I tried to maxwell filter meg data collected with 4D-neuroimaging/BTi
system as follows:
raw_tsss = mne.preprocessing.maxwell_filter(raw, st_only=True,
ignore_ref=True,destination=trans,st_duration=60)
However, the data looks very noisy after filtering:
Original data:
[image: original.png]
sss filtered data:
[image: maxfiltered_SSS.png]
tsss filtered data (left) and original data (right):
[image: original_vs_maxfiltered.png]
I wonder if anyone knows what could be wrong?
Is it a problem if there are discontinuities in the data (all data is not
collected on the same day).
I?m not one of the MNE experts but have some ideas about how MaxFilter should work.
Could you check some things:
* Are there some bad channels left that could be spread noise into the entire sensor field by the MaxFilter? For example sensor A248 looks like that you should mark it as bad and let MaxFilter reconstruct it based on the remaining good channels.
* Are the correct calibration files of the MaxFilter used? Wrong calibration files could also introduce disturbances into the data.
I hope that might help.
Is it important to run Maxfilter on your BTi data? As far as I remember the testing and adaptation to other systems than Neuromag is still in progress. Let?s see what Eric says.
And the online compensation by reference sensors of the BTi system usually does a good job in producing reasonably clean recordings.
You can also regress out explicitly the signals captured by reference channels, here the example with the HCP data.
The result images I sent in my previous mail were calculated using
ignore_ref=True.
I tried to remove bad channels before filtering but, unfortunately, this
didn't solve the problem.
Are calibration files needed if I only have magnetometer and not
gradiometer data? I think that I don't even have the calibration files.
However, I think that, as also Denis said, it is perhaps not that important
to maxwell filetr the data. I also managed to filter the data with another
filter that corresponds to maxwell filter and is previously used with
bti data. The main reason why I wanted to try maxwell was that aligning the
heads of the subject (i.e. destination option) would have been easy with
maxwell filter.
Best,
Maria
ma 3. jouluk. 2018 klo 21.02 Eric Larson (larson.eric.d at gmail.com)
kirjoitti:
External Email - Use Caution
However, the data looks very noisy after filtering:
Does it look any better if you use ignore_ref=True in maxwell_filter?
Eric
_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu Mne_analysis Info Page
our Triux is useing fine calibration files and I was assuming that such files, or similar, might be available in your system as well. But it is possible that this is specific for Elekta systems.
