Error when adding parcellation under non-linear normalization

Discussions
1

Dear all,

I'm using Brainstorm for SEEG electrode localization. I have raw T1w, CT, and SEEG data as input. However, I encountered some normalization-related errors when adding parcellation.

The localization procedure is as follows:

  1. Import MRI and do CAT12 segmentation (it seems that normalization is automatically done during this process)
  2. Import CT and co-register it to MRI using ct2mrireg
  3. Locate all electrodes

Here is the result tree and the parcellation option (I choose AAL3 as an example here):

image
image1323×1033 65.2 KB

I get the following error when the MNI normalization was non-linear (SPM or CAT12) but everything goes well when it's linear (maff8).

image
image972×578 24.9 KB

I'm not sure if this issue is due to an error in my procedure, the data, or if it's a bug. Please let me know if you need more information.

Thank you!

Best Regards,
Kun

2

@chinmay.chinara can you check this

3

@const would it be possible to share the data to check for the root cause. I tried with a bunch of our data but was not able to reproduce it.

also which version of Brainstorm are you using ?

4

@const, as you have noticed, the issue is the type or normalization.

By using non-linear normalization: there is a transformation for each voxel in the individual brain to the MNI space, and that transformation is independent of the other voxels transformations.

https://neuroimage.usc.edu/brainstorm/CoordinateSystems#Non-linear_normalization

The error that you get, happens as the MNI atlas (AAL3 in this case) has (MNI) voxels for which there is not a transformation subject space <--> MNI space. Thus it is not possible to find what is the subject voxel for those MNI voxels. When using a linear normalization method it is possible, as the transformation is the same for all the voxels.

Check this other post:

5

@chinmay.chinara Sorry, I may not allowed to share it due to privacy reasons since it's clinic data. But I found another post that followed the official tutorial and got the same error here (Error trying to use AAL3), which might be helpful.

And here are some environment-related information:

=== Brainstorm === 
  Version:        3.240808
  Release:        240808
  Variant:        souce
  Plugins:        cat12 ct2mrireg fieldtrip* spm12*
  SPM version:    SPM12 (7771)
  CAT12 version:  2577 (read from Brainstorm)
  BrainSuite version: BrainSuite23a (build #4199)
  ct2mrireg version: 14-Aug-2024 (read from Brainstorm)
 
=== Brainstorm directories === 
*** Directory paths may contain sensitive information, check before sharing *** 
  Brainstorm :    D:\OneDrive\Documents\MATLAB\brainstorm3
  DataBase   :    D:\OneDrive\working_data\sEEG_data\brainstorm
  Bst_User   :    C:\Users\kun\.brainstorm
  Temporary  :    C:\Users\kun\.brainstorm\tmp
 
=== Matlab === 
  Matlab version: R2021b (9.11)
  Java version:   1.8
 
=== System === 
  OS name:        Microsoft Windows 11 Pro (10.0.22631 N/A Build 22631)
  OS type:        win64
  Mem total:      97942 MiB
  Mem avail:      78301 MiB

@Raymundo.Cassani Thanks for your explanation. I still don't quite understand why some voxels fail to find a corresponding mapping, and whether this is a common issue.

Additionally, I would like to ask if the linear/non-linear normalization is solely influenced by the normalization operation in brainstorm? (Because this issue has never occurred with the previous ten or so subjects, where I used direct SPM register & reslice for co-registration. This is the first time I've tried using ct2mrireg for co-registration, so I'm not sure whether it's induced by my procedure or data or something else).

Finally, are there any other solutions besides switching to linear transformation?

Thank you again for your kind and quick reply!