Run ICA

ICA is run twice in the pipeline. The first time IC decomposition is simply used to detect time periods in which the IC time courses are noisy. The second ICA is the final decomposition, which is passed to the ICLabel classifier.

There are two YAML fields for tuning ICA parameters. the ica field will contain parameters for detecting noisy time periods in the data with outlying IC time courses, similar to the approach in the Flag Noisy Epochs step. The ica_args field is where you should pass keyword arguments that are accepted by MNE-Pythons mne.preprocessing.ICA class.

For a detailed description of the ica field parameters, please see the Flag Noisy Epochs step, as the approach and definitions are identical.

ica:
    flag_crit: 0.2
    outlier_method: quantile
    outliers_kwargs:
        k: 6
        lower: 0.25
        upper: 0.75

# See arguments definition from mne.preprocessing.ICA
ica_args:
    run1:
        method: fastica
    run2:
        method: infomax
    fit_params:
        extended: True