thalamic segmentation out of date

The current segmentation approach https://freesurfer.net/fswiki/ThalamicNuclei seems to have been superceded by a couple other approaches:
https://freesurfer.net/fswiki/ThalamicNucleiDTI
including a Bayesian one and a CNN..

I am thinking it may be useful to swap in the CNN version (likely the most computationally efficient option)..

No implementation file, test, or entry point is named in the issue. Start by comparing the current ThalamicNuclei approach with the linked ThalamicNucleiDTI documentation and its Bayesian and CNN alternatives. Done would mean selecting an approach and integrating it into scattr with a way to verify the resulting segmentation.

Written by the indexing model from the issue text.