[DAR-4940][External] - Remove the need to pass the legacy flag when importing and converting NifTI annotations

[JBWilkie]

Problem

Currently, medical files in the platform may be divided into two groups:

• 1: Files that were isotropically scaled upon upload - These are all files uploaded with the MED_2D_VEWER FF disabled
• 2: Files that were not isotropically scaled upon upload - These are all files uploaded with the MED_2D_VIEWER FF enabled

Now that MED_2D_VIEWER is globally enabled, all further uploaded files will not be isotropically scaled

When importing NifTI annotations to these items, or when converting their annotations to the NifTI format, we need to apply exactly the same scaling as was applied to the item itself upon upload. Currently, we control this with the legacy flag, but it's very manual and can be tricky for users to manage

Solution

This PR removes the need to pass the legacy flag in both the import and convert flows by:

• Importing annotations: Using the API to lookup the state of each file targeted by the import. If it was scaled, we apply scaling to the imported annotations, and vice versa
• Converting annotations: We use the presence of a newly added Darwin JSON medical metadata field "handler": "MONAI" to determine if a file was scaled or not, and apply scaling during the conversion if required

After this is merged, we'll release a new darwin-py version and update backend with it. This will enable backend to automatically scale exported NifTI annotations per-file, and then we'll fully remove the legacy flag and all related code some weeks in the future

Changelog

Removed the need to pass a flag when importing or converting NifTI annotations. Instead, we handle it automatically, significantly reducing confusion and the likelihood of errors

[linear]

DAR-4940 Remove the need to pass the `legacy` flag when importing & converting `NifTI` annotations

[dorfmanrobert]

I think we also need to check if is_volume is true

and in is_volume=false cases, we should apply no scaling at all right?

[JBWilkie]

Notes from our in-person conversation - These are the scenarios:

Before MED_2D_VIEWER:

• Not MONAI: Files are scaled --> It's possible to upload unscaled files, but these are files uploaded as slices (not volumes), which NifTI annotations cannot be exported from

After MED_2D_VIEWER:

• MONAI: Files are guaranteed to be unscaled
• Not MONAI: Files are scaled --> It's possible to upload unscaled files, but these are files uploaded as slices (not volumes), which NifTI annotations cannot be exported from