50Hz artefact for time frequency analysis

Dear MNE users,

I have some data in which we did have some problem during the acquisition of
MEG/EEG combined.

Indeed some EEG electrodes picked up the frequency oscillation of the HPI
coils. It should not be a problem as the coils frequency is above 300 HZ.
However I just noticed in my data that some EEG electrodes (4 out of 70 in
average) have an harmonic around 50HZ.

If I'm doing ERF/ERPs, my data will be filtering this artefact, but
unfortunately I want to do time-frequency analysis on sensors & source
level.

On unfiltered data (1-100Hz) , those EEG electrodes are picked as artefact
and rejected (amplitude of oscillation ~ 300-400uV). To pass the data I
needed to mark those electrodes as bad in MNE.

Now I thought about different solution to remove this and would like to have
your advice on which one to do, or maybe you will have a better suggestion..

a) Use a notch filter, but then what about the gamma band? Will I still
be able to analyse it or not?

b) Interpolate those specific electrodes, there are maximum 4 out of 70

c) As power computation is baseline corrected it should not be a
problem; the oscillations are also in the baseline and then will be
automatically removed (same for PLV?)

Thanks for your help

Elisabeth

hi Elisabeth,

I have some data in which we did have some problem during the acquisition of
MEG/EEG combined.

Indeed some EEG electrodes picked up the frequency oscillation of the HPI
coils. It should not be a problem as the coils frequency is above 300 HZ.
However I just noticed in my data that some EEG electrodes (4 out of 70 in
average) have an harmonic around 50HZ.

If I?m doing ERF/ERPs, my data will be filtering this artefact, but
unfortunately I want to do time-frequency analysis on sensors & source
level.

On unfiltered data (1-100Hz) , those EEG electrodes are picked as artefact
and rejected (amplitude of oscillation ~ 300-400uV). To pass the data I
needed to mark those electrodes as bad in MNE.

Now I thought about different solution to remove this and would like to have
your advice on which one to do, or maybe you will have a better suggestion..

a) Use a notch filter, but then what about the gamma band? Will I still
be able to analyse it or not?

why not?

b) Interpolate those specific electrodes, there are maximum 4 out of 70

I would say it depends on where are the 4 electrodes.

c) As power computation is baseline corrected it should not be a
problem; the oscillations are also in the baseline and then will be
automatically removed (same for PLV?)

yes indeed.

Best,
Alex

Hi Elisabeth,

A note related to the origins of your 50-Hz artefact: The MEG HPI coils
should not induce any detectable currents in your EEG electrodes.
Moreover, 300 Hz (or, the 290-33 Hz range that is typically used for cHPI)
does not have sub-harmonics at 50 Hz. It is much more likely that you had
a bad EEG connection/electrode/wire at those 4 EEG electrodes, resulting
in that they picked up the 50 Hz in the electric mains.

Best regards,

Tommi

Hi Elisabeth,

I tend to agree with Tommi. We did not experience this sort of noise with
combined MEEG. When you look at the power spectra of your EEG signals, do
you actually see the HPI peak responses? Such high amplitude in the signals
likely correspond to lower frequency - e.g. muscle artifacts? Or possible
arterial artifact (were these specific EEG electrodes located at the point
of contact between the skull and the dewar?)

If the spectral artifact is steady over time, baseline correction will work
and so will contrasting conditions against each other - if you want to
avoid notch artifacts in higher frequency bands.

Best,
Virginie