Ki67 stain in cells labeled with Quantum dots

Hi Cell Profiler Team,
I just started with cell profiler and try to figure out a pipeline for following problem:
I have co-culutures where one cell type is stained with Quantum dots (QT705) to distinguish the cells, furthermore the cells are stained with DAPI and a KI67 (Alexa488) antibody to identify proliferating cells. Now I like to know how

  1. how many cells are in my picture (DAPI)
  2. how cells are Quantum dot positive
  3. how many cells are stained with Ki67, how many of these KI67 stained cells are stained with Quantum dots.

With the attached pipeline I identify always as many DAPI positive cells as labeled cells (which is not true). I might be that I did not choose the perfect threshold and identification method for the Quantum dots since there dotted structure is not easy to identify for the programm. The identification of the KI67 seems to work (just needs optimisation for the threshold).

Hope you can help me solving my problem. Thanks in advance!


DefaultOUT__7.mat (219 KB)

I would suggest doing the following (which is not too different from this response):

  • Identify the nuclei (as you already have, although some parameter adjustment is necessary for optimal segmentation)

  • Use IdentifySecondary with the Nuclei as the primary objects, and the Ki67 image to identify the secondary objects as Cells. This is not to identify which cells are stained for Ki67 or not, because all
    the cells seem to have some staining associated with them. In this case, use Background or RobustBackground thresholding to detect all the cells. This gets around using the Qdots for the same purpose.

  • Use the Crop module to crop the OrigDot image into the shape of the Cells

  • Use IdentifyPrimAutomatic to identify the dots from the OrigDot image using the per-object thresholding method of your choice. You may need to adjust the settings to insure that only the true positive dots are detected.

  • Relate the dots to the Cells and use ClassifyObjects as before to find the ones that are labeled positively for Qdots.

  • Use IdentifySecondary to identify the Ki67-labeled cells from the Ki67 image. This time, you will need to adjust the settings to insure that only the truly labeled cells are detected.

  • Relate the same dots to the Ki67-Cells and use ClassifyObjects to find the ones that are labeled positively for Qdots.

Hope this helps!

Hey MArk,

Thanks forthe very fast and helpful answer. Now I have the pipeline running, but it needs improvement for the identification of the Ki67 dots in the nucleus. Since there are dots in each nucleus I tried to define a treshold, but anyhow Ifurther improvement has to be done to really identiy the more intense spots and not the whole nucleus
Furthermore I could not use the per-object thresholding for the OrigDot image since this led to a error in processing.
Thanks for further suggestions!


sorry the pipeline…
161109PIPE.mat (1.5 KB)

Hi Nicole,

In this case, then, you will need to crop the Origdot image into the shape of the nuclei and not the cells. This cropped image is then used as input into IdenitfyPrimAuto in order to apply per-object thresholding. The reason you were getting the error with the per-object method is that the input image had not been cropped beforehand, in order to define the object regions.

Another issue is that with per-object thresholding, the setting “Discard objects touching the border…” applies to the object borders, and not the image borders as it does normally. Unless you want qdots touching the nuclei edges to be thrown out, you should change this setting to ‘No’.