mne_icalabel.megnet.megnet_label_components

mne_icalabel.megnet.megnet_label_components(raw, ica)

Label the provided ICA components with the MEGnet neural network.

For more information, see Treacher et al.[1].

Parameters:

rawRaw

Raw MEG recording used to fit the ICA decomposition. The raw instance should be bandpass filtered between 1 and 100 Hz and notch filtered at 50 or 60 Hz to remove line noise, and downsampled to 250 Hz.

icaICA

ICA decomposition of the provided instance. The ICA decomposition should use the infomax method.

Returns:

labels_pred_probanumpy.ndarray of shape (n_components, n_classes)

The estimated corresponding predicted probabilities of output classes for each independent component. Columns are ordered with 'brain/other', 'eye movement', 'heart beat', 'eye blink'.

References

[1]

Alex H. Treacher, Prabhat Garg, Elizabeth Davenport, Ryan Godwin, Amy Proskovec, Leonardo Guimaraes Bezerra, Gowtham Murugesan, Ben Wagner, Christopher T. Whitlow, Joel D. Stitzel, Joseph A. Maldjian, and Albert A. Montillo. Megnet: automatic ica-based artifact removal for meg using spatiotemporal convolutional neural networks. NeuroImage, 241:118402, 2021. doi:https://doi.org/10.1016/j.neuroimage.2021.118402.