Resting-state MEG: no empty room recording

Discussions
1

Hi everyone,
I am analyzing resting state MEG data and would like to do source analyses, but this dataset does not include empty-room recordings.
I would like to know what would be the best approach to compute the noise covariance matrix: Would it be correct to compute the noise covariance matrix on the whole resting-state period ? Is there a better alternative ?

Thanks,

Thomas

2

Hi Thomas,

The best would be to use empty room measurements, as recommended for all types of MEG recordings:
https://neuroimage.usc.edu/brainstorm/Tutorials/NoiseCovariance#Variations_on_how_to_estimate_sample_noise_covariance

If it’s already too late and you don’t have access to them, maybe try without any noise modeling (noise covariance matrix = identity).

Francois

3

Hi Francois,
Ok thanks I will do that. Unfortunately there is no empty room measurements in the database I am using.

Thomas

4

I saw that in the Brainstorm tutorial for the HCP dataset the use of a noise covariance matrix including the whole file is recommended.
Do you think that no noise modeling is better than a noise modeling using the whole duration ?

Thomas

5

I have no idea, sorry…
The most active brain regions in the recordings used to compute the noise covariance matrix will tend to disappear in the source model. But it could be still better than using a completely uninformed model.
Try both and see…

6

Ok I will let you know