mne/freesurfer tools compatibility with neuromag dicom access

šŸ’¬ Support & Discussions
1

Hi all,
Apologies for the screen shots :\ I can provide script snippets, but not sure what might be useful at this point.

Iā€™m working with a team that includes people trained to use neuromagā€™s dicom access to generate the fif format MRI (this is the pipeline they use for clinical data). They have done co -registration in neuromagā€™s MRIlab from this dicom access created fif files.

Iā€™m also working with some legacy code which uses mneā€™s freesurfer tools.

If I load the COR.fif file and the bem meshes created with my legacy mne/freesurfer script into MRIlab, everything looks wonky.

newAlign_mesh
newAlign_mesh1920Ɨ1200 198 KB

If I load the COR.fif file (as the trans), and use mne.viz.plot_alignment, things look wonky.

alignment_dgleichmann
alignment_dgleichmann1618Ɨ1522 121 KB

Are these processes compatible?

The legacy script was written for
mne 0.15.2
and is running on Python 3.6.9 (default, Jan 26 2021, 15:33:00) on
Distributor ID: CentOS
Description: CentOS release 6.7 (Final)
(this os is also what we need to use for the neuromag softwares)

I am using a newer python/mne environment/distribution to do the visualization.
mne ā€˜0.23.0ā€™
python 3.9
Distributor ID: Ubuntu
Description: Ubuntu 18.04.5 LTS
Release: 18.04

As I was poking around in the documentation, I found some instructions for using MRIlab for the coregistration, but this also used freesurfer for the MRI processing (not dicom access). I seem to recall some difference between freesurfer internal units and other MRI processing ones, but I canā€™t find that information now that Iā€™m writing up this topic.

We agree that having the team use the Freesurfer generated MRI images would likely relieve this situation, but they have already done about 45 of these coregistrations, so thereā€™s some motivation not to have to redo them.

Thanks for any thoughts,
Megan

2

hi @mschendel

you wrote a trans file with MRILab ?
the trans is the ā€œhead to MRIā€ transformation so to work the MRI coordinate
should be the same as with freesurfer. If itā€™s not the case there is probably a
not so hard fix. I am ok to help you debugging but I need to replicate the problem.
Maybe using the sample dataset?

Alex

3

Thanks for your suggestion. I think the issue is with dicom access, so let me take a look at the sample dataset and see if I can use dicom access to generate the initial fiff file.
Megan

4

Hi again,
I looked at the COR.fif file in the sample data, as that is what we typically would load into MRI lab, but the file in the sample data is already stripped to brain so I canā€™t use it to align the landmarks.
I also think the problem is creating the fiff file from dicom access, and testing this would require access to the mprage files, which I couldnā€™t find in the sample data.

Any advice?
Thanks,
Megan

5

can you join the next MNE office jours meeting on friday ?

A

6

Hi Alexandre,
Yes, I think I can.
Thanks,
Megan

7

Hi there,
It just doesnā€™t seem worth it to try and jam this old process into our new pipeline. Things work fine with using freesurfer and the mne process.
Thanks for the help, though!
Megan