Hello!
I have two questions:
Thank you,
George
Indeed, one problem of the artifact cleaning with ICA is the selection of the components to remove.
The EEGLAB Infomax algorithm (used in Brainstorm) gives components that all capture a similar amount of information, therefore if you have a really powerful artifact in the signals, it might be split in multiple components.
In MEG, we can sometime observe different spatial topographies for the different components of a heartbeat. You may end up with one component for the QRS complex, and one for the T wave.
Your first screen capture could be something in this range.
If your subject is slightly moving, it may result in two slightly spatial components, with varying contributions over time. This could be what we see between IC1+IC3 and IC2 in your second screen captures.
With MEG recordings, we tend to recommend the SSP approach instead: it is not as refined in terms of signal separation, but it is much easier to handle (semi-) automatically. You identify where your artifact is in your data, and in most cases, you would get good results by removing only the first component, sometimes the second and/or third.
https://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsSsp
https://neuroimage.usc.edu/brainstorm/Tutorials/SSPCookbook
What if I do not see any components associated with heartbeats throughout whole record? Is it ok to not choose any heartbeat-component?
The projection of the heartbeats on the MEG sensors is heavily subject-dependent. Some subjects have anatomical configurations that lead to almost no artifacts.
