Skip to content

HT local subregions testing - #1

Open
htregidgo wants to merge 4 commits into
dti_integrationfrom
HT-local-subregions-testing
Open

HT local subregions testing#1
htregidgo wants to merge 4 commits into
dti_integrationfrom
HT-local-subregions-testing

Conversation

@htregidgo

Copy link
Copy Markdown
Collaborator
  • Add local subregions test config profiles
  • Document subregions smoke-test outputs

htregidgo added 3 commits May 14, 2026 14:56
Refactor the shared subregions smoke-test harness so Henry and Jackson can
run it without editing hard-coded paths in the script.  The script now looks
for repo-relative JSON profile pairs and uses the first one whose required
FreeSurfer, subject, segmentation, and DTI paths exist.

Add Henry-specific DTI and multichannel JSON files while leaving Jackson's
existing JSON files as the fallback profile.  Keep temp and output locations
separate for DTI, standard thalamus, and multichannel runs so results remain
inspectable and avoid filename collisions.

Replace the hard-coded means_groupings.json path in thalamusDTI.py with a
path derived from the module location, so the support file is loaded from the
repo on either machine.
Add notes to the local subregions test harness explaining the
three objects it drives: the DTI-aware run, the standard thalamus
run, and the multi-channel DTI/FA run.

Annotate each process.py-style command group with the temp files or
final outputs it writes. This keeps the dirty shared harness easier
to inspect after long local smoke tests without changing runtime
behavior.
@jnolan14
jnolan14 marked this pull request as ready for review May 27, 2026 14:22
Add shared Gaussian helpers for diagonal and full-covariance subregions
fitting, including likelihoods, posterior parameter updates, covariance
conversion for GEMS, and validation for covariance-mode inputs.

Wire covariance_mode through MeshModel and ThalamicNucleiDTI while preserving
diagonal mode as the default. Refactor image fitting and final posterior
extraction to use the shared helpers, and pass full covariance matrices to
GEMS in both covariance modes.

Update DTI hyperparameter estimation so multi-channel runs produce
channel-aware meanHyper arrays while keeping nHyper one-dimensional. Add
focused pytest coverage for the helper math, covariance shape handling,
low-mass fallbacks, singular covariance rejection, and the full mean-prior
cost.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant