MNI normalization: “MNI-CAT12” vs “MNI-Segment”

Discussions
, ,
1

Hi all,

I am using Brainstorm for SEEG electrode localization.
Following the tutorials, I first imported the subject’s MRI and ran CAT12 segmentation, then imported the post-implant CT and labeled the contact locations.

When I wanted to export the MNI coordinates and AAL labels of the contacts, I got a bit confused about the normalization step.
Some tutorials mention that I need to run MNI normalization before exporting, but after performing CAT12 segmentation, I already saw a coordinate system called “MNI-CAT12” in the MRI viewer. I am not sure if that means the MRI has already been normalized.

Later, I tried running MNI normalization again before exporting.
At that point, the options were only SPM / maff8, and after choosing SPM, the viewer label changed from “MNI-CAT12” to “MNI-Segment.”

So I would like to confirm:

  1. Does CAT12 segmentation already include an MNI normalization step?

  2. If I then run MNI normalization (SPM-Segment) afterward, does that mean I applied normalization twice, or is this the correct workflow?

Thank you very much for your clarification and for maintaining such great documentation!

Best regards,
Kun

2

Yes, CAT12 it does MNI normalization using a non-linear method.

You are replacing the MNI normalization that has been already computed by CAT12 with a new MNI normalization, this time performed with SPM. Note that both methods, the CAT12 and Segment are non-linear.

Check this post:

1 Like
3

Thanks for your quick and clear explanation!

Would it be possible to switch back from the SPM normalization to the CAT12 version without affecting other parts of the workflow? For example, by reapplying CAT12 segmentation on the MRI?

I’m not sure whether this would interfere with the electrode localization I’ve already done.

4

Not really, the CAT12 normalization is imported when you run the CAT12 segmentation. As such you would need to re-run CAT12 segmentation to have its MNI normalization.

The MNI coordinates for the electrodes will be modified as a different conversion from the SCS (subject coordinate system) to MNI will be used.

One thing you to re-compute the CAT12 MNI-normalization without changing your electrode localization would be to do the following:

Note: Before following the steps duplicate the Subject, so it is possible to go back if you are not satisfied with the results.

  1. Export the contact localizations as EEG: BIDS electrodes.tsv, subject space mm (*.tsv). Link

  2. Perform CAT12 segmentation, to re-compute the CAT12 MNI-normalization

  3. Import the contact locations using the same format as in Step1 link

1 Like
5

Well noted with many thanks!

6

As mentioned there, CAT12 segmentation already performs normalization automatically, so an additional normalization should not be necessary before running iEEG atlas labels. However, I often encounter the issue described here: https://neuroimage.usc.edu/forums/t/some-points-are-outside-the-definition-of-the-non-linear-mni-registration/53625

To resolve it, I usually apply an extra maff8 normalization after electrode localization, then add the parcellation and export the iEEG labels. Based on earlier explanations, this step should replace the existing CAT12 normalization.

However, when I normalize using SPM’s segment, this replacement is reflected in the MRI viewer (MNI fields) and in the exported labels.csv (the parcellation column gets the “segment” suffix). But when I apply maff8 normalization, nothing appears to change — both still show “cat12.” The only sign that maff8 normalization worked is that the “points outside” error disappears.

So I’d like to ask:

  1. In this case (CAT12 normalization already done, then followed by maff8 linear normalization), what exactly happens internally?
  2. Does Brainstorm treat the MRI as normalized only by maff8 at that point?
  3. When exporting labels.csv, are the MNI coordinates and parcellation labels derived from CAT12 or from the maff8 normalization?
  4. And what exactly does “Add parcellation” do? Why does it throw an error when some points fall outside the non-linear registration instead of just ignoring them?

Thanks again for your helpful and detailed reply!

7

The MRI will have information for both MNI normalizations, the non-linear (with CAT12) and the linear (with maff8).

It has both normalizations. For each conversion it will try to do it with the non-linear. If it fails (points outside the definition), it will use the linear normalization for only those points.

As per the previous point, it depends the points that failed for CAT12 non-linear normalization will be done with maff8.

It reslices the MNI parcellation into subject space to create the same parcellation for the subject MRI volume

1 Like
8

Thank you. That’s really helpful!