Dear list,
I wonder whether there is a MNE-way of removing power-chor artifacts
from wide-band MEG data (1-200 Hz).
As there does not seem to be a notch filter available, empty room SSPs
might be the way to do it.
What do you think?
Denis
hi Denis,
I wonder whether there is a MNE-way of removing power-chor artifacts
from wide-band MEG data (1-200 Hz).
you mean the AC current and the harmonics of 50 or 60Hz?
As there does not seem to be a notch filter available, empty room SSPs
might be the way to do it.
yes don't use SSP for this.
You need a notch like filter but I am not aware of any with mne or mne-python.
Alex
2012/9/12 Alexandre Gramfort <gramfort at nmr.mgh.harvard.edu>:
hi Denis,
I wonder whether there is a MNE-way of removing power-chor artifacts
from wide-band MEG data (1-200 Hz).you mean the AC current and the harmonics of 50 or 60Hz?
yes.
As there does not seem to be a notch filter available, empty room SSPs
might be the way to do it.yes don't use SSP for this.
ok, thanks though
Hi Denis,
As Alex said SSP cant be used to remove line frequency, as its comes from
electronics.
If you go mne-Matlab way, you can use eegfilt to have notch filter or try
harmonics filter which works better for line frequency. For reading and
writing raw fif file look at mne_ex_read_write_raw.m.
sheraz
Hi Denis,
The multitaper based rmlinesc() routine in Chronux to remove line noise
works well:
http://www.chronux.org/downloads/chronux/chronux/documentation/chronux_2_00/spectral_analysis/continuous/rmlinesc.html
The following is a reference that describes the technique:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1300074/
Also, most of the time I don't remove it so when I do time-frequency
decompositions I know that I have the scales correctly done when I see the
line components at the correct frequency..
Regards,
Hari
Dear Sheraz and Hari -- thanks for the good pointers!
(I hope you did not receive this already, my mail application was
going wild this morning...)
2012/9/13 Hari Bharadwaj <hari at nmr.mgh.harvard.edu>:
Hi Denis,
The multitaper based rmlinesc() routine in Chronux to remove line noise
works well:
http://www.chronux.org/downloads/chronux/chronux/documentation/chronux_2_00/spectral_analysis/continuous/rmlinesc.htmlThe following is a reference that describes the technique:
Analysis of dynamic brain imaging data. - PMCAlso, most of the time I don't remove it so when I do time-frequency
decompositions I know that I have the scales correctly done when I see the
line components at the correct frequency..
Actually very neat. I haven't thought of that..
Hello,
If I may put my two cents in... Most of the 50/60-Hz interference is
really the ambient magnetic field. This is the case at least with the
Elekta Neuromag systems and I believe with most other systems, too.
Thus, a properly-derived SSP operator should reduce the 50/60-Hz level
considerably but as such a projection has no notion of time, the same
spatial filtering is done unnecessarily at all frequencies. If you want
to keep the spatial patterns as intact as possible, then a notch is a
better option as Alex pointed out.
Cheers,
Lauri