Defining regions of interest

Hi there,

I’m relatively new to CellProfiler, but familiar with the most basic functions and modules. One thing I’m struggling with, however, is to “define” the regions marked with WGA (plasma membrane marker, red channel) as regions of interest (ROI).

In my experiment, I have marked Cx43 (green), the nucleus (blue) and WGA (red), and I’m trying to quantify the signals from the green channel within defined ROI (i.e. the cytoplasm, plasma membrane). I tried to identify the WGA-stained parts as primary objects, but with this approach I get a lot of objects and thus huge output files, because each object gets a value. Is it possible to define the plasma membrane (WGA stained regions) as one single object?

Also, is my method for defining the cell and cytoplasm ROI good? I ask because my approach was more or less trial and error.
To define the “cells” ROI, in IdentifySecondaryObjects: I used the propagation method, WGA Outlines (from IdentifyPrimaryObjects) as an input image and the nuclei (from IdentifyPrimaryObjects) as input objects. Judging by the output it looked OK (see the attached image). Further, to identify the cytoplasm ROI, I substracted the nuclei from the “cells” in IdentifyTertiaryObject.

Thanks in advance. I’m really liking this software and hope to master it in the future.

1 Like

Also, is my method for defining the cell and cytoplasm ROI good? I ask because my approach was more or less trial and error.

Without seeing the original images, I can’t say for sure, but your description plus your screenshot of what your output is look quite reasonable to me!

In my experiment, I have marked Cx43 (green), the nucleus (blue) and WGA (red), and I’m trying to quantify the signals from the green channel within defined ROI (i.e. the cytoplasm, plasma membrane). I tried to identify the WGA-stained parts as primary objects, but with this approach I get a lot of objects and thus huge output files, because each object gets a value. Is it possible to define the plasma membrane (WGA stained regions) as one single object?

I’m still not totally clear on your use case here- what precisely is the measurement you want? Is it something like “The average Cx43 at the membrane in each cell”? “The average Cx43 intensity at the membrane in each image?” etc etc etc. There are definitely ways I think to get you the info that you want (some of them hackier than others), but exactly how to go about doing it will depend on exactly what you want to know.

2 Likes

Hi Beth and thank you for replying,

To clarify, I want to quantify the Cx43 (green channel) intensity in the plasma membrane (WGA stained regions, red channel). In order to do this, however, I need to (I think) define the WGA stained-plasma membrane as one single object and MeasureObjectIntensity of the green image in the plasma membrane. In other words (kind of), measure the colocalization of Cx43 (green) and WGA (red).

I attached my current pipeline and a test image so you can better see my thinking. Again, thanks for your help!

CP_test_qcoloc.cppipe (19.6 KB)

Image1

I’m still not totally clear on what your question is (or if it’s a combo of a couple of questions)- are you asking how to better isolate just the membrane, how to do the measurement across the whole image rather than per object, or both?

If it’s how to better isolate the membrane:

  • After your IDSecondary to find cells, you can shrink each cell by ~3-4 pixels using the ExpandOrShrinkObjects module- call it ShrunkenCells say. Now do IDTertiary using that and the ShrunkenCells to make an object called, for example, Membrane.

If you want per-cell measurements then at this stage you add MeasureObjectIntensity; if you want to measure the intensity in a given channel across all the objects in that image, you add the MeasureImageIntensity module and set Measure the intensity only from areas enclosed by objects? to Yes; you can now measure just the intensity inside your Membrane objects.

1 Like

I want several things, but what I am currently struggling to accomplish is to measure the green signal within the plasma membrane (defined by the WGA staining).

I really liked your approach (ShrinkObjects) and I will review the results that gives me and compare it with the results of other approaches. What I really have in mind, though, is to “convert the (red) image to an object/ROI”, so that all the green signal within the area/object (defined by the red image) can be measured.

I have tried to ApplyThreshold, then select the output as a binary image and finally IDPrimaryObject, but this approach is no better than IDPrimary Objects directly. Is there any way to convert an image “as is” to an Object?

Thanks for you time and help!

Is there any way to convert an image “as is” to an Object?

No, sorry! But you can try turning off declumping in IdentifyPrimaryObjects (set Method to distinguish clumped objects? to None) to get CP to try to make as few objects as possible.