Originally posted by @jcohenadad in spinalcordtoolbox/spinalcordtoolbox#4042 (review)
We should choose another output suffix for the inference. Currently it says: _lesion-manualNeuroPoly.nii.gz
, which is absolutely not intuitive (ie: it is not a manual segmentation)
Originally posted by @joshuacwnewton in spinalcordtoolbox/spinalcordtoolbox#4042 (comment)
Good observation!
Though, this seems to be coming from the trained models themselves. Looking at model_seg_lesion_mp2rage.json
, which is provided via the test release r20230210
, we see:
"target_suffix": [
"_lesion-manualNeuroPoly"
],