Problem with Identify Secondary Module

Image Analysis
#1

I am having problems with the Identify Secondary Module.

As part of an experiment, we labeled three separate populations of cells with different colored membrane dyes, we then combined the populations, let them grow, fixed and added DAPI. So in any given image, I have a combination of green, red, and far red labeled membranes along with nuclei.

The Identify Prim works perfect. But since the Identify Secondary is using the Primary objects when it goes to identify the cell edges, it will “find” cell bodies that aren’t actually there. I have tried a variety of methods and haven’t come up with anything.

Attached is a sample image of my DAPI stain along with a membrane labeled image. I also included a sample of what cell profiler generates.

This seems like such a simple thing, any suggestions.

Thanks

Position 13_ex387 em415-470 (Dapi,Hoechst)-01.tiff512×512 513 KB


Position 13_ex482 em513-575 (FitC, Alexa 488, Cy2)-01.tiff512×512 513 KB


Picture 3.png936×838 92.1 KB

#2

Hi,

The issue you’re describing is actually expected behavior on the part of IdentifySecondary. The module attempts to use the input image to find a border between the primary object and the background based on the local image intensity. If the image has no variation to speak of (e.g. the primary object is sitting on a background region), this border doesn’t extend much beyond the object, and may be almost identical with the primary object border itself. This is why you are seeing secondary bodies with the same shape as the primary bodies in areas where the input image is mostly black.

Fortunately, you can then exclude these bodies in the following way:
(1) MeasureObjectIntensity using the secondary objects and the input image used for IdentifySecondary.
(2) FilterByObjectMeasurement using the secondary objects as the object to filter by, Intensity as the category and the input image used for IdentifySecondary as the image measurements. For the feature, you can choose one of the intensity measurements (e.g, MeanIntensity, feature #2) or a similar one and then set the minimum threshold that must be satisfied over each secondary object. In this way, secondary objects which are sitting on dim/dark backgrounds are excluded whereas those that encompass actual cells are kept.

Hope this helps!
-Mark

#3

That was most helpful. Thank you very much.

Heather