Hi Denis,
Thanks for your reply. I am concerned about the empty room noise removal
during the *pre-processing stage (artifacts removal using SSP)* as
recommended in the MNE website ( https://mne.tools/stable/auto_tutorials/preprocessing/plot_50_artifact_correction_ssp.html).
*Another related question: *How to choose the number of noise components or
projectors (empty room,ecg,eog) during artifact removal when using SSP? Is
there any automated way or setting any threshold to automatically remove
depending upon the subjects?
my suggestion is to look at the data before and after SSP, including the evoked
response. Pre-processing is all about increasing SNR so you need to remove noise
making sure that you don't loose too much signal by doing this.
with mne you can use proj="interactive" option is evoked.plot to see how SSPs
affect the data. Make sure the projections are not already applied though.
I would add: one single cardiac / occular SSP vector is often good enough and safe.
For empty room SSS it really depends on your data. I think on some sites people include up to 16 vectors for the magnetometers.
My impression is that empty room SSP is much less critical and being generous with it is not a bad idea.
But always do plots to analyze the impact of your preprocessing.
Thank you, Denis & Alex for your quick replies!
What we were trying to clarify is whether adding ssp-projections computed
from emptyrooms
to Neuromag data to which Maxfilter and SSS have been applied should have
additional benefits, theoretically and practically.
Related, a quick side question: the system-based projectors that appear in
the raw.fif after recording, are they just discarded during Maxfilter, or
have they been applied?
Hi there, I am not sure to follow entirely what the question is (but it may
be my jetlag)?
1) If you wish to know the environmental/site noise level, this is given
by the set of SSP that we read when maxfiltering (cf config in the pipeline)
2) for your purpose, you may be interested in ER (sensor) noise which
Merlin is currently investigating so you may want to chat with him
3) in addition to the site noise, you apply SSP or ICA for biological
artifacts (as previously mentioned 2 to 3 components are typically far
enough per biological artifact, e.g. 2 for EOG and 2 for ECG)
It uses the extended SSS (eSSS) method to improve the external SSS basis
during Maxwell filtering. This method was originally presented back in 2016
-- if you're curious to learn more about it, there is a poster link in the
PR above.