Multiplex image problems in Qupath

Hi there,

I’m trying to analyse my multiplex images using version Qupath 0.2.0-m8. I scan my images (.ndpi format). Using Qupath I select specific areas within my tissue and I send them to Image J using a downsample factor of 2. Then I align that images on Image J, so I end up having a 8bit stack image with z= number of markers I have (5 in my specific case, z=5) in a .tiff format. Then, still in Image J, I transform stack to images and I merge all of them assigning a different colour to every image and inverting LUT. Once I set up the brightness and contrast for every marker I save the merge image in OME.TIFF. The problem is visible at this point, when I open the image with Qupath to proceed with segmentation. I can see the colours are all inverted, background is the color I assigned to my marker of interest, and therefore the program is just analysing that background. Could you please help me with this? I have the feeling my problem is image processing related but I’m unable to discover where the problem is. BTW, the example image in 32bit works perfectly in Qupath 0.2.0-m8, so my understanding is Qupath hasn’t any problem.

Thanks a lot in advance for any help and feedback I could have


My first recommendation would be to not invert, since the inversion seems to be the problem. But also provide a sample image if possilbe (even a small area extracted using the QuPath ImageJ functionality) for testing along with the exact steps (Record Macro helps here) if possible.

I just did this recently, and my biggest problem was naming the different channels, which is something Pete has made simpler in M9.

*You mentioned multiplex images, which is usually fluorescence. My example was actually aligning brightfield images and extracting the DAB channel, and I’m not really sure what you are working with from your description.