Segment SC using contrast agnostic model #44
Conversation
…tible with soft segmentations)
… light CSF to improve SC seg
| @@ -0,0 +1,352 @@ | |||
| """ | |||
There was a problem hiding this comment.
Are there specific differences between this script and https://github.com/sct-pipeline/contrast-agnostic-softseg-spinalcord/blob/4694169352fb962f607e7684fe778677adc123dc/monai/run_inference_single_image.py to justify duplication?
There was a problem hiding this comment.
Scripts are the same. There is a note about that at the end of this README. I duplicated the script because I did not want to create dependency on another repo's branch. I am aware that this duplication is suboptimal. We could move the run_inference_single_image.py script to the ivadomed/utilities repo to avoid further duplications. Tagging @naga-karthik.
There was a problem hiding this comment.
i think the plan is to have it inside SCT, right @naga-karthik? The faster we get it inside SCT, the faster we can remove the duplicates and all work on the SCT-based implementation
There was a problem hiding this comment.
Yes, the plan is to get it inside SCT. That is my action task for this week!
| @@ -240,6 +240,36 @@ def keep_largest_object(predictions): | |||
| return predictions | |||
|
|
|||
|
|
|||
| # Adapted from ivadomed: | |||
| # https://github.com/ivadomed/ivadomed/blob/e101ebea632683d67deab3c50dd6b372207de2a9/ivadomed/postprocessing.py#L224-L245 | |||
| def remove_small_objects(data, size_min): | |||
There was a problem hiding this comment.
i suggest working with atomic PRs-- ie: first, merge this one, and then create a new PR where you implement the remove_small functionality
…t_label_vertebrae as well as sct_label_utils (to avoid SC straightening)
|
The |
This PR adds a
segment_sc_contrast-agnostic.shbash script.The script:
run_inference_single_image.pyscript is just a copy of the contrast-agnostic script, see this comment for details)sct_label_vertebrae