Trouble selecting the threshold for speckles in my pipeline

Hi,

I’m trying to modify a pipeline downloaded from cellprofiler ‘Example Speckles’ for my assay. My goal is to find speckles (stress granules) in two separate channels and then use the relate objects module to find out how many of the speckles overlap. This is to classify the speckles as stress granules or not (since by our definition, both proteins would need to be present for the speckle to be designated as a stress granule). I am using two of the three channels imaged to assess this, and there is no cellular marker stain. I am however unsure about a couple of things:

  • What does the EnhanceFeature module do?
  • Using the IdentifyPrimaryObjects after EnhanceFeatures I’ve chosen Otsu (and three classes). If I do not modify the lower and maximum values for the threshold I get speckles both in my positive and negative control images . In fact, I get thousands more in my negative control due to the diffuse staining. If I set the lower threshold to 0,08, all of the speckles in both positive and negative control images dissappear. This seems very low already, but when I modify this to 0,001 I get quite a few speckles in my positive control, and a very small number in the negative control. Of those in the negative control, none are related in the RelateObjects module. This seems like the sweet spot, but arbitrarily set, so but do you have any guidance on how to determine this lower threshold value?
  • In a different pipeline in CellProfiler, ‘Example Colocalization’ there are some CorrectIllumination modules added. When is that helpful? Can it hurt to add it and align the images in a speckle finding pipeline?
  • Feel free to comment on other things in the pipeline if you find that I have done added something unneccessary, or have mis-used a function.

See the pipeline I’m working on and a positive and negative control in attached files

Thank you for your help.

Kind regards,
Frida N
img_000000000_568 Ex_000.tif (8.0 MB) img_000000000_633 to combine with 568_000.tif (8.0 MB)img_000000000_568 Ex_000.tif (8.0 MB) img_000000000_633 to combine with 568_000.tif (8.0 MB) Stress granules_20200318_revised pH2AX pipeline.cpproj (961.1 KB)

Hi @FridaN,
In general, to learn more about each module and what it does, you can select the module (by clicking) and use the “?” button on the lower left of the pipeline panel (right above the Start Test Mode button).
There are also tutorials and manuals available on the website as you may have already seen.

To know what Threshold to use, you could hover over different parts of your image, and you’ll see intensity values on the lower right of the window.
You could also use the MeasureImageIntensity module to have a better overall intensity measurements of your image.

CorrectIllumination module could help to calculate and then reduce uneven illumination/background in images.

EnhanceOrSupressFeatures module could help to enhance or suppress the intensity of certain pixels relative to the rest of the image.

Hope this helps, let us know if you still have trouble analyzing your images.

Hi @Nasim
Thank you for your help. I am using the functions you mention, but when it comes to thresholding I’m still unsure of when to feel confident that I chose the correct one. It just seems a little subjective, and if I go back in 6 months I might choose a different threshold. If I interpret the description of the Otsu thresholding method with three classes correctly, I need to pick a lower threshold between the foreground and middleground pixel intensities to exclude my diffuse staining. I will try running through all my negative control images and determine what range of values the diffuse stainings have and then go from there.
Again, thank you for your time.

Regards,
Frida

Hi @FridaN,

I checked your pipeline. I had change primary object parameters a bit, it looks better I guess though I have a slight confusion with your positive & negative control.
Regarding your illumination query with respect to example, I think it depends on the image based on what kind of illumination you want to remove you should choose the method.
Please find attached screenshot & pipeline.
Hope it helps.

Stress granules_20200318_revised pH2AX pipeline1.cpproj (1.3 MB)

  1. Regards,
    Lakshmi
    Fujifilm Wako Automation (Consultant)
    www.wakoautomation.com
    For CellProfiler training or optimised pipeline write to,
    lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Hi @lakshmi,

Thank you so much for having a look at my Pipeline for me. I tried running it and I’m getting a little more granules in my negative control than I would like though. Going from your changes, I raised the threshold for one of my stainings and was able to eliminate the problem.

Thanks again for your time!
Frida

Hi @FridaN,

Great!! You are most welcome…

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging