Label-studio: Ability to have alternate versions of the same image to annotate with RoiPolygon

[p-sodmann]

Why

In biomedical images there are usually different image modalities of the same object. For example different weightings in MRI for the same slice or different staining / immunofluorescence channels with different proteins in the same specimen in microscopy.
Just having the possibility to make a ROI selection based on one modality limits the use of label-studio.

Conditional Labeling with choices provides an Opportunity but cannot share the annotation across different images.

Solution

One possibility would be to give Annotation objects the ability to attach toName to a list of objects
<PolygonLabels name="label" toName=["image_x", "image_y", "image_z"]>

Or Group the images together, since only one can be displayed at every time anyway

  <View visibleWhen="choice-selected"
        whenTagName="image_modality" whenChoiceValue="Vx">
    <Image name="image_x" group="image" value="$image_z" zoomControl="true" zoom="true"/>
  </View>
  <View visibleWhen="choice-selected"
        whenTagName="image_modality" whenChoiceValue="Vy">
    <Image name="image_y" group="image" value="$image_y" zoomControl="true" zoom="true"/>
  </View>

<PolygonLabels name="label" toName=image>

[makseq]

@theudas We have discussed this issue with the team, I hope we implement it soon.

[makseq]

@hlomzik Have you any news?

[makseq]

@p-sodmann Sorry for a late answer, but we have implemented a multi-layer image approach finally:
https://github.com/heartexlabs/label-studio-frontend/pull/164

I think we will include it into release 0.9.1.

Feel free to reopen this issue if you have any problems.