Fluorescent cells

He,

I am trying to create a pipeline to count the ratio between green and red microspores (stained with FDA/PI). However I am having difficulties with the CorrectIlluminateionCalcute and Apply modules. My settings are somehow incorrect because I get blackimages, white images or no differce after Apply.
If i remove Illumination calcute and apply it doesn’t really go any better. Because when I identify the primary objects for Green and Red I get clumps and other things and I think this is due to the error in the illumination settings.
At the moment I have setting to A. count all the green cells and then B. count the total amount of cells but then I can still calculate a ratio so this is fine for my purpose. I added my pipeline and images in the attachments, can you help me out?

Thanks! Anne





pipeline FDA.cp (10.7 KB)

Hi Anne,
(Note: I moved this from the CP1 forum to the CP2)

First, a couple pieces of advice:

  • if you can, avoid the JPG image format. They are a lossy – TIFs are a better choice, or PNG as they produce smaller files.
  • Save the images in separate channels as grayscale. It speeds processing time and disambiguates the color unmixing step.

Re: the pipeline:
(1) LoadImages is looking for files with “30” in them. I needed to change this to something common to all files, like “14-12-2010” . Presumably this is just an oversight for these examples.
(2) Use Test Mode for debugging. Mine shows “Caution” symbols where settings are obviously incorrect.
(3) CorrectIllumCalculate, in the “Each” mode, almost certainly should have some smoothing, and not all of yours do.
(4) CorrectIllumApply should use the “Divide” option, not Subtract, when using the “Each” mode and especially when using “Rescale” in CorrectIllumCalculate. This is a common mistake, and likely the cause of your blank images. This is documented in the Help buttons (labeled with “?”).
(5) You are not using the Corrected images (e.g. “CorrRed”) anywhere later in your pipeline. You likely should change the inputs of IDPrimaryObjs to input the Corr images
(6) I would try some other smoothing options in CorrIllumCalc, since the Splines option gives some odd looking illumination functions. Too peaked?
(7) Most of the green cells are being excluded because the size range of 32-60 pixels was too large. Lower it, or don’t exclude objects outside this diameter range.
(8) You can add a CalculateMath module to calculate the ratio that you desire.

I am attaching my pipeline. Let us know how it looks.
David
pipeline_FDA_DL.cp (10.5 KB)

He David,

Sorry for the late response, I was waiting for an email but the system doesn’t seem to do that!
Thanks for the CorrectIlluminationCalcute and Apply help. That seems to work fine but the identifying of the red cells doesn’t seem to work yet.
The range from 17-60 pixels seems to work for the green cells ( 9 calcuted and 9 present) but for the red cells too many cells are counted. Should I use the CorrRed or should I create a new gray picture ( and then CorrectIlluminationCalcute and Apply) and use CorrGray? Or should I do something else?
This problems accours with both JPEG and PNG pictures.

I added the newly adjusted pipeline in the attachments

Thanks! Anne
pipeline_FDA_A 14-1.cp (9.99 KB)

Hi Anne,

You may want to try changing the thresholding method for the red cells, perhaps to either:

  • Otsu global 3-class with the middle class set to background
  • RobustBackground global

In either case, you may need to adjust the threshold correction factor until the identification performs to your satisfaction. Based on a test run, RobustBackground seems to do better with a minimum of later adjustment.

Also, you no longer need the second ColorToGray module.

Regards,
-Mark