MEG-MRI registration

Discussions
, ,
1

Dear Francois and all,

I am sorry to bother you. I have been troubled in the MEG-MRI registration for a long time and I really appreciate your help.
I have imported MRI with anatomy folders (using freesurfer) and MEG .fif raw files (using the Elekta-Neuromag option) and performed MEG-MRI registration with reference to the tutorial https://neuroimage.usc.edu/brainstorm/Tutorials/ChannelFile#Automatic_registration.
I also conducted the manual registration, and the final results are shown in the follwing pictures.
It can be seen from the figures that the Polhemus fiducial ear points are inside the head and it also a little bit distant from the MRI ear points. Some head points along are also inside the head, especially on the left-hand side.
What effect will this have on the source imaging results? And how should I adjust this?
Very looking forward to your replies.
Liu Xinyan

图片1
图片1651×518 92.9 KB
图片2
图片2656×515 86.3 KB
图片3

2

It seems to be not a big problem with the registration process however you need to adjust the helmet (green dots) as some points are outside the helmet. You can adjust it by selecting one or all X, Y, Z coordinates and move the helmet with arrow keys, once it is corrected you can press OK.
The source analysis depends on the accuracy of alignment between MR registration and sensors. The sensor positions in green dots outside the head will create more distance from the real head which eventually leads to poor source analysis.
image

3

Dear Abrar,
Thank you for the clear answer. I adjusted it according to your suggestion. However, it seems that I couldn’t guarantee that all the points are on the head, because in this way, the helmet and the head will intersect. Do you have any suggestion for me?
Additionally, In my second subject, I couldn't align the helmet, as shown in Figure 2. If I try to align it, the points on the nose and the NAS will shift to the left, just like in the third picture. Which outcome should I prefer?
Thanks again for your help.
Best,
Liu Xinyan

图片4
图片4674×518 93.8 KB
图片5
图片5690×550 81.5 KB
图片6
图片6700×529 84.2 KB

4

It can be seen from the figures that the Polhemus fiducial ear points are inside the head and it also a little bit distant from the MRI ear points.

The registration in your first post doesn't look too bad.

For fiducial-based registration:
However, it looks like you are using different points for the LPA/RPA points in the MEG acquisition (orage points, somewhere in from of the ear) and in the anatomy preparation in Brainstorm (blue points, between the helix and the tragus). If you MEG was recorded using the "real pre-auricular points", you must use the same points when importing the subject's anatomy in Brainstorm.

For automatic registration:
Remove the nose points before processing: right-click on the channel file > Digitized head points > Remove points below nasion. Example:
https://neuroimage.usc.edu/brainstorm/Tutorials/VisualSingle#MRI_registration

For manual registration:
You have to be extremely careful with what you are doing. Whatever you do manually cannot be reproduced, and introduces a bias between subjects. Not recommended when you have other options. Depending on the quality of your digitized head shape (which looks indeed quite noisy...), I recommend you try using either the fiducial-based or the headshape-based registration for all your participants, and fix manually only the issues you may have for isolated subjects with very poor head shape quality.

What effect will this have on the source imaging results?

Moving your sensors by 1cm would result in shifting your estimated sources by ~1cm.
The same goes for the rotations along the 3 axes.

In my second subject, I couldn't align the helmet, as shown in Figure 2. If I try to align it, the points on the nose and the NAS will shift to the left, just like in the third picture

Regarding your second post: the head shape (the green points) look a lot larger than the participant's head. Are you sure you didn't make a mistake matching the anatomy and recordings?
If you didn't, you may have some distortions cause by your digitizing environment (magnetic fields altering the Polhemus measures). Double-check this before recording more subjects.

image
image690×550 81.6 KB

5

Dear Francois，
Thanks for your patient reply, which is very helpful to me.

I read the fiducials and the measured headshape through fieldtrip and ft_read_headshape.m:
dataset='meg01.fif';
shape = ft_read_headshape(dataset);
headshape = ft_convert_units(shape, 'mm');
ft_plot_headshape(headshape);
and got something like this:
headshape.fid.pos =
[ -70.7639837265015 2.62228383007823e-06 -6.45878913019260e-06
2.22098623936517e-07 93.8694953918457 -4.44904912910715e-06
69.3212890625000 5.53332768049586e-06 -6.11441748787911e-06 ]
headshape.fid.label =
'LPA'
'Nasion'
'RPA'
headshape.unit = ’ mm ’
图片7

  1. How do I understand the meaning of headshape.fid.pos? Does it have anything to do with SCS coordinates?
  2. Can I specify the fiducials position of MRI as headshape.fid.pos? Just like this,
    图片8
    图片8832×503 9.55 KB

    I tried to do it, but these coordinates can't be saved. That is, when I clicked OK and then opened “edit fiducial positions” again, the coordinates would automatically changed. Does this mean that it is not feasible to specify the financial position of MRI as headshape.fid.pos?
  3. After I reassigned the coordinates of the MRI fiducial points, I got the following registration results:
    图片9
    图片9742×630 102 KB
  4. After that, I removed points below nasion and refined using head points. I got the following registration results:
    图片10
  5. As can be seen from the figure above, the MEG helmet was tilted to the right . Does this mean that I have to perform manual registration? After the manual registration, the final results are shown in the follwing pictures. It looks like the NAS is on the left. Is that reasonable?
    图片11
    Did I do anything wrong in the process? I will feel grateful extremely if you could give me any help.
    Best,
    Liu Xinyan
6
  1. How do I understand the meaning of headshape.fid.pos? Does it have anything to do with SCS coordinates?

