Calculate forward model using digitization from eximia system

Mailing List Archive (read-only)
1

Dear List,

I'm following this tutorial

http://martinos.org/mne/stable/auto_tutorials/plot_forward.html

and trying to compute the forward model using individual anatomical
information.

So far I got the skull/skin surfaces using mne watershed_bem, and now I'm
trying to use the mne.gui.coregistration the generate the -trans file.

However, the digitization information is an eximia .nbe file. I found a
thread about converting it to FIF

https://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/2013-March/001465.html

but it suggests using the mne_eximia2fif, and my eeg data is not acquired
with the eximia system, only the digitization.

The digitization consists of 3D positions for each electrodes + fiducial,
which I have imported into python as numpy.array.

Should I use mne.channels.read_dig_montage and update info['dig'] manually?

I tried

import mne.channels.read_dig_montage
digitization = read_dig_montage(elp=my_data[:, 1:], point_names=ch_names)

where my_data[:, 1:] are the coordinates of the electrodes in mm, and the
first 3 lines are ['nasion', 'lpa', 'rpa'] (ch_names has the strings with
the fiducial and electrodes names)

However I get this error:

2

Hi Leonardo,

I'm following this tutorial

http://martinos.org/mne/stable/auto_tutorials/plot_forward.html

and trying to compute the forward model using individual anatomical
information.

So far I got the skull/skin surfaces using mne watershed_bem, and now I'm
trying to use the mne.gui.coregistration the generate the -trans file.

However, the digitization information is an eximia .nbe file. I found a
thread about converting it to FIF

[Mne_analysis] mne_eximia2fiff: digitizer data ordering

but it suggests using the mne_eximia2fif, and my eeg data is not acquired
with the eximia system, only the digitization.

The digitization consists of 3D positions for each electrodes + fiducial,
which I have imported into python as numpy.array.

Should I use mne.channels.read_dig_montage and update info['dig'] manually?

yes that's the entry point.

see the structure of a DigMontage object and maybe see if you can create one
directly.

let us know if it works

Alex

3

I know it has been a while, but I finally went back to this project and
managed to create the DigMontage using the mne.channels.read_dig_montage

        digitization = read_dig_montage(hsp=positions.copy(),
elp=positions.copy(), point_names=labels, unit='mm')

positions is a ndarray (N,3) and labels is a list (N), where N is the
number of entries in the digitization file: 257 electrodes + 3 fiducials
here is an example of .sfp file I get from the digitization process and
used to generate the position and labels objects through csv.reader() :

https://drive.google.com/open?id=0B0p7WZ2wlTFSazJWbENnN0JnMkU

I then used set_montage to update the info['dig'] entry ( knowing that I
already called this function before with a normal Montage, to set the
electrodes positions, since info['ch_names'] is not updated by
read_dig_montage)

        raw.set_montage(digitization)

It works, but I'm not sure if this is the correct way to do it.
*Is it ok to use electrodes as head points?*
*Is it not enough to use the fiducials alone?*

Perhaps I should open a new thread less specific about eximia, since my
question is more generic:* how can I align electrodes to individual MRIs
just having the fiducials from the digitization process in a text file?*

I found this thread, but no it is not quite the same since I do have the
MRI, but the co-registration is not in any supported format (Polhemus, etc)

https://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/2016-June/003414.html

In any case, now I can save the raw with the info['dig'] information (lpa,
rpa and nasion) and use it to create the trans file using the
mne.gui.coregistration (through "mne coreg" in the command line,
http://www.slideshare.net/mne-python/mnepython-coregistration)

However, for some reason all the electrodes are quite far from the scalp,
even though the fiducials were aligned:

https://drive.google.com/open?id=0B0p7WZ2wlTFSNklmcTlPakJERXc
https://drive.google.com/open?id=0B0p7WZ2wlTFScXF4NGMtNXBiZ2M

since it happened with 5 subjects, and the error is always similar, I think
it is unlikely to be a simple switch between subjects digitization (I
double checked the origin of the files in any case)
it could be, however, a systematic error in the acquisition... But this
system was successfully used with brainstorm, so I would not start with
this hypothesis.

*Does anybody have another idea why this could be happening?*
*Or how could I fix that?*

I could morph the MRI if needed, but since I'm using individual MRIs with
the co-registration, wouldn't that be a waste and similar to using
generic/fsaverage MRI?

