Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CIT168 node count #15

Open
andreifoldes opened this issue Sep 21, 2023 · 7 comments
Open

CIT168 node count #15

andreifoldes opened this issue Sep 21, 2023 · 7 comments
Labels
bug Something isn't working

Comments

@andreifoldes
Copy link

Summary

I realize its rude to ask question about "dev" features, but when opening the atlas-4S152Parcels_dseg.tsv I noticed that there are 28 (lateralized) nodes, but in the original CIT168 there are 20 (bilateral) nodes: https://neurovault.org/collections/3145/.

Is this because they get squished in MNI template space?

@andreifoldes andreifoldes added the bug Something isn't working label Sep 21, 2023
@tsalo
Copy link
Member

tsalo commented Sep 21, 2023

I believe so. I think @mattcieslak handled the initial resampling, so he could speak to that, but it makes sense that we would lose some regions when downsampling to 2 mm3 and warping to MNI152NLin6Asym.

@mattcieslak
Copy link
Contributor

In the AtlasPack repo you can see all the steps that were taken to make these atlases. I don't think, particularly in BOLD, that averaging signal from analogous regions in different hemispheres is what most people intend to do. This script is where the splitting and renaming happens.

There are also lines in that script where you can see that some tiny structures are merged together to prevent the downsample-disappearance that can happen when you go to 2 or 3 mm. The names in the sidecars tell you what the original CIT168 regions were and how they're been combined.

@andreifoldes
Copy link
Author

andreifoldes commented Sep 21, 2023

In the AtlasPack repo you can see all the steps that were taken to make these atlases. I don't think, particularly in BOLD, that averaging signal from analogous regions in different hemispheres is what most people intend to do. This script is where the splitting and renaming happens.

Right! I was trying to understand why there are 14 nodes instead of 20 shown in neurovault.org/collections/3145.

There are also lines in that script where you can see that some tiny structures are merged together to prevent the downsample-disappearance that can happen when you go to 2 or 3 mm. The names in the sidecars tell you what the original CIT168 regions were and how they're been combined.

Yes, thank you for linking that. I can see that node 10 and 11 are merged into node 7, which explains the difference. Would it be possible for XCP_D to consider the resolution of the provided functional files and if sufficient apply the "original" CIT168? If I understand correctly the original atlas was designed with 0.7mm anatomical and was later upsampled to 1mm in template space.
This would be very useful for using XCP_D with 7T files.

@tsalo
Copy link
Member

tsalo commented Sep 26, 2023

Would it be possible for XCP_D to consider the resolution of the provided functional files and if sufficient apply the "original" CIT168?

I don't believe the other atlases that compose 4S are available at a higher resolution.

@tsalo tsalo transferred this issue from PennLINC/xcp_d Sep 26, 2023
@andreifoldes
Copy link
Author

andreifoldes commented Sep 27, 2023

I see! However for the atlases like the Tian or the currently discussed CIT168 this "resolution-aware" functionality would be useful no?

@tsalo
Copy link
Member

tsalo commented Sep 27, 2023

In the case of individual atlases used by XCP-D, I agree that having the highest-resolution version as the "base atlas" is best. However, CIT168 is only used in XCP-D through the combined 4S atlas, so the overall atlas resolution is dependent on the lowest-resolution component atlas in that case.

@tsalo
Copy link
Member

tsalo commented Sep 27, 2023

If XCP-D isn't using the highest-resolution (NIfTI) version of any of its other atlases, please let me know. I think I tried to grab the best version of each, but I might have missed something.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

3 participants