Correlating 3 stains to one object

I have images in three bands: red, green, and blue. Blue is a DNA nuclear stain, red is also a nuclear stain, and green is a surface stain. The cells that are positive for the green stain are vascular in structure, so they are elongated cells. I want to try and see if there is correlation between high/low intensity in red, and high/low intensity in green.
I have created a pipeline which identifies primary objects in all three of the bands, but I’m not entirely sure what to do next. I can relate the objects to try and get a 1-1-1 ratio where green and red exist. But I think what I really want to do is then open the data in CellProfiler Analyst and create a scatterplot that maps the red intensity vs green intensity per object.
I just am not entirely sure what extra pieces in the CellProfiler pipeline I might still need to add?
Thank you for the help.

Rather than creating 3 different IdentifyPrimaryObjects modules, I’d use IdentifyPrimaryObjects on your blue channel to find Nuclei, followed by IdentifySecondaryObjects on your green channel (with Nuclei as seeds) to find Cells, then measure intensity of Nuclei and Cells in the green and red channels. This should give you the information you want.

Good luck!

The problem with using IdentifySecondary, is there are way more nuclei (primary objects) than either of the other stains. It isn’t a 1:1 ratio.

Okay, so I found the ClassifyObjects Module, which I think is a good portion of what I want to do. However I’m a confused about the output image? I have attached the module output, and it clearly looks like there should be at least 3 populations according to the histogram on the right. But the image only shows blue populations, also, that doesn’t look like near enough objects to actually be the blue population. So I must be misinterpreting what it’s actually showing me, can someone clarify?

The original nuclei image is below, I would expect every one of these nuclei to fall into one of the 4 categories from the ClassifyObjects module:

Also, I really want to get this kind of information for all the objects in several images. I actually cropped my original image into several pieces so that I can run it through CellProfiler in a reasonable amount of time. But I want the statistics and data from the sum of all the images. Can CellProfiler Analyst give me this type of classification product for the sum of my images/objects?

ClassifyObjects when used on two stains at a time only lets you create a single threshold on each variable, so you can’t create 3 populations for any stain- it’s designed for [-/-,-/+, +/-, +/+] classification only. I agree the image on the bottom left doesn’t look right but it’s hard to diagnose from a screenshot.

Analyst can definitely allow you to create multiple gates and populations for each stain, to classify into as many subgroups as you want, and then to apply the model you’ve created to more images. You’ll want to include an ExportToDatabase module at the end of your pipeline and tell it to create a CPA Properties file for easiest use. For something a bit more quick and dirty, you can run your pipeline on all your images and then use CellProfiler’s own built in DataTools on the measurements file created at the end of processing.