Thank you for any help,
Leonardo

2016-09-09 8:07 GMT-05:00 Alexandre Gramfort <alexandre.gramfort at telecom-
paristech.fr>:

Hi Leonardo,

> I'm following this tutorial
>
> http://martinos.org/mne/stable/auto_tutorials/plot_forward.html
>
> and trying to compute the forward model using individual anatomical
> information.
>
> So far I got the skull/skin surfaces using mne watershed_bem, and now I'm
> trying to use the mne.gui.coregistration the generate the -trans file.
>
> However, the digitization information is an eximia .nbe file. I found a
> thread about converting it to FIF
>
> https://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/2013
-March/001465.html
>
> but it suggests using the mne_eximia2fif, and my eeg data is not acquired
> with the eximia system, only the digitization.
>
> The digitization consists of 3D positions for each electrodes + fiducial,
> which I have imported into python as numpy.array.
>
> Should I use mne.channels.read_dig_montage and update info['dig']
manually?

yes that's the entry point.

see the structure of a DigMontage object and maybe see if you can create
one
directly.

let us know if it works

Alex

> I tried
>
> import mne.channels.read_dig_montage
> digitization = read_dig_montage(elp=my_data[:, 1:],
point_names=ch_names)
>
> where my_data[:, 1:] are the coordinates of the electrodes in mm, and the
> first 3 lines are ['nasion', 'lpa', 'rpa'] (ch_names has the strings with
> the fiducial and electrodes names)
>
> However I get this error:
>
> ===========================================
>
> In [48]: read_dig_montage(elp=my_data[:, 1:],
point_names=montage.ch_names)
> ------------------------------------------------------------
---------------
> TypeError Traceback (most recent call
last)
> <ipython-input-48-c58c337206f3> in <module>()
> ----> 1 read_dig_montage(hsp=None, elp=my_data[:, 1:],
> point_names=montage.ch_names)
>
> /home/leonardo/projects/mne-python/mne/channels/montage.pyc in
> read_dig_montage(hsp, hpi, elp, point_names, unit, fif, transform,
> dev_head_t)
> 608 fids = apply_trans(neuromag_trans, [nasion, lpa,
rpa])
> 609 elp = apply_trans(neuromag_trans, elp)
> --> 610 hsp = apply_trans(neuromag_trans, hsp)
> 611 else:
> 612 fids = [None] * 3
>
> /home/leonardo/projects/mne-python/mne/transforms.pyc in
apply_trans(trans,
> pts, move)
> 172
> 173 # apply rotation & scale
> --> 174 out_pts = np.dot(pts, trans[:3, :3].T)
> 175 # apply translation
> 176 if move is True:
>
> TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'
>
> ===========================================
>
> Looks like hsp is mandatory, but I don't have the Polhemus information...
>
> Is there an easier way to create the -trans file and calculate the
forward
> model using individual electrodes positions from the eximia system (or
> np.array with the electrodes coordinates for that matter)?
>
> Thank you very much in advance for any help,
>
> Leonardo
>
>
>
>
>
>
>
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>
>
> The information in this e-mail is intended only for the person to whom
it is
> addressed. If you believe this e-mail was sent to you in error and the
> e-mail
> contains patient information, please contact the Partners Compliance
> HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you
in
> error
> but does not contain patient information, please contact the sender and
> properly
> dispose of the e-mail.
>
_______________________________________________
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/20161026/6727e387/attachment.html

4

I suspect the point digitized was not actually on the scalp. I have
seen other groups have systematic issues like this, when they don't
place the digitizer pen directly on the scalp (typically on top of an
electrode: which is often a bad idea for multiple reasons; although
occasionally unavoidable). MNE should handle this fine and use the
closest point on the scalp, but I would recommend for future data
collection that the pen is put inside the electrode on the scalp
itself (be careful, some digitizer pens are sharp).

hth
d

5

I know it has been a while, but I finally went back to this project and
managed to create the DigMontage using the mne.channels.read_dig_montage.
..
It works, but I'm not sure if this is the correct way to do it.

Given that your alignments look approximately correct (other than the
scaling issue!) I suspect these steps are correct. Can you open an
MNE-Python issue with this procedure? It would be good if we could turn it
into a tutorial or improve the docs to describe it.

