ica_processing.classify_ica_components#

autoclean.functions.ica.ica_processing.classify_ica_components(raw, ica, method='iclabel', verbose=None)[source]#

Classify ICA components using automated algorithms.

This function uses automated classification methods to identify the likely source of each ICA component (brain, eye, muscle, heart, etc.). Supports both ICLabel and ICVision methods for component classification.

Parameters:

rawmne.io.Raw: The raw EEG data used for ICA fitting.
icamne.preprocessing.ICA: The fitted ICA object to classify.
methodstr, default “iclabel”: Classification method to use. Options: “iclabel”, “icvision”.
verbosebool or None, default None: Control verbosity of output.

Returns:

component_labelspd.DataFrame: DataFrame with columns: - “component”: Component index - “ic_type”: Predicted component type (brain, eye, muscle, etc.) - “confidence”: Confidence score (0-1) for the prediction - Additional columns with probabilities for each component type

See also

fit_ica: Fit ICA decomposition to EEG data
apply_ica_rejection: Apply ICA to remove artifact components
mne_icalabel.label_components: ICLabel implementation

Examples

>>> labels = classify_ica_components(raw, ica, method="iclabel")
>>> labels = classify_ica_components(raw, ica, method="icvision")
>>> artifacts = labels[(labels["ic_type"] == "eye") & (labels["confidence"] > 0.8)]