CrossLong Data

[ageoly-git]

Hi Nick,

I hope this message finds you well.

I am reaching out because I was interested in integrating the cortical thickness data you used in the NeuroImage paper into a longitudinal cortical thickness analysis I am performing on veteran TBI patients that relies on ANTs SST (labeled with DKT 31 atlas).

My hope is to use the region-wide baseline data from the "CN" group as a healthy control sample and my goal is to evaluate an individual's trajectory to/away from a normative age-matched profile. However, when examining the ANTs SST data you have in the "Data" branch, I noticed that all of the participants included are > 60 years old (my data set ranges from 30 y/o to 60 y/o). In the NeuroImage paper I see that Figure 8. did the Age prediction with participants ranging from ~10 y/o up to ~80 y/o.

Is there any chance you might be able to make the full data set available? Additionally, if I am blindly missing something perhaps you could point me in the right direction?

Additionally, I wanted to nerd out and say thank you so much for your contributions to NeuroImaging and for the wonderful ANTs toolbox. It is something I use frequently and your work has truly guided my passion for neuroimaging !

Warmest regards, Andrew

[ntustison]

Hi Andrew,

Thanks for your interest and the kind words.

So there's a couple of papers which you might be mixing up. This CrossLong repository corresponds to a 2019 JAD paper which, as the title indicates, uses ADNI data which skews older so it's no surprising that the ages are all > 60. I have an earlier paper that was published in NeuroImage in 2014. That repository is here. The results from the four public data sets (Oasis, NKI, Kirby, IXI) are in the analytics2 subdirectory in .csv files and are probably what you're looking for. If you're interested in those, let me know and I can help you find what you need. Admittedly, this repository isn't as clean as I generally tend to make them. Finally, the most recent cortical thickness paper came out in 2021 with the repository here. Basically we converted much of the cross-sectional and longitudinal pipelines to employ deep learning resulting in better measurements. It includes the additional SRPB data set and the data is much better organized. My guess is this latter data set is what you'd be most interested in.

Just an FYI but the deep learning work is all found in our ANTsXNet Python and R packages. An extended tutorial is available here.

Good luck and let me know if you have any questions.

Nick

[ageoly-git]

Thank you so much for your help and quick response, Nick.

I just found the data from the 2021 thickness paper, and it is neatly organized and easy to parse.

I am really excited to get started with this data to see what comes out!

Best, Andrew

---

From: Nick Tustison @.> Sent: Wednesday, January 17, 2024 3:37 PM To: ntustison/CrossLong @.> Cc: Andrew Dedinas Geoly @.>; Author @.> Subject: Re: [ntustison/CrossLong] CrossLong Data (Issue #2)

Hi Andrew,

Thanks for your interest and the kind words.

So there's a couple of papers which you might be mixing up. This CrossLong repository corresponds to a 2019 JAD paperhttps://pubmed.ncbi.nlm.nih.gov/31356207/ which, as the title indicates, uses ADNI data which skews older so it's no surprising that the ages are all > 60. I have an earlier paperhttps://pubmed.ncbi.nlm.nih.gov/24879923/ that was published in NeuroImage in 2014. That repository is herehttps://github.com/ntustison/KapowskiChronicles/. The results from the four public data sets (Oasis, NKI, Kirby, IXI) are in the analytics2 subdirectory in .csv files and are probably what you're looking for. If you're interested in those, let me know and I can help you find what you need. Admittedly, this repository isn't as clean as I generally tend to make them. Finally, the most recent cortical thickness paperhttps://www.nature.com/articles/s41598-021-87564-6 came out in 2021 with the repository herehttps://github.com/ntustison/PaperANTsX. Basically we converted much of the cross-sectional and longitudinal pipelines to employ deep learning resulting in better measurements. It includes the additional SRPB data set and the data is much better organized. My guess is this latter data set is what you'd be most interested in.

Just an FYI but the deep learning work is all found in our ANTsXNet Python and R packages. An extended tutorial is available herehttps://gist.github.com/ntustison/12a656a5fc2f6f9c4494c88dc09c5621.

Good luck and let me know if you have any questions.

Nick

— Reply to this email directly, view it on GitHubhttps://github.com/ntustison/CrossLong/issues/2#issuecomment-1897476072, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AQWCYCDNK3JOGKPX44CN2O3YPBOCFAVCNFSM6AAAAABB7IBUEKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQOJXGQ3TMMBXGI. You are receiving this because you authored the thread.Message ID: @.***>

[ageoly-git]

Hi Nick,

I hope all is well!

I wanted to reach out and say thank you very much for the incredibly helpful resources you pointed me towards a while back. The data from the 2021 Thickness paper have proven to be very useful for normative modeling.

More recently, our group has an interest in utilizing the "SubjectToTemplateLogJacobian.nii.gz" outputs from the antslongitudinalcorticalthickness pipeline to examine expansion/contraction for a given participant's time point image.

I was wondering if you might be able to confirm/clarify the following?

Please correct me if I am wrong, but my understanding is:

1) The log-Jacobians for a given time point in the single subject template space reflect intra-subject deformation relative to that SST.

2) For a given time point: if voxel > 0, then voxel expands towards the time point (i.e. time point voxel > SST); if voxel < 0, then voxel contracted towards the time point (i.e. time point voxel < SST).

I apologize if these are simple questions but I just want to ensure I am fully understanding the outputs before moving forward. I have reviewed some earlier documentation / issue threads and I think I am correct (but not totally sure).

Thank you so much for your continued support and I hope you have a wonderful weekend.

Best, Andrew

[ntustison]

Hi @ageoly-git ,

1. The log-Jacobians for a given time point in the single subject template space reflect intra-subject deformation relative to that SST.

2. For a given time point: if voxel > 0, then voxel expands towards the time point (i.e. time point voxel > SST); if voxel < 0, then voxel contracted towards the time point (i.e. time point voxel < SST).

It's not the SST but the group template. Otherwise, what you write is correct. You can see further information here.