Clarification on general SEEG localization procedure

Discussions
1

Dear all,

I’m a newbie in this field, and I'm using Brainstorm for SEEG electrode localization. However, I’m a bit unclear about the underlying processes in Brainstorm’s operations.

My current workflow is as follows:

  1. Import MRI and perform CAT12 segmentation.
  2. Import CT and co-register it to MRI using ct2mrireg (or SPM).
  3. Load SEEG channels and manually localize all electrodes.

Initially, I thought these steps corresponded to the following (please correct me if I'm mistaken):

  1. MRI normalization to MNI space.
  2. CT/MRI coregistration (which implies the CT is also in MNI space now?).
  3. Manual electrode localization (resulting in the contact positions being in MNI space?).

I have some questions for clarification:

  1. Which operations actually modify the data (e.g., reslicing) versus simply calculating transformations (like affine calculations without applying the transformation)?
  2. What exactly does CAT12 segmentation contribute in this context? Does it improve MRI normalization or transform MNI-based parcellation coordinates back to native space?

Thank you for your help!
Kun

2

Very interesting questions, let me try to clarify them:

  1. CAT12 does non-linear MNI normalization, it means that there is a transformation saved alongside the MRI volume to go from CT voxels to MNI space. Moreover, other processing in the volume (voxel) and surface spaces are performed, check some of these at CAT12 page:
    CAT12 Manual

  2. The CT is MNI normalized. (there is a transformation CT <--> MNI space)

  3. In Brainstorm, all coordinates are in the Subject Coordinate System
    https://neuroimage.usc.edu/brainstorm/CoordinateSystems
    However, the MRI and CT volume have all the information to go from SCS to MNI space. In addition, you can export the SEEG positions in MNI coordinates.

  1. To start, when importing a volume in Brainstorm, a copy of the volume is saved in Brainstorm and that is the one modified, your original volume files are never modified.
    Slicing will certainly modify the data, to have the same volume grid as the reference MRI.
    Moreover, co-registration methods can as well modify the volume data. E.g. CT2MRI, rotates the volume to match the MRI, and the CT volume is cropped to match the MRI space. On the other hand, the SPM method does not rotate nor crops the CT data.

  2. In your pipeline CAT12 will compute a (non-linear) MNI normalization, will automatically identify as well the fiducials that are used to generate the SCS, extract surfaces that may be useful when analysing the recordings, provide also anatomical atlases (volume labels) which can be used to label SEEG contacts:
    https://neuroimage.usc.edu/brainstorm/Tutorials/ECoG#Anatomical_labelling

3

Thanks for your insightful reply!

So, the MRI is directly normalized to MNI space using CAT12. The CT is first co-registered with the MRI, and then the MRI normalization parameters are applied to align the CT with MNI space. Is that correct?

If that's the case, how is the CT normalized to MNI space when only using SPM?

Is it possible to export all SCS coordinates and the transformation/normalization parameters to calculate the MNI coordinates outside Brainstorm?

I recently changed my referencing from average to bipolar, so I want to calculate the new channel coordinates in SCS and then convert them to MNI. However, modifying my current analysis pipeline to perform bipolar referencing directly in Brainstorm would be quite cumbersome. So, I’m wondering if it’s possible to export the data from Brainstorm and use other tools, like Python, to perform the calculations.

Thank you!

4

Yes that is correct. Just a detail, the MRI normalization parameters (MRI <--> MNI) are not applied directly to the CT, but are used to compute the CT<-->MNI parameters. The are not the same as the MRI and CT are coregistered in world space, not in the MRI (or CT) volume space.

When coregistering with SPM, there is not MNI normalization of the CT.
The MRI and CT will share the world space.

If the MRI was MNI-normalized before this transformation is used for the CT.

Yes, there transformation are stored in the volume files (MRI and CT) in Brainstorm, in the fields:

  • SCS: Defines the Subject Coordinate System, and transformation MRI <--> SCS
  • NCS: Defines the MNI coordinate system, and transformation MRI <--> MNI

Moreover, there is the function cs_convert() that allows you to convert between coordinate systems:
https://neuroimage.usc.edu/brainstorm/CoordinateSystems#Converting_between_coordinate_systems

In Brainstorm once you have imported your (average ref) data, it is possible to create the bipolar montage, and then apply this montage to the data to generate a new recording, then you would need to update the coordinates (in any space) for these channels.

Check this links on creating montages, and a tutorial that uses montages (for EEG):

This seems more like a topic for a different thread.
Do not hesitate in creating a new post regarding montages if you need more info for this topic.

5

Thank you for your detailed response! I’ve learned a lot.

I have a small follow-up question regarding the SPM-based and CAT12-based pipelines. From this explanation, it seems the main difference between the two pipelines lies in the tool used for MRI normalization. The subsequent steps for CT to MRI alignment and normalization of the CT to standard space should be quite similar, right?

I believe the information and functions provided above regarding the transformation should be enough to solve my problem. If I have any further related questions, I’ll start a new thread.

Thanks again for your patience and thorough explanations!