make measurements in one image using objects generated in reference image

[AnthonyOrvedahl]

I'm interested in analyzing serial sections for overlap of different IHC stains (or intensity of one stain in different populations). Is it possible to export the objects generated in one file and make measurements within these objects in a second image (with some rotation or adjustments as necessary to make them overlap?). Real world example, I have one slide stained for F4/80, and an adjacent section stained for cleaved caspase 3. I would like to compare the amount of cleaved caspase 3 signal in the F4/80 positive and F4/80 negative cells. Can I use the cell objects generated from the F4/80 image to measure signal in the cleaved caspase 3 image? I looked around the posts and didn't find an answer.

[Svidro]

It is possible, but depends on how regular the sequential sections are. I have seen it done before, but the registration was not done in QuPath. You can definitely take whole sets of objects and alter their position by translation and rotation, but determining how much of a translation and rotation are needed are somewhat up to you.

In short, I would recommend finding an external solutions to match up your images as perfectly as possible, and then you can copy and paste your detections between images, and use "Add Intensity Features" for whole cell measurements or a script for cytoplams/nuclear measurements to generate values using the new image data.

Further information, though much of it references each other.
#171
#162

A couple of posts on image registration in general.
https://groups.google.com/forum/#!searchin/qupath-users/registration%7Csort:date/qupath-users/5-JMvmCKRBo/6NeAyDwsBQAJ
https://groups.google.com/forum/#!searchin/qupath-users/registration%7Csort:date/qupath-users/VLJL6UCXqEk/lvu6LO0bBAAJ

[petebankhead]

@AnthonyOrvedahl Figuring out the rotation or adjustments necessary to make the cells overlap is a difficult problem to solve, and beyond the scope of QuPath currently.

Potentially anything is possible, since QuPath is open source and supports scripts and extensions, and so someone might take on the task of adding this functionality in the future. I have looked a bit into aligning slides to transfer larger regions, but the kind of fine-grained alignment necessary for cell-by-cell measurements is much more awkward and it isn't something I am actively working on myself.

As @Svidro says, if you take care of registering the slides elsewhere then it may well be possible to hack together something in QuPath to transfer the detected cells and that could be useful... but it would take some effort and would probably not be ideal in terms of workflow or accuracy.

Two other ways in which QuPath might help with looking at multiple markers per cell:

• Support for brightfield and fluorescence multiplexing
• Ability to align restained sections of the same tissue

Both are things I'm looking to add or improve in QuPath. To some extent, the first is already present; a nice solution for the second one would also them mean better support for images registered-outside-of-QuPath... but it's not quite there yet.

One other things: if you've any screenshots of example images that would be helpful. I've assumed you're working with brightfield whole slide images. Registering smaller image regions is more doable, as there are a range of relevant registration plugins within Fiji.