*Is it ok to use electrodes as head points?*

If you don't have "extra" points directly on the scalp, I would say that
electrodes can be used, yes. But keep in mind how your digitization was
actually done during the experiment; some electrodes (e.g., BrainVision
actiCap <http://www.brainproducts.com/productdetails.php?id=4&gt;\) are
actually a bit thick, say half a centimeter, so if they were digitized on
the center of those electrodes, you'd expect them to be that far off of the
MRI during coregistration.

*Is it not enough to use the fiducials alone?*

You can use just the fiducials, but fitting a the rotation + translation
using three points manually placed on an MRI surface is might be less
robust than fitting extra dig points. So it depends on how much you trust
your fiducial digitization and your ability to pinpoint those equivalent
locations no an MRI.

Perhaps I should open a new thread less specific about eximia, since my

question is more generic:* how can I align electrodes to individual MRIs
just having the fiducials from the digitization process in a text file?*

We have an eXimia data-reading pull request open
<https://github.com/mne-tools/mne-python/pull/3523&gt;, and support is a
work-in-progress. So I suggest we discuss eXimia things there to converge
on something usable.

The digitization issues are more generic, and hopefully someone else can
comment more meaningfully. I don't have much experience with it. But in
theory once you have your info['dig'] populated properly via DigMontage,
using the coreg gui in MNE-Python, or even saving the data to FIF and using
mne_analyze should be possible. But it sounds like you have already gone
that route...

However, for some reason all the electrodes are quite far from the scalp,

even though the fiducials were aligned:

https://drive.google.com/open?id=0B0p7WZ2wlTFSNklmcTlPakJERXc
https://drive.google.com/open?id=0B0p7WZ2wlTFScXF4NGMtNXBiZ2M

since it happened with 5 subjects, and the error is always similar, I
think it is unlikely to be a simple switch between subjects digitization (I
double checked the origin of the files in any case)
it could be, however, a systematic error in the acquisition... But this
system was successfully used with brainstorm, so I would not start with
this hypothesis.

I assume by this you mean the plots looked okay in BrainStorm, so we need
to dig into it a bit.

It's possible that BrainStorm projects the electrodes onto the scalp for
visualization. Do you have data from one subject you could share? If so,
feel free to open an MNE-Python issue and we can work through it. For
example, as a first pass you could look at the distribution histogram of
the dig points in X and Y, and do the same thing for the MRI head surface
coordinates. You can also look at the transformation matrix and ensure that
it's a pure rotation + translation matrix, with no scaling (but it really
should be!).

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161026/9c9d3478/attachment.html

6

2016-10-26 11:59 GMT-05:00 dgw <dgwakeman at gmail.com>:

I suspect the point digitized was not actually on the scalp. I have
seen other groups have systematic issues like this, when they don't
place the digitizer pen directly on the scalp (typically on top of an
electrode: which is often a bad idea for multiple reasons; although
occasionally unavoidable).

I understand, since I was not there I do not know if this is the case.
Although I though it was a bit to far for that. I'll check if the distance
match the height of the electrodes!

MNE should handle this fine and use the
closest point on the scalp,

Cool, I saw this before in fieldtrip but I didn't see the option in
mne.channels.coregistration() gui
Which function is supposed to do that? mne.make_forward_solution() ?

Thanks!
Leonardo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161026/94e929a3/attachment.html

7

Cool, I saw this before in fieldtrip but I didn't see the option in
mne.channels.coregistration() gui
Which function is supposed to do that? mne.make_forward_solution() ?

Yeah, deep within the forward code it does this projection. I suppose we
could add that an option to project electrodes in plot_trans and/or the
GUI. Feel free to open an issue for that while we're at it.

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161026/dfd4b049/attachment.html

8

2016-10-26 12:02 GMT-05:00 Eric Larson <larson.eric.d at gmail.com>:

I know it has been a while, but I finally went back to this project and

managed to create the DigMontage using the mne.channels.read_dig_montage.
..
It works, but I'm not sure if this is the correct way to do it.

Given that your alignments look approximately correct (other than the
scaling issue!) I suspect these steps are correct. Can you open an
MNE-Python issue with this procedure? It would be good if we could turn it
into a tutorial or improve the docs to describe it.

