Polhemus fiducials wrong (CTF)

Hello there,

I found the fiducials in my dataset's polhemus files are wrongly defined, thus all the head models are unaligned with the skulls. Is there a way to create head models without polhemus files? I learned that the CTF pipeline has ways to create head models based on only the MRI, I wonder if there's similar approach in Brainstorm as well.
Thank you so much!

If you don't know where the head is located in the MEG helmet, unfortunately there is no way to reconstruct this information. If you can't align correctly the sensors on the anatomy, there is no way to calculate a correct head model.

The best you can do is align manually the helmet on the head (right-click on the channel file > MRI registration > Edit), and move the head where you think it was during the MEG acquisition.

Thank you Francois for your reply! But if I were to create the head model in CTF softwares, can I import it to Brainstorm to continue with source modelling in Brainstorm?
Thanks again!


No, you can’t import the CTF forward models in Brainstorm.
If the position of the head is missing, the CTF software would not be able to invent one either... you would just have fewer possibilities to visualize that your results are wrong with a default head position...

But I learned I can mark the fiducials on the structural MRI data in MRIviewer and to create a head model that way. Was that also impossible to do? Thanks!
I do have the individual MRI scans for each participants, only the polhemus fiducials are wrongly placed.

I'm not sure I understand what you are expecting to do here.
In order to register the anatomy and the sensors, you need the fiducials to be defined TWICE: in the MRI and in the MEG sensor space. If the fiducials are incorrect in one of the spaces, the sensors and the anatomy are not registered and the forward model does not make sense.
Please read the tutorial I pointed out previously.

Maybe there is an issue with the vocabulary. A "headmodel" in FieldTrip or CTF terms defines only the anatomy, indeed, and you don't need the MEG-space fiducials. But this is completely unrelated with the MEG activity. What do you expect from it?

Hi Francois, you are right! I did confuse the two definitions of head models.
My current goal is to do coregistration. But as you know now, the fiducials in the polhemus files are off (I was not involved in the data collection), which makes the coregistration inaccurate.
I think now the only solution would be to manually coresgistrate it as you suggested. But my biggest concern is whether the movements of the helmet surface would be justified, and to what degree.
Thank you very much for your time!


But my biggest concern is whether the movements of the helmet surface would be justified

No, this is not acceptable technically. But it's either this or you discard all the data of these participants.

and to what degree

The importance of the anatomy-MEG registration depends on what you do with the recordings.
If your registration is off by a translation of 1cm, your brain sources will be be off by one 1cm. Combined with a rotation of 10-20 degrees, you can easily reach a few cm of localization error in some regions.

The dynamics will remain unchanged, therefore if source localization is not your primary concern, it might not be a major issue. If you guide the analysis with the data, and look for instance at the temporal evolution of a ROI designed for each subject at the peak of activity, your results won't be much affected.