Opening and cropping large tifs

I have coverted images of scanned pathology sections from MRXS to TIFF using the 3D histotech slide converter then processed with Fiji into smaller tiffs. I do this with some script provided to me.
However, there seems to be a threshold of 6GB for Fiji to process these converted TIFFS.
My patholoy slides have two pieces of tissue on them so I have attempted to crop out one piece to reduce the image size but I can’t manually crop anything in the order of 1.5GB.
Many thanks,

Hi there.

From the documentation of QuPath, there is native support for your MRXS format.

To avoid all the suffering of resaving and cropping these images, please give it a shot. It’s free, open source and very well documented.

And whatever you want to do in Fiji should be doable as it packs ImageJ in there :slight_smile:



Here’s some more updated info about .mrxs specifically from the QuPath docs:

1 Like

Thanks Oli I’ll try it out.

Hey guys,

Sorry if this is hijacking the thread. I’m having a similar issue.

I have large Phillips TIFFs and am trying to crop the WSI. QuPath is GREAT. For smaller images I would export the region to ImageJ with a downsample of 1 but for larger images, this wouldn’t work.

Is there a work around? Ideally I would like to crop WSI without down sampling the images.

Thanks so much.

I suspect that can be done, but I’m not sure if the code is already written. If you figure it out, it would probably be very useful to others!
I don’t do a lot of exporting, but my idea would be to combine the tile writer here:
with the ome-tiff writer and tile combiner for the vectra files here:

Obviously both scripts would need to be modified, as the export would need to only include some sort of annotated area, but that would probably only take modifying some of the true/false settings.

Rebuilding it would be more work as the image type is probably different for the Vectra files that script is intended to stitch into a whole slide image. Information relating to the position of the tiles would probably be somewhere different as well! But at least the method of using the ome-tiff pyramid builder is there as an example.

*of course it would be easier if you can figure out how to run the ImageJ part of the analysis through QuPath instead.

I don’t know exactly what you mean with ‘this wouldn’t work’, but my guess is that you hit a fundamental limit in the size of image that can be supported by ImageJ. There is more technical information about it in the ImageJ FAQs here (note that you might be able to open the image with ImageJ2, but not display it in a viewer).

A major part of the reason QuPath exists is because of this limit, which affects pretty much any image analysis software not explicitly designed to handle WSI.

In general, for exporting via ImageJ using QuPath you have to have an image small enough that ImageJ can handle - either by cropping it, downsampling it, or both.

Your other option is to export it as an OME-TIFF pyramid. But that will still be a WSI that can’t be used everywhere.

There is more information about exporting images from QuPath at

1 Like