I opened the issue in github, I can help to write the tutorial.

*Is it ok to use electrodes as head points?*

If you don't have "extra" points directly on the scalp, I would say that
electrodes can be used, yes. But keep in mind how your digitization was
actually done during the experiment; some electrodes (e.g., BrainVision
actiCap <actiCAP slim/snap | Brain Products GmbH) are
actually a bit thick, say half a centimeter, so if they were digitized on
the center of those electrodes, you'd expect them to be that far off of the
MRI during coregistration.

Indeed, I think this is what is happening. I'm checking to see if this was
the case.

*Is it not enough to use the fiducials alone?*

You can use just the fiducials, but fitting a the rotation + translation
using three points manually placed on an MRI surface is might be less
robust than fitting extra dig points. So it depends on how much you trust
your fiducial digitization and your ability to pinpoint those equivalent
locations no an MRI.

I see, the problem is the noise in the acquisition of the fiducials, which
could be minimized by extra points assuming the error in measurement is
gaussian. I always wandered why the extra points if three is enough in the
ideal world :wink:

Perhaps I should open a new thread less specific about eximia, since my

question is more generic:* how can I align electrodes to individual MRIs
just having the fiducials from the digitization process in a text file?*

We have an eXimia data-reading pull request open
<https://github.com/mne-tools/mne-python/pull/3523&gt;, and support is a
work-in-progress. So I suggest we discuss eXimia things there to converge
on something usable.

Since I do not have Eximia EEG data in this project (.nxe), I think it is a
bit different from this thread.

The digitization issues are more generic, and hopefully someone else can
comment more meaningfully. I don't have much experience with it. But in
theory once you have your info['dig'] populated properly via DigMontage,
using the coreg gui in MNE-Python, or even saving the data to FIF and using
mne_analyze should be possible. But it sounds like you have already gone
that route...

yes, I'm just not sure which is the best/simpler way of doing it.
For instance I think it is weird to call set_montage twice...

However, for some reason all the electrodes are quite far from the scalp,

even though the fiducials were aligned:

https://drive.google.com/open?id=0B0p7WZ2wlTFSNklmcTlPakJERXc
https://drive.google.com/open?id=0B0p7WZ2wlTFScXF4NGMtNXBiZ2M

since it happened with 5 subjects, and the error is always similar, I
think it is unlikely to be a simple switch between subjects digitization (I
double checked the origin of the files in any case)
it could be, however, a systematic error in the acquisition... But this
system was successfully used with brainstorm, so I would not start with
this hypothesis.

I assume by this you mean the plots looked okay in BrainStorm, so we need
to dig into it a bit.

Yes

It's possible that BrainStorm projects the electrodes onto the scalp for
visualization.

I also think this was the case

Do you have data from one subject you could share? If so, feel free to
open an MNE-Python issue and we can work through it.

I'll look for one example.

For example, as a first pass you could look at the distribution histogram
of the dig points in X and Y, and do the same thing for the MRI head
surface coordinates. You can also look at the transformation matrix and
ensure that it's a pure rotation + translation matrix, with no scaling (but
it really should be!).

It has to be, no? I never clicked the option to do scaling in the
mne.gui.coregistration!
I'll check the histograms, also of the distances betweem electrodes and
scalp (they seem to be quite constant, hence probably the elctrode height)

Thank you,
Leonardo

Eric

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

The information in this e-mail is intended only for the person to whom it
is
addressed. If you believe this e-mail was sent to you in error and the
e-mail
contains patient information, please contact the Partners Compliance
HelpLine at
MyComplianceReport.com: Compliance and Ethics Reporting . If the e-mail was sent to you in
error
but does not contain patient information, please contact the sender and
properly
dispose of the e-mail.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161026/75722c0e/attachment-0001.html

9

2016-10-26 12:35 GMT-05:00 Eric Larson <larson.eric.d at gmail.com>:

Cool, I saw this before in fieldtrip but I didn't see the option in

mne.channels.coregistration() gui
Which function is supposed to do that? mne.make_forward_solution() ?

Yeah, deep within the forward code it does this projection. I suppose we
could add that an option to project electrodes in plot_trans and/or the
GUI. Feel free to open an issue for that while we're at it.

I opened an issue about that

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161026/2312becc/attachment.html