Update ANTs tag?

[ntustison]

Hey all,

I'm trying to debug a subtle KK issue and tracing it to a discrepancy between ANTs and ANTsPy. I realized that the core ANTs used in ANTsPy is a bit dated:

% git show 276cf0717945d3dd3c4298c607d9d6a788ba574e
commit 276cf0717945d3dd3c4298c607d9d6a788ba574e
Merge: 6eb6bf05 1b2b5412
Author: Philip Cook <cookpa@pennmedicine.upenn.edu>
Date:   Tue Mar 21 22:20:45 2023 -0400

    Merge pull request #1516 from ANTsX/parseWinPointers

    ENH: Parse windows pointers for write operations

Anybody have any thoughts? Particularly, @cookpa and @ncullen93 who've been doing maintenance, containerization, etc.

[ncullen93]

Yes I think it's a good idea to update the ants tag and also increment the antspy version.

[cookpa]

Agree, we're due a new release and I think it would be good policy to bump the version when ANTs gets update, which often implies an ITK update also.

Could we increment to 0.5.0?

[ntustison]

Thanks @ncullen93 and @cookpa . I would have to defer to y'all. I imagine what I'm working on isn't as crucially tied to a specific ANTsPy version as what others are doing.

Specifically, @cookpa incrementing to 0.5.0 is fine with me. Would you like to do the updating or would you rather I did it? I don't think I've ever formally updated versions of any of the ANTsX packages.

[ncullen93]

Any good places to announce the new release? Neuroimaging forums, etc..

I'd also like to revamp the docs and move them to ants.dev/docs/antspy for a new release... I just haven't found a good tool yet to generate them in markdown.

[ntustison]

Yeah, promotion is something we've definitely struggled with.

[ntustison]

Okay, the KK issue I was having in ANTsPy where thickness values would occasionally blow up to ridiculous values does not exist in the KK version in ANTs. I don't know if this is with the DiReCT image filter itself or something internal ITK but it would be good to update just for this reason alone.

[stnava]

yeah - I have seen that as well .... thanks for looking into

[ncullen93]

If anyone wants to update it then feel free, otherwise I can do it tomorrow.

[cookpa]

Okay, the KK issue I was having in ANTsPy where thickness values would occasionally blow up to ridiculous values does not exist in the KK version in ANTs. I don't know if this is with the DiReCT image filter itself or something internal ITK but it would be good to update just for this reason alone.

I have a vague memory of seeing this and then thinking I fixed it with an explicit t=10 kwarg to the kelly_kapowski call, but I don't understand how that was needed because the default should have always been 10.

Anyway, I can tag a Github release today. I don't know if the last section of wheels.yml, sending things to PyPI, actually works though. But it should make wheels and attach them to the release.

[ntustison]

Yeah, I thought I remembered this issue, as well, but I thought it was much further back in the history then within the past year or so.

I'm actually working on a mouse-based analog to the cortical thickness pipeline and so the scale is obviously completely different but I haven't had any success in determining any sort of correspondence with any of the parameters. It was only when I tried to compare with the ANTs version that I noticed the differences in code although nothing in the diff output strikes me as being the culprit.

[ntustison]

Thanks @ncullen93 .