search

ivadomed / ms-lesion-agnostic

Deep Learning contrasts "agnostic" tool for MS lesion segmentation in the spinal cord
MIT License
3 stars 0 forks source link

Manual segmentation of MS lesions in the spinal cord to be reviewed by neuroradiologist #4

Open plbenveniste opened 4 months ago

plbenveniste commented 4 months ago

Creating this issue to centralize all images which are tricky to annotate and which could be discussed with a neurologist.

Tagging @Nilser3

plbenveniste commented 4 months ago

This is image from CanProCo is tricky : sub-tor016_ses-M0_PSIR.nii.gz

ezgif com-animated-gif-maker

Should the spinal cord central canal be segmented or not ? How many lesions should be counted ? Also the M12 image has a lot more lesions, did we miss some ?

[!NOTE] Code to get image : fsleyes sub-tor016/ses-M0/anat/sub-tor016_ses-M0_PSIR.nii.gz sub-tor016/ses-M0/anat/sub-tor016_ses-M0_T2w.nii.gz derivatives/labels/sub-tor016/ses-M0/anat/sub-tor016_ses-M0_PSIR_lesion-manual.nii.gz & Code to get M12 image : fsleyes sub-tor016/ses-M12/anat/sub-tor016_ses-M12_PSIR.nii.gz sub-tor016/ses-M12/anat/sub-tor016_ses-M12_T2w.nii.gz derivatives/labels/sub-tor016/ses-M12/anat/sub-tor016_ses-M12_PSIR_lesion-manual.nii.gz &

plbenveniste commented 4 months ago

This case is not that easy as well to know if it's one big lesion or multiple lesions : Canproco - sub-tor029_ses-M0_PSIR.nii.gz

ezgif com-animated-gif-maker (1)

[!Note] Code to get image : fsleyes sub-tor029/ses-M0/anat/sub-tor029_ses-M0_PSIR.nii.gz sub-tor029/ses-M0/anat/sub-tor029_ses-M0_T2w.nii.gz derivatives/labels/sub-tor029/ses-M0/anat/sub-tor029_ses-M0_PSIR_lesion-manual.nii.gz &

plbenveniste commented 3 months ago

In this case, lesions have been annotated with multiple values (1 and 2). One theory is that the annotation comes from the level of confidence of the lesion annotation. Subject : sct-testing-large - sub-rennesMS002_acq-sagcerv_T2w.nii.gz

ezgif com-animated-gif-maker (2)

The lesions in yellow could be focal lesions ? Dirty white matter ?

[!Note] Code to get image : fsleyes sub-rennesMS002/anat/sub-rennesMS002_acq-sagcerv_T2w.nii.gz sub-rennesMS002/anat/sub-rennesMS002_acq-inf_T2star.nii.gz derivatives/labels/sub-rennesMS002/anat/sub-rennesMS002_acq-sagcerv_T2w_lesion-manual.nii.gz &