Nick Tustison

> Specifically, it appears that the "individual timepoints" which make up the SST are being under-estimated relative to the subject-specific template itself: I'm not surprised. I don't believe the SST...

Sure, and yes to Phil's suggestion. I'm pretty sure it was @jeffduda who discovered this issue early on and led to its correction in the current longitudinal script. I believe...

[Yes.](https://github.com/ANTsX/ANTs/pull/502). Perhaps it was an email Jeff sent and not an issue.

I'm guessing it's a Python 3.9 issue.

Have you looked at this [example repository](https://github.com/stnava/chicken)?

You're right but you can convert those labeled images to .csv point set files using ``ImageMath ... LabelStats ...``, ensure that it works as expected, and then map it to...

I hadn't noticed that. Thanks for pointing that out. Two points should work, I think, but I wouldn't use it as a case for debugging.

In my suggestion above, my assumption was that you'd do a bit more digging on your end. The original chicken example works so the key question is "at what stage...

I don't know what "landmark disks" are but the input data structure of the point set metric is the same [for images](https://github.com/ANTsX/ANTs/blob/master/Utilities/itkLabeledPointSetFileReader.hxx#L510-L576) as it is for [.txt files](https://github.com/ANTsX/ANTs/blob/master/Utilities/itkLabeledPointSetFileReader.hxx#L167-L197).

I believe I wrote it such that one could specify a mask to provide the necessary physical domain without having to use an image-specific metric. Also, you might want to...