Correctly display fractional segmentations with TILED_FULL dimension organization and multiple segments

[CPBridge]

@igoroctaviano @fedorov @dclunie

Currently, the specific combination of segmentations with SegmentationType="FRACTIONAL", DimensionOrganizationType="TILED_FULL" and multiple segments do not display correctly in Slim.

I tested the following combinations locally with dcm4chee:

SegmentationType=BINARY, DimensionOrganizationType=TILED_FULL, multiple segments => displayed correctly
SegmentationType=FRACTIONAL, DimensionOrganizationType=TILED_FULL, single segment => displayed correctly
SegmentationType=FRACTIONAL, DimensionOrganizationType=TILED_SPARSE, single segment => displayed correctly
SegmentationType=FRACTIONAL, DimensionOrganizationType=TILED_FULL, single segment => displayed correctly
SegmentationType=FRACTIONAL, DimensionOrganizationType=TILED_SPARSE, multiple segment => displayed correctly
SegmentationType=FRACTIONAL, DimensionOrganizationType=TILED_FULL, multiple segment => incorrectly displays as blank

While the others display correctly, in the last case the segmentation appears to load but displays all "empty" (i.e. purple). See https://github.com/ImagingDataCommons/IDC-ProjectManagement/issues/1811 and the screenshot therein.

The issue is likely related to the way slim/dmv interprets the arrangement of the tiles across the different segments for TILED_FULL. The relevant logic is described here. It should be noted that we recently put in a correction to this section of the standard to add the line that appears there now "then along the segments, if applicable, where the direction is defined by ascending numeric values of Segment Number (0062,0004) as defined in the Segment Sequence (0062,0002).". I.e. when slim/dmv was originally written, the way that multiple segments should be stacked within a TILED_FULL segmentation was not defined. It's perhaps more surprising that this does work for multiple-segment TILED_FULL binary segmentations, than that it doesn't for multiple-segment TILED_FULL fractional ones.

There is a bucket containing the problematic multiple-segment TILED_FULL fractional segmentations for images in the RMS cohort (within IDC) here that may be used for testing.

[CPBridge]

In contrast with my findings with dcm4chee it appears that FRACTIONAL, TILED_FULL segmentations can be displayed by slim on the google DICOM store. See this comment. I'm unable to explain this difference in behaviour between different archives