I'm sorry, I'm not competent to tell you how FieldTrip reading functions work exactly, or what you would get from reading a FIF file with ft_read_headshape. Please refer to the FieldTrip documentation or the FieldTrip support channels (Support - FieldTrip toolbox) for more information.

When reading files in Brainstorm, everything is converted to CTF/SCS which is clearly documented in the online tutorials:
https://neuroimage.usc.edu/brainstorm/CoordinateSystems

  1. Can I specify the fiducials position of MRI as headshape.fid.pos? Just like this,

You can, but it sets the MRI fiducials, not the MEG fiducials, so I'm not sure what you'd expect to do here. Make sure you understand clearly all the documentation about coordinate systems and MEG/MRI co-registration before trying to play randomly with the registration information. This is a critical point of the MEG processing pipeline, which is also highly prone to misunderstanding and manipulation errors.
https://neuroimage.usc.edu/brainstorm/Tutorials/ChannelFile#Automatic_registration

Does this mean that it is not feasible to specify the financial position of MRI as headshape.fid.pos?

The fiducials coordinates you want to use (from this .fif file) are the coordinates that were digitized during the MEG acquisition. They make sense only if they are used together with the MEG sensors coordinates and all the extra digitized head points. Using them arbitrarily for any other purpose than applying a spatial transformation to the MEG sensors+head points does not make sense.

  1. After I reassigned the coordinates of the MRI fiducial points, I got the following registration results:

As explained, I don't think this makes much sense.

  1. After that, I removed points below nasion and refined using head points. I got the following registration results:

As mentioned in my previous post: if you use this automatic registration based on the head shape, all the refinements you can bring to the NAS/LPA/RPA fiducials before (your step #3) are useless. The previous MRI-MEG registration is completely replaced.

  1. As can be seen from the figure above, the MEG helmet was tilted to the right . Does this mean that I have to perform manual registration? After the manual registration, the final results are shown in the follwing pictures. It looks like the NAS is on the left. Is that reasonable?

This is up to you to decide, you should know where the subject was in the MEG OR trust your methodology. Unfortunately, we can't magically fix poor acquisition conditions or procedures.
Either you trust the NAS/LPA/RPA fiducials in the FIF files (which are already correctly taken into account without you trying to do any additional stuff with FieldTrip functions), or you trust the head shape (which looks noisy, and probably too large - as I mentioned in my previous post), or you consider that none of this stuff works and you will try to adjust manually the position of the head in the helmet.

Just remember that what to want to see is the real position that the participant had when he/she was sitting in the MEG, not what you think looks better. If the head of the subject was titled, or far from down away from the sensors, or if he/she was moving a lot: there is nothing you can do about that.

7

Dear Francois,
I'm sincerely grateful for all your help. With your help, I have a more profound understanding on MEG/MRI co-registration.
Best wishes,
Liu Xinyan

8

Dear Francois,
If the head shape (the green points) look a lot larger than the participant's head, could I use the EEG edit for head scaling and apply it to MEG? Just as shown in the figure below:

图片27
图片271396×963 123 KB

Thanks for your kind attention and look forward your prompt reply.
Best Regards,
Han Xiaoke

9

If the head shape (the green points) looks a lot larger than the participant's head, there was probably an issue during its acquisition. Something in the room could be distorting the magnetic fields recorded by the Polhemus. You should be very suspicious about this file, and not rely much on the MEG head registration...
Do the best you can in guessing where the head was in the helmet.

Projecting the EEG electrodes on the head is a good idea.
But you can't apply anything similar to the MEG sensors: this is a rigid array and its shape does not depend on the head shape.

10

Dear Francois,
Thank you for your reply. All in all, the approach is inappropriate, right?
But I still don't understand. After EEG edit adjusts the size of the head shape, there will be a prompt asking Do you want to apply the same transformation to the MEG sensors . If I choose Yes, the head shape will change, but then the source result may be wrong, right?
图片28
In addition, it seems that the head size can be adjusted in Fieldtrip. Is this different from the adjustment in Brainstorm?

图片29
图片29891×524 62.4 KB

I'm sorry to bother you again. I'm not very familiar with the acquisition process of MEG, so these questions may also be very basic. Thank you again for your patience.
Yours,
Han Xiaoke

11

This refers only to the rigid transformations you applied (rotation + translation), not to the resizing or projections on the scalp.

In addition, it seems that the head size can be adjusted in Fieldtrip. Is this different from the adjustment in Brainstorm?

I don't know how FieldTrip handles this, I'm sorry I can't help you with this.

12

Dear Francois,
Thank you very much for your help.
Best Regards,
Han Xiaoke