Interactive Image Alignment Across Image Modalities

Hi all,

I’m trying to use the interactive image alignment tool to generate the matrix to transfer multiplexed data from a darkfield imaged slide to its H&E brightfield imaged counterpart. However, it throws this error:

Does this refer to the fact that the image is darkfield? Are there any workarounds?

Thank you!!

Hi David,
A few questions: What are the image types of the two images? For the alignment, are you using image intensity or area annotations? At what stage in the process does this error pop up?

2 Likes

The two image types are Fluorescence and Brightfield (H&E). I was planning on using the affine tranform alginment, but I wasn’t able to get to that point. The error pops up when I hit “ok” after selecting the image I want to overlay.

I have been able to overlay BF onto IF and IF onto BF, and just tested it again now, so the default functionality is fine. There might be something specific about the format of your images, though. The error sounds familiar, but it has been a long time.

Just to verify, you are using 0.2.3?

2 Likes

Usually my workflow with the image alignment is open Image A, then use the interactive image alignment to open and line up Image B, and then I manually copy the matrix into a script I have that transforms the objects on Image B to Image A.

Unfortunately, the error is preventing me from successfully loading Image B (Fluorescence) into the visualizer to generate the matrix.

This actually helps! So I think the problem is this: my Fluorescent image is a 16bit tif, and my HE is a 8bit tif. Could this be it?

If so, I’m at a complete loss for how to resolve this issue…

1 Like

I can overlay 8, 16 and 32 bit images, so it is not just that. It may be some other aspect of the file format, however. Perhaps one of them is pyramidal and the other is not? I am not sure.

I tested BioFormats onto Openslide and vice versa, but not any ImageJ servers. THAT does seem to ring a bell…

Can you check the Log in the View menu to see if there is an error?

1 Like

I’m not sure if this would matter anyway, but what are the image server types?

2 Likes

Fluorescence:


H&E:

Ok, I have never seen that Server type that I can recall. I also went ahead and searched for the error message on the forum and now realize why it sounded familiar.

And this mentions the error as well.

I guess to provide some context, my lab’s imaging computer can’t export 16bit tifs. I had to take them to a neighboring lab computer and export them on their newer software (Zen 2 or 3). The only way I could get this to work was by exporting the individual channels and then concatenating them with an old script of Pete’s I found somewhere online. That’s where the server type comes from :slight_smile:

Can you try doing the registration on 1 channel from the fluorescence image?

1 Like

Just out of curiosity, why do you need to export the images?

You mentioned Zen, and that’s usually converting CZI files into something else, which usually is not necessary.

1 Like

They’re mosiacs - if I open them in Qupath in the proprietary image format, it takes the mosaic image and breaks each tile into its own image in the project.

@smcardle is this where you normally export as BigTIFF? I need to test, but I noticed the OME-TIFF writer is only in certain versions of Zen and I wonder if that would work as well.

@davidnascari There may be other options, but I suspect the easiest way forward, if they are distinct regions anyway, would be to treat them as such by splitting the image into Scenes/files in Zen.

And I can confirm that exporting as OME-TIFF from Zen retains the positions of the different tile regions and can be opened and overlaid on a brightfield image in QuPath. The only downside is that the pixel size is lost and would need to be reapplied using setPixelSizeMicrons()
Oh, and it is a non-pyramidal file, so @smcardle’s recommendation to export via QuPath is still probably the way to go.

Hi Sara, so, if I do it with just one of the individual channel images (which is still 16b) it does allow me to align it and generate the matrix. I guess this means that the problem is the server type. Is there anyway to go in and change it?

1 Like

I’ve never worked with that channel concatenation script, so I don’t know for sure. My instinct says there probably isn’t a way to change the server type on the existing image. You could maybe export from QuPath an OME-TIFF with all the channels and then load that into the projects with normal bioformats. But, then you have to duplicate the data and deal with extra files. But, since you are copy-pasting the alignment anyway, maybe just using 1 channel would be good enough for your purposes?

4 Likes