Chapter 4: Artifact Removal with ICA¶
Continuing from Chapter 3 — Preprocessing — Filtering and Referencing.
Filtering and re-referencing handle broad, predictable noise; independent component analysis (ICA) handles artifacts that overlap in frequency with real brain signal — chiefly eye blinks, eye movements, and muscle activity — the same problem tackled with code in the MNE-Python track's Artifact Removal with ICA chapter.
Running ICA¶
Process → Run ICA... decomposes the recording into statistically independent components, with a choice of algorithm — FastICA, Infomax, or PICARD are all supported, matching the same algorithm choices available to mne.preprocessing.ICA() in code. Depending on the algorithm and how much data you have, this can take anywhere from a few seconds to a few minutes.
Looking at components¶
Once ICA finishes, two plot commands help identify which components are artifacts rather than brain signal:
- Plot → Plot ICA Components shows a scalp topography per component; a blink component typically shows a sharp, symmetric pattern over the frontal electrodes.
- Plot → Plot ICA Sources shows each component's time course side by side, so you can line up large, sparse spikes with visible blinks or a rhythmic pattern with heartbeat.
Letting a classifier help: ICLabel¶
Judging every component by eye works but is slow and inconsistent. Process → Label ICs... runs ICLabel, an automatic classifier, over every component and labels each one — Brain, Eye, Muscle, Heart, Line Noise, Channel Noise, or Other — with a confidence score, turning "does this look like a blink" into a number you can threshold consistently across recordings.
Removing the artifact components¶
Once you've marked which components to exclude (by eye, by ICLabel, or both), Process → Apply ICA subtracts exactly those components' contribution and reconstructs a cleaned recording — the same "remove, don't just ignore" approach as ica.apply() in code. Process → Import/Export ICA... lets you save a fitted decomposition to reuse on another recording, or load one back in.