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.

Sign up for GitHub

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

Jump to bottom

Update README.md #1

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Update README.md #1

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 40 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,2 +1,40 @@
# AtlasPack
Combined cortical/subcortical atlases for xcp-d and qsiprep
# Combining cortical and subcortical atlases for use in XCPD and QSIPrep

There are many popular brain atlases, but they're all in different versions of MNI space. They also cover
different structures. We want to get them all into one MNI version correctly - using the templateflow
framework - so they can be used to make structural and functional connectivity matrices.

## Things to worry about

We ultimately want square connectivity matrices with exactly the same regions represented in the rows/columns to come out of xcpd
and qsiprep. Here are some of the problems we will run into:

1. Some regions are so small that when they're mapped to a low-res BOLD/DWI volume they will disappear. Should we allow this?
In some cases they are from dividing a region into subcomponents. We could recombine them.
Comment on lines +12 to +13
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder what nilearn's behavior is? I assume it just returns NaNs for the missing regions. I'd be fine with NaNs if the region disappears or is entirely outside the brain mask, generally speaking.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm mostly worried about what this means for those interpreting the connectivity matrices without thinking too hard about it. A column of NaNs isn't a big deal software-wise, but if someone is doing network science stuff on these matrices it's a big deal to have a fully disconnected node. We also know that there should almost never be an actually disconnected node in the brains we'll be analyzing. We looked at this problem here and it still haunts me

2. Regions may overlap each other from different atlases. Which should have precedence?
Copy link
Member

@tsalo tsalo Nov 30, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I remember correctly, qsiprep needs everything to be in the same atlas file, right? Because xcpd can just concatenate time series files if necessary, I think. Just saw the next bullet point 😆

If there are overlapping regions, we could use a 4D atlas file, and basically pretend it's a probabilistic segmentation.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Mrtrix, DSI Studio, etc need a single 3D volume :(

3. For DWI the regions need to be defined *all in the same image*. BOLD regions can be in different images and you can calculate
connectivity from their timeseries later. But this isn't possible for DWI. Therefore, we can't include Tian and dwi/CIT168 atlases
in the same image
Comment on lines +15 to +17
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can the atlas image be 4D?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unfortunately not for the DWI usage


## Strategy to for making the cross-software atlases

Due to worry #3 we are going to pick subcortical atlases that DO NOT have multiple resolutions. Since extremely popular cortical
atlases have multiple resolutions (i.e. Schaefer) multiple subcortical atlases would explode the number of combined atlases.
For example you would need cortical100+subcortical100, cortical100+subcortical200 ... cortical1000+subcortical1000.

### QSIPrep atlases

| Atlas Name | MNI Version | Subcortical | Cerebellum | Thick |
| :------------: | :------------------------: | :---------: | :--------: | :-----: |
| AAL116 | NLin6 | Yes | Yes | Yes |
| Aicha384 | NLin6 | Yes | No | Yes |
| Brainnetome246 | NLin6 | Yes | No | No |
| Gordon333 | NLin6 | No | No | NOO |
| power264 | NLin6 | Yes | Yes | Spheres |
| Schaefer | NLin6 | No | No | No |

### XCPD atlases

Should go here