the Microstate EEGlab toolbox⁶⁴ and tis Python package⁷⁶ in which we have applied for our analysis select 10 re-run as a default value. In addition, Thomas Koenig's manual⁷⁷ has recommended that a range from 20 to 50 re-runs could be sufficient for a proper analysis. Furthermore, we have found that there are several existing EEG microstate analysis literatures that set 10 re-runs^65,78 as well as papers that use 30²⁰ as a proper re-run number.

Another parameter for K-means clustering is iteration, which means that in each re-run the K-means algorithm keeps iterating until some stopping criteria (convergence threshold) are satisfied. In this analysis, we used the convergence threshold, which stops the algorithms when the relative error change between subsequence iterations is below the threshold. Here, we set the threshold at 10-6. The maximum number of iterations set to 1000 which means the algorithm can stop if the maximum iteration is reached before convergence for computation time efficiency.

Microstate features were then extracted. In each microstate, duration was defined as the average duration of microstates per second. Occurrence was defined as the average frequency of microstates observed. Coverage was defined as the percentage of each microstate appearing in each epoch. Mean GFP was defined as the average GFP for a microstate. Next, two features were extracted for epochs across all types: mean duration was defined as the average duration of all types in a specific epoch and mean occurrence was defined as the frequency of all microstates per second in each epoch. In summary, a total of 18 microstate features were used in this study: four features for each type and two across all types. Transition probabilities, the percentage of transition from one to the different microstates, were also calculated. Thus, there were a total of 12 pairs. The directional predominance introduced by Lehmann et al. were also analyzed. It reveals the asymmetries of transition between two types of microstates in their sequences. Using the permutation test, the difference between the expected transition and the observed transition was calculated. This process was run 10,000 times to obtain the p-value.