Co-localization pipeline?

Dear all

I’m trying to make a co-localization pipeline which in my mind should be very easy to make but I just can’t make it work. I really hope for some good advice.

I have cells stained with:
CD8 antibody (RED)
KV13 antibody (GREEN)
DAPI nuclei (blue)

I have the following goals that I would like to achieve:

  1. Calculate the percentage of All cells (blue) that are CD8 positive.
  2. Calculate the percentage of CD8 cells (red) that are KV1.3 positive (green) - in the picture it is the yellow ones.
  3. Calculate the intensity of the fluorescence - to get a quantitative measure of protein level in the cell (this goes for the KV1.3 positive cells).

In my pipeline I’ve managed to get the CD8 identified as primary objects and also KV1.3 (green) cells as primary objects. I have tried the relate command to get some “ratio” of the red cells that also where Green, but this doesn’t give me a number or ratio?

I 'm looking forward solving these issues :smile:

DefaultOUT.mat (57.6 KB)

I’ve just looked through the examples again. I used some wrong words for the search. I’ve been able to make a pipeline (attached) that I my opinion does the job.

Any comments on the pipeline?

Regarding the quantification of the fluorescent signal I’ve used the IntensityMean value (as far as I remember). Is this how it should be done?

Thanks in advance.

DefaultOUT.mat (18.9 KB)


Your pipeline looks good. The only advice I have is the following:

  • Using the median intensity is more statistically robust than the mean.
  • Using an absolute intensity threshold cutoff in ClassifyObjects to define high vs. low intensity can be tricky. It’s not uncommon for the intensity to change between samples or between images, so your values may not work consistently. I would suggest perhaps using IdentifyPrimaryObjects to define the high-intensity objects; that way, you can leverage the automated thresholding functionality towards dealing with intensity variations.

Lastly, I noticed you seem to be using a pipeline from one of our developer trunk builds, rather than the release version. Can you tell me why you felt the need to upgrade? It doesn’t seem like you have anything in your pipeline which requires it.

Good luck with your analysis!

I tested this pipeline on brain sections having GFP, DAPI and BrdU. It has very good detection for nuclei but 2 problems I would like to correct if possible:

  1. Very bright GFP cells are excluded
  2. cell counts are not exported to the CSV file


(Sorry for the long delay in responding)

You can use MeasureObjectIntensity to measure the intensity of each cell. Then, you can use DisplayDataOnImage to preview the intensity measurements overlaid on the image of the cells, for use in selecting an intensity cutoff. Since MeasureObjectIntensity produces several measurements to choose from, I would recommend using the median intensity as a statistically robust measurement to look at. Then you can use FIlterObjects to exclude cells that fall above the measurement and threshold of your choice. You may want to examine a number of images to insure that the choice of cutoff is consistent across multiple images in your assay.

In ExportToSpreadsheet, is “Select the columns of measurements to export” checked or unchecked?

  • If unchecked, then the counts should be included if you are exporting the per-image measurements.
  • If checked, click the “Press to select measurments” button and make sure that the items under *Image > Count *
    on the tree are checked off.