User's Guide‎ > ‎Analysis‎ > ‎

7 - Segmentation into micro-states

What is it for? 

The temporal segmentation analysis identifies periods of stable electric field topographies or 'functional microstates'. The segmentation generates  hypotheses, at the group-level, about when topographic modulations occur. These hypotheses will be then tested statistically using the fitting procedure (when, over a period of stable topography, different electric field topographies describe, e.g. condition A vs. B) .

The periods of stable topography are functionnally relevant periods of interest over which, for instance, GFP, ERP waveforms or IS can be averaged and then compared across conditions. Averaging data in time based on the period of interest revealed by the segmentation can help increasing the SNR.

Segmentation:

    - press Tools >>> Segmentation of EEG Files
    - define group of files to be compared
    - take overall grand_avr. eph C1 and C2 (…and C3…..)
    - optional but helpful check: click on "sort files within lists" 
    - choose clustering method (K-means or (T)AAHC)
    - add epoch (i.e. overall file length of grand_avr.eph or post-stimulus period)
    - choose no. of clusters (e.g. 3-20)
    - define output folder


>>> RUN

    - from the output folder use *.error.data to define appropriate no. of maps potentially explaining the observed differences between conditions

        o        Global Explained Variance (GEV) ↑
        o        Cross-Validation Criterion (CV) ↓   (very important!!!)
        o        KL criterion ↑ (even more important!!!)

    - click on respective peak/ valley
    - Go to “Options” and “Open segmentation at cursor position” which opens the convergent *.seg file (which comprises the landscape view of the map topographies or appearances, resp.)

        o        This *.seg-file also has got an equivalent *.ep which you need for the next processing step >>> i.e. the Fitting procedure

    - Choose time period(s) or epochs, respectively, with onset or offset or map presence differences (e.g. map 3 starts at the same time in condition X and Y but is present longer in condition Y) >>> pick all relevant TF (and map numbers present within these TF) since they provide the input for the successive fitting procedure





Comments