axondeepseg / ometiff-tests Goto Github PK

I tested the notebook on Ubuntu 20.04 in WSL 1:

• I used the conda install from pyvips (https://libvips.github.io/pyvips/README.html#conda-install). It already includes the libvips library so I commented those lines in the notebook.
• I had to pip install matplotlib, jupyter and imagecodecs as well.

I have a few questions/observations about the tests.

1) Read OME-TIFF image + metadata:

• I got this message when running the cell: OME series: failed to read 'tubhiswt_C1.ome.tif', which is normal because the file C1 is not in the repo.
• However, in the displayed metadata, there is some metadata for C1, ex: {'UUID': {'FileName': 'tubhiswt_C1.ome.tif', 'value': 'urn:uuid:f801ea0a-e93e-4f4f-99b3-7ecc15048c12'}, 'FirstC': 1, 'FirstT': 2, 'FirstZ': 0, 'IFD': 2, 'PlaneCount': 1}
• And the display of "timepoints" gives me the following image, with the second row being empty:
  
• It looks like the 2 files are "linked" somehow through metadata in OMETIFF? (I suspect Etienne had the two files on his drive, but only 1 is in the repo). If that is the case, this is an OMETIFF feature that I was not aware of...

Write OME-TIFF image + metadata:

• If I understand correctly, I think the file should have 10 "timepoints" and 3 "channels". However, when I read the OMETIFF file written by tifffile in Fiji/Bioformats, I get an image and metadata with 30 "timepoints" (SizeT = 30) and 1 "channel" (SizeC = 1).
• Also, we should document what compress=7 means in the call of tifffile.imwrite.

3.1) TIFF to OME-TIFF:

• Why do we need to add an alpha channel? This should be documented.
• When I read the OMETIFF file written by pyvids in Fiji/Bioformats, I got the right dimension in metadata but the image is a repetition of a tiles?
  
• There is also a warning in the console, I don't know if it's related:

[WARN] SamplesPerPixel mismatch: OME=-1, TIFF=1
[WARN] Image ID 'Image:0': missing plane #0
[WARN] Using TiffReader to determine the number of planes.

3.2.2) Generate the OMETIFF with the NDPI data:

• For some unknown reason, when I open the generated OMETIFF file in Fiji/Bioformats, the image on the 3 channels appears totally black, although the validation at point 3.2.3 shows the correct image. I updated the Bioformats version to the latest i.e. 6.6.1.