Best Approach to Get Multichannel MRXS Files into QuPath

So I understand the open-ness issues that make .mrxs files tricky, but I’m wondering if there is some best practice approach to get multi channel (4 in my case) mrxs data into QuPath. If I try to open it directly I’m seeing 3 of the 4 channels. (Ultimately I just need to use the dapi to guide per cell measurements of the other 3 channels)

I’ve tinkered with exporting the relevant (smallish, couple gig total) section of the slide from caseviewer to a tiled tiff and following the instructions here QuPath multiple channel in separate files - how to merge them? to combine the 4 channels. I haven’t quite gotten it to work for all 4, but given just for 2 channels it locked up for 20 min running the script and once have the image in it qpath is unresponsive I have doubts about it. (I was trying with milestone 8)

If I manually combine those same images in fiji first into a 2gig tif everything works fine,tho the import is a little slow but i wonder if there is a smoother path, or indeed any path for bigger ROI?

A bit of a newbie to QuPath so might be something I’m missing but if anyone has an approach that is working for them I’d love to hear about it.

Have you created the TIFFs in pyramidal format, or just basic TIFFs? If they are not pyramidal, that would be very difficult and could be the cause of your problems as QuPath tries to load each of the entire images. QuPath should have options to write out the pyramidal format in M8, in the File menu.

Related posts involving large TIFF files


Maybe others will have some more useful suggestions.

Thanks, Happy New Year, I admit to being a little naive about the subtleties of tiff formats, but I always assumed the ‘tiled tiff’ export option in caseviewer created a pyramidal format, but I could be wrong. I’d also assume having imported them to fiji with bioformats (and not having gotten the which pyramid level do you want dialogue it gives you for imaris files btw, so maybe it isn’t pyramidal?) and saved them out as a tiff, that the result is a plain tiff. The first grinds to a halt with the import script with qupath using 20gbish memory, the second works fine with a slight initial load hiccup (taking 13gb). The data on disk is only 2gbish. I guess In a worst case I can just always go though fiji until something breaks as the files get bigger.

If you open any of the files in QuPath, you can see whether they are pyramidal in the Pyramid entry in the Image tab.


If the entry says “1” it also is not pyramidal.
image

Not sure about the file size, I expect a dramatic expansion in file size when converting compressed images, as seen in the second link above (50GB files), but if your files are not compressed I am not sure.

Thanks again.

Both the individual channels and the composite from fiji show up as pyramidal in qupath as it turns out (says 1,4,16,64) so I guess fiji is clever enough to carry that along. Would still be very curious what anyone else is doing in practice since the scanner is what it is and 4 or 5 channels is typical here but would be nice to provide users with a smooth protocol they can use themselves and view interactively.

I’m unclear on what exactly are the steps you used but if you saved as a tiff in Fiji then I am fairly sure it will not be pyramidal. However there is a recent QuPath option that will dynamically create a pyramid if desired - which could explain the apparent pyramid levels under the ‘Image’ tab… but this will still ultimately require everything being read into RAM at some point, which would explain the extremely long delay.

If you do have a true pyramidal image (not dynamically generated) it should open in QuPath quickly and require substantially less memory than the full uncompressed image size (which is displayed under the ‘Image’ tab).

The trouble with exporting as tiled TIFF is that it may or may not be pyramidal, and even if it is pyramidal it may be so in a way Bio-Formats doesn’t recognise because pyramidal TIFFs in the wild appear not to be terribly standardised (I’ve seen this reported for some Philips TIFF images). Is there any other way you can export the channels?

In the end, I think the best workflow is likely to be the one from the Glencoe blog post - but I haven’t tried it and don’t work with mrxs images myself.

1 Like

Thanks, I’d put my money on neither being used as pyramidal whatever they are in principle. There’s something probably about the concatenation script that slows down, while the already combined data it can muscle through and make the pyramid on the fly given its not too big. I’ll look at the whole slide to ome tiff conversion program, which looks really useful if I can get it running and it works correctly on these images.