Overall atlas segmentation pipeline in Python#69
Overall atlas segmentation pipeline in Python#69dzenanz merged 7 commits intoKitwareMedical:masterfrom
Conversation
As ApplyToImageMetadata is not available in Python and might never be, remove the commented-out code which mentions it. See this commit for an attempt to make it available in Python: dzenanz/ITK@db8157e
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
tbirdso
left a comment
tbirdso
left a comment
There was a problem hiding this comment.
Code is looking good overall. A few questions in regards to the vision for this code and how/whether it will need to be generalized for arbitrary sample sets.
| # we don't need to change laterality of atlas landmarks | ||
| # as they all lie in a plane with K coordinate of zero |
There was a problem hiding this comment.
Not sure I fully understand the geometry here. Will this generalize to other sample sets?
There was a problem hiding this comment.
It won't. But I guess I could (and should) apply some permutation matrix transform on the atlases (and some landmarks files) so the mirroring here is along LR axis instead of SI. Thanks for reminding me of this.
| # we don't need to change laterality of atlas landmarks | ||
| # as they all lie in a plane with K coordinate of zero | ||
|
|
||
| atlas_bone_label_filename = root_dir + bone + '/' + atlas + '-AA-' + bone + '-label.nrrd' |
There was a problem hiding this comment.
Is this filename convention documented and enforced elsewhere in the project as well?
There was a problem hiding this comment.
It is probably not documented yet. This is just where I have put files on my computer. I guess we should document it in the future, or make it configurable.
There was a problem hiding this comment.
Overall, we should move away from file IO, file name conventions, to in-memory data structure input output, dagster.
| label_names = { | ||
| 1: 'Diaphysis', | ||
| 2: 'Metaphysis', | ||
| 3: 'Growth Plate', | ||
| 4: 'Epiphysis', | ||
| } | ||
| # TODO: add proper reading of labels from .seg.nrrd files so this is not hardcoded | ||
| # label_names = { | ||
| # 1: 'Cortical', | ||
| # 2: 'Trabecular VOI', | ||
| # 3: 'New Trabecular VOI', | ||
| # 4: 'New Cortical VOI', | ||
| # } |
There was a problem hiding this comment.
Is there a strategy for revisiting this later? Possibly accepting an arbitrary list of label names?
There was a problem hiding this comment.
I was hoping to be able to read label names from .seg.nrrd files, which would make this hard-coded list unnecessary. That would also involve less manual steps, and therefore be less error-prone. I was planning to do this for the full set of atlases (6 per bone).
|
|
||
| if __name__ == '__main__': | ||
| if len(sys.argv) == 1: # direct invocation | ||
| atlas_list = ['901-L', '901-R', '907-L', '907-R', '917-L', '917-R', 'F9-3wk-02-L', 'F9-3wk-02-R'] |
There was a problem hiding this comment.
Perhaps not necessary to change for this PR, but would be good to clarify what steps will be needed to generalize pipeline for arbitrary samples. We probably would not want a developer to have to import these lines even if they never run __main__ directly.
There was a problem hiding this comment.
Currently, this is the entry-point for the analysis. We plan to use https://github.com/dagster-io/dagster later, so this will have to change too.
thewtex
left a comment
thewtex
left a comment
There was a problem hiding this comment.
Good progress, more todo to make it more general and move away from file IO.
Before merging, please rename the file from segmentation_pipeline.py to something more specific, e.g. mouse_femur_ct_morphometry.py
Handles laterality: Atlas and case under consideration can have different laterality (e.g. atlas is from right leg, case under consideration from left leg), so we appropriately mirror the atlas. Crop atlas to the bounding box of the labeled region (bone in question) Crop case to the bounding box of the bone in question
Without this, elastix crashes on the second iteration of the loop `for case in data_list`.
Compress bone atlas As atlases are already axis-aligned, we don't need its landmarks
3 participants
No description provided.