[Mne_analysis] head position information when building forward model for different runs

Fri Mar 22 01:24:38 EDT 2019

        External Email - Use Caution        

Hi Lin,

You can use this to bring runs into a common head position (copy-pasted
from Alex's response to an older post):
https://gist.github.com/jasmainak/756cf02dce82ab2dc5c2c69722dd13d1

It's basically using the MNE method on a spherical surface to bring them to
a common head position. I will not vouch for its accuracy compared to
Maxfilter as I have not tested it. However, this is what the 2005 SSS paper
says:

"The problem caused by movement can be solved by using the minimum norm
estimate as a source model for
transforming the measured signals to correspond to a reference head
position. The device-independent components of SSS are a similar source
model with the benefit of modeling also the external interference signals."

I think it's a chicken and an egg problem to some extent. How can you get
an accurate noise covariance to bring the data to a common head position if
the head position is not fixed?

Best,
Mainak

On Thu, Mar 21, 2019 at 4:51 PM Lin Wang <wanglinsisi at gmail.com> wrote:

>         External Email - Use Caution
>
> Hi Ellen,
>
> Thanks for your response. Yeah, as Alex also pointed out, the averaging of
> the dSPM values is not linear because of the noise normalization, which
> takes the number of epochs into account.
>
> But if we average the MNE values along the 'normal' orientation for the
> loose orientation constraints, the operation should be linear, i.e. the
> average of MNE values across runs is the same as the MNE of the averaged
> ERF across runs.
>
> Best,
> Lin
>
>
>
> On Thu, 21 Mar 2019 at 16:36, Ellen Lau <ellenlau at umd.edu> wrote:
>
>>         External Email - Use Caution
>>
>> Hi Lin,
>>
>> In a related conversation I was recently reminded that the inverse
>> solution may not be linear when the orientation is loose rather than fixed,
>> maybe that is contributing to the difference?
>>
>> Ellen
>>
>> >
>> > Message: 1
>> > Date: Thu, 21 Mar 2019 13:05:37 -0400
>> > From: Lin Wang <wanglinsisi at gmail.com>
>> > Subject: [Mne_analysis] head position information when building
>> >       forward model for different runs
>> > To: Discussion and support forum for the users of MNE Software
>> >       <mne_analysis at nmr.mgh.harvard.edu>
>> > Message-ID:
>> >       <CADsMai1N2JQzovOAj7rt8aUT_gS6C=
>> SKyW1j40D9HtQMHr-6wQ at mail.gmail.com>
>> > Content-Type: text/plain; charset="utf-8"
>> >
>> >        External Email - Use Caution
>> >
>> > Hi MNE experts,
>> >
>> > I have a question about what head position information to use when
>> building
>> > a forward model for each run in one participant.
>> >
>> > We have eight runs of MEG data. At the beginning, we used the head
>> position
>> > of the first run to build just one forward model for one participant.
>> Then
>> > we thought it might be more accurate to use the run-specific head
>> position
>> > to build forward models separately for different runs. In both analyses,
>> > the forward models were used to calculate the inverse operators, which
>> were
>> > applied to the evoked response for each run. We then averaged the
>> > activation across runs within each participant. Finally, we compared the
>> > activation difference between two conditions at the group level.
>> >
>> > Although the group-averaged activation looks very similar from the two
>> > analyses, the use of run-specific head position reduced the statistical
>> > power at the group level. Do you know why?
>> > Is there a better way to account for the head movement within each run?
>> We
>> > didn't apply the MaxFilter to the data because the head movement was not
>> > considered to be serious during the data acquisition. Is there a way to
>> > incorporate the head movement within each run or the whole experiment
>> > without running the MaxFilter?
>> >
>> > Thanks a lot for your input!
>> >
>> > Best,
>> > Lin
>> > -------------- next part --------------
>> > An HTML attachment was scrubbed...
>> > URL:
>> http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190321/67dac610/attachment-0001.html
>> >
>> > ------------------------------
>> >
>> > Message: 2
>> > Date: Thu, 21 Mar 2019 21:11:44 +0100
>> > From: Alexandre Gramfort <alexandre.gramfort at inria.fr>
>> > Subject: Re: [Mne_analysis] head position information when building
>> >       forward model for different runs
>> > To: Discussion and support forum for the users of MNE Software
>> >       <mne_analysis at nmr.mgh.harvard.edu>
>> > Message-ID:
>> >       <
>> CADeotZogU6vbv85Se-oQH3odoEbv34bEVLK2TfOQcU8JF9JjSw at mail.gmail.com>
>> > Content-Type: text/plain; charset="utf-8"
>> >
>> >        External Email - Use Caution
>> >
>> > hi Lin,
>> >
>> > did you average dSPM values? as given the noise normalization the
>> average
>> > of the dSPM is not the dSPM of the average. If could make a big
>> difference
>> > unless all runs have the same number of epochs in all conditions.
>> >
>> > Alex
>> >
>> >
>> > On Thu, Mar 21, 2019 at 6:07 PM Lin Wang <wanglinsisi at gmail.com> wrote:
>> >
>> >>        External Email - Use Caution
>> >>
>> >> Hi MNE experts,
>> >>
>> >> I have a question about what head position information to use when
>> >> building a forward model for each run in one participant.
>> >>
>> >> We have eight runs of MEG data. At the beginning, we used the head
>> >> position of the first run to build just one forward model for one
>> >> participant. Then we thought it might be more accurate to use the
>> >> run-specific head position to build forward models separately for
>> different
>> >> runs. In both analyses, the forward models were used to calculate the
>> >> inverse operators, which were applied to the evoked response for each
>> run.
>> >> We then averaged the activation across runs within each participant.
>> >> Finally, we compared the activation difference between two conditions
>> at
>> >> the group level.
>> >>
>> >> Although the group-averaged activation looks very similar from the two
>> >> analyses, the use of run-specific head position reduced the statistical
>> >> power at the group level. Do you know why?
>> >> Is there a better way to account for the head movement within each
>> run? We
>> >> didn't apply the MaxFilter to the data because the head movement was
>> not
>> >> considered to be serious during the data acquisition. Is there a way to
>> >> incorporate the head movement within each run or the whole experiment
>> >> without running the MaxFilter?
>> >>
>> >> Thanks a lot for your input!
>> >>
>> >> Best,
>> >> Lin
>> >> _______________________________________________
>> >> Mne_analysis mailing list
>> >> Mne_analysis at nmr.mgh.harvard.edu
>> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>> > -------------- next part --------------
>> > An HTML attachment was scrubbed...
>> > URL:
>> http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190321/6d85588a/attachment-0001.html
>> >
>> > *********************************************
>>
>>
>>
>> _______________________________________________
>> Mne_analysis mailing list
>> Mne_analysis at nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>>
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190322/c0702c5e/attachment-0001.html 