Consistency of labels for dynamic parcellation

[kaitj]

The problem

Something that came up, but one thought was whether the indices of the merged labels would be consistent if the parcellations are dynamically created (e.g. thalamus segmentation per subject). The issue arises if different subjects have a different number of parcellations for whatever reason, leading to inconsistent labeling across subjects.

Currently, the the labels are merged in an incremental order based on what is present in the base and overlay images (correct me if I'm wrong here @tkkuehn. Need to give this some more thought, but one way could possibly to leverage the metadata to determine label indices (e.g. new overlay index = old overlay index + max(base overlay index)).

Environment

N/A

[kaitj]

I think this is now resolved by #36? @tkkuehn

[tkkuehn]

Right now (i.e. since #36 was merged), as long as the metadata is consistent across all subjects, the output labels will also be consistent. Notably if the metadata is different across subjects, this is still an issue.

[kaitj]

Ahh that is fair - that was not one I had considered when writing this issue. I can't think of any situation where the metadata would be different across subjects (at the moment) within a study.

We can put this on the agenda for next week to talk about?