Hi Team,
I was going through tutorials and few questions on forum which were asked few years ago. I would like to confirm if my understanding is right. Can you please help me with it ?
Bad channel interpolation: A few channels are marked as bad and I will be focusing on analysing reconstructed source time series. Thus, I don't have consider bad channel interpolation and I believe, there is no such feature in brainstorm either ?
I am analysing resting state MEG data and have marked some time segments as bad. I haven't clearly understood - if I need to interpolate/perform some operation for effective concatenation of the trimmed boundaries after I remove the bad time segment. Can you help me here ?
Also, for computation of data covariance matrix, what would you advice on selecting the baseline time segment ?
Appreciate all your patience and support!
Thanks!
As always, there is no generic answer to analysis questions without a better understanding on our end of your research questions.
But broadly speaking, it always recommended to remove the DC offset from raw MEG recordings, as these DC values are much larger than brain signals and high-pass filtering will cause use artifacts at the edges of the recording, which will spoil your analyses, especially if you epoch the data over shorter segments.
No need for channel interpolation if you focus on source mapping.
Brainstorm will take care of removing the BAD segments from further analyses.
If you're looking at resting state, you can indicate that the data covariance be derived from the entire recording. If you do not have empty-room recordings, the noise covariance can be set to Identity (no info on noise).
The online tutorial at the bottom left redirects to https://neuroimage.usc.edu/brainstorm/Tutorials/SourceEstimation#Z-score.
So, does this standardise the sensor time series or the computed source time series ?
Also, in the tutorials, baseline normalisation is done before epoching in activity based recording. I am analysing resting state MEG data. Should I be doing the baseline normalisation as I do not intend to epoch my data and analyse on the source time series.
I had already performed high pass filtering and removed the artefacts. I do not intend to epoch the data and will be the trimming the starting's 10 sec and ending's 10 sec data for further analysis. In this case, I believe - I wouldn't have to the DC correction. Is it right ?
Great, so if you have already high-passed filtered this long recording, this will have taken care of the large DC offset at each sensor. You therefore do not need to apply further DC/baseline correction.