Hello IIT forum,

thanks to the developers for providing this great atlas. When trying to use it with DTI-TK and the JHU atlas I've encountered the following challenge:

I want to use tensor based registration techniques by DTI-TK and register subject images to the IIT Atlas. Then I want to perform tractography on the IIT Atlas with the provided FOD data. Afterwards I would like to extract fiber bundles through JHU ROIs.

My understanding at the moment is as follows:

- DTI-TK requires the 256 voxel version of IIT. In this version, the origin is set to 0 and the orientation is LPI.
- The JHU atlas is, like the IITAtlas, in ICBM152 space. The transform of the non 256 voxel versions of the IITAtlas is:

1 0 0 -91
-0 1 0 -126
-0 0 1 -72

Therefore, I want to linearly transform the JHU atlas by this transformation (and probably also flip the x-axis ?) in order to be in the DTI-TK compatible space. Unfortunately neither just a linear transform (or the inverse) or a transform and a flip on the x axis has yielded good results.

What am I doing wrong?

Thanks for your help,
Darius

Hi Darius,

First, a few notes that may or may not apply to your project.

Note 1: I am not sure how committed you are to using JHU regions. If you want to conduct ROI analysis in white matter, you may want to use connectivity-based ROIs instead of the anatomically-defined JHU ROIs that may contain very different connections. The track-density images corresponding to the different edges of the connectome can be converted to masks (one mask per edge), transformed to the space of individual subjects, and used to extract average values of different MRI properties per edge.

Note 2: You mentioned you want to perform tractography on the IIT atlas. Note that the tractogram is already provided. Of course you can redo tractography on the HARDI template using your preferred approach and parameters.

Now to your questions:
The space of the IIT atlas and the ICBM152 space do not perfectly match. You should register one space to the other. If you do that, can't you apply the resulting transformation to the JHU regions to bring them to IIT 256x256 space?
Also, how did you apply the linear transformation?
Finally, if the above do not spark any ideas that lead to solutions, can you share a screen shot so that we get a feeling of the kinds of problems you have encountered?

Take care,
Konstantinos

Hey Konstantinos,

thanks for your comprehensive answer. I understand that I have to do a registration. Do I also have to flip the image in case my preferred atlas is in a different layout (not in 256 IIT LPI)? I've tried around with image transformations and header changes for over an hour now but I'm still a bit confused whether this is just something memory related or actual alteration of the interpreted data.

Thanks for your help
Darius

Hi Darius,

I am afraid I don't understand which step you can't get to work.
Are you trying to bring the JHU labels in IIT space? And how exactly are you trying to do that?
Please provide more details about your steps.

K.