Perform alignment in original image space

[arjunsridhar12345]

Hello, thanks for making this tool. It has a lot of helpful features to perform the alignment. An extension of this we have thought about is being able to perform the alignment in the original volume space. For now, we have defined a CustomAtlas class that subclasses BrainAtlas, and passing in the path to the original volume.

Running it with this modification leads to errors in getting the brain surface, so I think there is some steps needed regarding the xyz picks passed in (tried passing in non-transformed xyz picks but unsure of ordering of coordinates, etc.) I have a few questions below regarding this desire:

• Is this even possible given the way the current alignment code is structured?
• Has anyone already done something like this?
• Modifications needed to the app? This will probably require a deeper discussion. Can post the xyz json and image volume and labels if that's helpful.

[mayofaulkner]

Hello! Thanks for your message.

This would be the first time that anyone is trying this as far as I am aware. In theory I don't see why it shouldn't be possible.
The xyz picks should be given in the order ml ap dv in um with respect to bregma in the atlas.

Happy to take a look if you can provide me with the files!

[arjunsridhar12345]

Thanks for the response. When you say with respect to bregma in the atlas, does this mean the xyz picks need some transformation I assume? The xyz picks I have are done in the raw image space, and the volume/labels have been warped back from the ccf to the individual brain space.

I've attached the xyz picks (in order ml ap dv). I can't attach the volumes due to the file type not supported and when I try to zip them it is too big. The volumes are 25um spacing with shape (634, 406, 225) and origin (-1.5, -1.5, 1.5) in (ap, ml, dv) order if that's helpful. Thanks and happy to provide anything else to help out!

xyz_picks.json