Converting MRXS to OME-TIFF

Hi All,

I’m hoping to get some insight/suggestions for using bioformats2raw and raw2ometiff to convert 3DHISTECH MRXS files to OME-TIFF to be imported into Omero. I am a Sysadmin working to integrate/automate some processes and have very limited background in these imaging formats. I’ve done some reading starting here and the github repos & issues for both tools and generally have the process working, that’s not really the problem. The concern I have is the drastic increase in size from the MRXS to OME-TIFF.

I’m seeing anywhere from a 4x to 6x increase in size after conversion using JPEG-2000 compression on the OME-TIFF (the default LZW is obviously much larger yet). At first I thought maybe it was a problem with the source images due to our scanner settings, so I tried converting some examples from here with similar results.

At this point I have stuck to what seems like the common CLI options for the tools, here’s some examples of what I’m using:

bioformats2raw --resolutions=6 CMU-2.mrxs CMU-2_raw
raw2ometiff --compression=JPEG-2000 CMU-2_raw CMU-2-JPEG-2000.ome.tiff

The above starts with ~1.3 GB at the source and results in ~7.1 GB OME-TIFF file. As our real world source images are anywhere from 10-30 GB usually, this is a concern.

Am I missing anything? Are there any other options that would help with the file size without noticeable quality loss? I’ve also seen this Github issue which is starting to lead me to believe this is expected and its not me doing something wrong.

Any insight is greatly appreciated, thanks!

Hi @brianc

We implemented the workflow and it works for us.
We did notice few problems, one of them is of image size. For us the number of pixels in the output image was increased.
We reported this issue and provided sample data. I suggest you inspect your input/output images, eg using QuPath to see if you have the same issue.

best
Ofra

Hi @brianc ,
The original CMU-2.mrxs file is compressed with a lossy compression method (not exactly sure which one). When you convert to CMU-2_raw the result is lossless-ly compressed (using blosc) and thus much larger. The 2nd step conversion to OME.TIFF then uses JPEG-2000 lossless compression, thus also creating a much larger output than the original. You could instead use the lossy JPEG-2000 compression method (raw2ometiff --compression=“JPEG-2000 Lossy” CMU-2_raw CMU-2-JPEG-2000.ome.tiff) and that may help a bit but in my experience that is not a very efficient compressor and it’s quite slow. If your downstream use allows it, you could use the regular JPEG compressor instead, which will compress quite a bit. But of course, beware of the serious artifacts from JPEG especially if any quantitative analysis is planned.
Cheers,
Damir

3 Likes

Thank you both for the input! Unfortunately it seems our size increase is just the normal behavior as @dsudar explained.