I was going through the tutorials and will start using Brainstorm on my data. On accessing my data, I realized that the NeuroMag based data that we use in our lab has SSS (Signal-Space Separation) applied to it. Does it make sense to apply SSP to this data or just skip the steps in this part of the tutorial?
We recommend that you follow the introduction tutorials using the example dataset provided, following all the steps, so that you get all the necessary training for being autonomous with the software.
You might be interested in following also additional advanced Elekta-based tutorials in the section "Other analysis scenarios":
https://neuroimage.usc.edu/brainstorm/Tutorials
Later, when processing your own recordings, you can mix and match all the pre-processing techniques depending on the type of noise present in your signals. Do not do any filtering or artifact cleaning presented in the recordings if you don't need to. The least amount of preprocessing you do on your data, the better.
There is no incompatibility between using MaxFilter SSS (which removes sources of signals that are located outside of the head) and SSP for eye blinks and residual heartbeats in Brainstorm. Evaluate the noise before applying anything.
1 Like
Awesome, thanks for the input.
Hello Francois,
I am wondering, did you perhaps mix SSP and ICA...? In terminology that I am used to, SSP is typically aiming at rejecting external interference, done based on empty-room noise covariance.
@MattiS Indeed, you can compute SSP from noise recordings and remove some components corresponding to ambient noise. I think this is sometimes done for reviewing raw MEG signals in MNE or Elekta viewers.
I was referring to the computation of SSP from examples of reproducible artifacts (heartbeats, blinks), as I thought it is what @Darah's question:
https://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsSsp#SSP_Theory
ICA includes additional constrains of independence in time of the different components, and can also be an interesting option for cleaning MEG recordings:
https://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy#Artifact_cleaning_with_ICA
The way how @Darah asked about SSP and SSS together, I interpreted that he is talking about SSP as an older or alternative way for removing external interference, like SSP was originally spexed & like how it is applied in Neuromag / Elekta / MNE pipeline... it is indeed typically done for during-measurement viewing, but some labs still use that as their main artefact rejection (in some rooms it apparently works very well). In that context, if data are SSS-processed, one would not apply SSP after that.
But I agree with you that when one works thru a tutorial, it would be good to do all steps using the tutorial dataset.
I wonder, how well SSP would work in removing physiological artefact signals that have time-varying topography (like MCG, blinks). Do you have experience on that?
Yes, we do have a lot of experience with this. Heartbeats and eye blinks have time-varying topographies, but that can typically be approximated with a succession of 1 to 3 stable topographies. By removing 1, 2 or 3 SSP components, you can expect to remove a lot of the artifact, and fully automatically in the good cases (not having the problem of picking a somewhat random number of ICA components).
Of course, it won't work all the time, it depends on the MEG system, the shielding, the subject anatomy and behavior...
You can check out all the tutorials related with MEG on the Brainstorm website, you'll find well illustrated examples:
https://neuroimage.usc.edu/brainstorm/Tutorials
Apologies for not being more clearer, I did mean, as @Francois stated, to ask if SSP (for reproducible artifacts) can be applied after SSS. I was erroneously assuming that SSS also gets rid of artifacts.