mne_icalabel.iclabel.run_iclabel

mne_icalabel.iclabel.run_iclabel(images, psds, autocorr, backend='pytorch')

Run the ICLabel network on the provided set of features.

The features are un-formatted and are as-returned by get_iclabel_features. For more information, see Pion-Tonachini et al.[1].

Parameters:

imagesarray of shape (n_components, 1, 32, 32)

The topoplot images.

psdsarray of shape (n_components, 1, 1, 100)

The power spectral density features.

autocorrarray of shape (n_components, 1, 1, 100)

The autocorrelation features.

backendNone | torch | onnx

Backend to use to run ICLabel. If None, returns the first available backend in the order torch, onnx.

Returns:

labelsarray of shape (n_components, n_classes)

The predicted numerical probability values for all labels in ICLabel output. Columns are ordered with 'Brain', 'Muscle', 'Eye', 'Heart', 'Line Noise', 'Channel Noise', and 'Other'.

References

[1]

Luca Pion-Tonachini, Ken Kreutz-Delgado, and Scott Makeig. Iclabel: an automated electroencephalographic independent component classifier, dataset, and website. NeuroImage, 198:181–197, September 2019. doi:10.1016/j.neuroimage.2019.05.026.