OMERO import of directory with `DICOMDIR` behaves unexpectedly

[melissalinkert]

A directory import of https://downloads.openmicroscopy.org/images/DICOM/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/ results in two Images, where only one is expected.

$ omero import -f PARNED-0BNNF4_B2_Q30/
...
2025-11-06 17:04:22,124 8453       [      main] INFO      ome.formats.importer.ImportCandidates - 8 file(s) parsed into 2 group(s) with 2 call(s) to setId in 7686ms. (7690ms total) [0 unknowns]
#======================================
# Group: /mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_1 SPW: false Reader: loci.formats.in.DicomReader
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_1
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_0
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_2
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_3
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_4
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_6
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_5
#======================================
# Group: /mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DICOMDIR SPW: false Reader: loci.formats.in.DicomReader
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DICOMDIR
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_0
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_2
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_3
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_4
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_1
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_6
/mnt/bioformats_test_data/automated-tests/curated/dicom/public/wsi/2023-04-28/PARNED-0BNNF4_B2_Q30/DCM_5

Only the second group of files is expected, since that is the minimum number of groups to cover all files in the directory. Suspect this is something to do with how DICOMDIR files are treated in DicomReader, but this requires more investigation.

[melissalinkert]

Looking more closely at ImportCandidates, I think this is the intended behavior (I was just surprised). Picking any file other than DICOMDIR intentionally does not detect the DICOMDIR file, which means that the DICOMDIR group has to exist. Since picking any other file will end up with multiple used files, the check in https://github.com/ome/omero-blitz/blob/master/src/main/java/ome/formats/importer/ImportCandidates.java#L624 does not result in the DCM_0 group being removed.

I don't think we want to update DicomReader to always look for a DICOMDIR, so it looks like the only other way to fix this would be a substantial change in ImportCandidates behavior to exclude any group whose entire used file list is contained within another group. Moving to the omero-blitz repo in case anyone else has additional thoughts, but I am OK with closing since I am now convinced this isn't a bug, just a misunderstanding. We'll need to keep this in mind for any future testing though.