CP exceeds the counting?

cellprofiler

#1

Dear CP team,
Thanks a lot for yr alltime support.
Im facing some problem with counting the number of objects.

Briefly, Im working on a CNT assay where Im using three channels: nuclei (Blue), protein 1 (Green GFP) and protein 2 (Red) obtained by Zeiss confocal microscope. Using LSM imager, I exported the images into bmp format.
Im using the following pipeline:
LoadImages
LoadSingleImage
LoadText
CorrectIllumination_Apply
CorrectIllumination_Apply
CorrectIllumination_Apply
IdentifyPrimAutomatic
IdentifyPrimAutomatic
IdentifySecondary
IdentifySecondary
IdentifyTertiarySubregion
IdentifyTertiarySubregion
MeasureCorrelation
MeasureObjectIntensity
MeasureObjectIntensity
MeasureObjectIntensity
MeasureObjectAreaShape
MeasureTexture
MeasureTexture
MeasureTexture
CalculateRatios
CalculateRatios
CalculateRatios
CalculateRatios
ClassifyObjects
ClassifyObjects
ClassifyObjects
CalculateStatistics

But when see the excel file result, The number of objects counted is same (as manual counting) in some images whereas its quite high in other images.
I have tried with various min,max diameter ranges. Also, with MoG global and Adaptive, Otsu global and Adaptive and threshold correction factors.
tried with Tif images in addition to bmp images.
The result hardly changes.

Plz, guide me.
Thanks







#2

Here are some other images whose counting no. is higher than the actual manual no.







#3

Hi,

Without further details on the parameters used in the pipeline, it is hard to diagnose your problem. I presume the problems with identification are occurring in IdentifyPrimAutomatic? If so, would you be able to post pictures of the output of that module on the image where the count is inaccurate?

Regards,
-Mark


#4

Thanks Mark,
Im posting the IdentifyPrimAutomatic module here. Also, the output of that module.

Module #7: IdentifyPrimAutomatic revision - 12
What did you call the images you want to process? CorrBlue
What do you want to call the objects identified by this module? Nuclei
Typical diameter of objects, in pixel units (Min,Max): 25,45
Discard objects outside the diameter range? Yes
Try to merge too small objects with nearby larger objects? No
Discard objects touching the border of the image? Yes
Select an automatic thresholding method or enter an absolute threshold in the range [0,1]. To choose a binary image, select “Other” and type its name. Choosing ‘‘All’’ will use the Otsu Global method to calculate a single threshold for the entire image group. The other methods calculate a threshold for each image individually. “Set interactively” will allow you to manually adjust the threshold during the first cycle to determine what will work well. MoG Global
Threshold correction factor 1.2
Lower and upper bounds on threshold, in the range [0,1] 0.04,1
For MoG thresholding, what is the approximate fraction of image covered by objects? 20
Method to distinguish clumped objects (see help for details): Intensity
Method to draw dividing lines between clumped objects (see help for details): Intensity
Size of smoothing filter, in pixel units (if you are distinguishing between clumped objects). Enter 0 for low resolution images with small objects (~< 5 pixel diameter) to prevent any image smoothing. 5
Suppress local maxima within this distance, (a positive integer, in pixel units) (if you are distinguishing between clumped objects) Automatic
Speed up by using lower-resolution image to find local maxima? (if you are distinguishing between clumped objects) Yes
Enter the following information, separated by commas, if you would like to use the Laplacian of Gaussian method for identifying objects instead of using the above settings: Size of neighborhood(height,width),Sigma,Minimum Area,Size for Wiener Filter(height,width),Threshold /
What do you want to call the outlines of the identified objects (optional)? Do not save
Do you want to fill holes in identified objects? Yes
Do you want to run in test mode where each method for distinguishing clumped objects is compared? No

Module #8: IdentifyPrimAutomatic revision - 12
What did you call the images you want to process? CorrGreen
What do you want to call the objects identified by this module? ThresholdedCells
Typical diameter of objects, in pixel units (Min,Max): 50,999999
Discard objects outside the diameter range? Yes
Try to merge too small objects with nearby larger objects? No
Discard objects touching the border of the image? No
Select an automatic thresholding method or enter an absolute threshold in the range [0,1]. To choose a binary image, select “Other” and type its name. Choosing ‘‘All’’ will use the Otsu Global method to calculate a single threshold for the entire image group. The other methods calculate a threshold for each image individually. “Set interactively” will allow you to manually adjust the threshold during the first cycle to determine what will work well. .03
Threshold correction factor 1
Lower and upper bounds on threshold, in the range [0,1] 0,1
For MoG thresholding, what is the approximate fraction of image covered by objects? 40 Method to distinguish clumped objects (see help for details): None
Method to draw dividing lines between clumped objects (see help for details): None
Size of smoothing filter, in pixel units (if you are distinguishing between clumped objects). Enter 0 for low resolution images with small objects (~< 5 pixel diameter) to prevent any image smoothing. Automatic
Suppress local maxima within this distance, (a positive integer, in pixel units) (if you are distinguishing between clumped objects) Automatic
Speed up by using lower-resolution image to find local maxima? (if you are distinguishing between clumped objects) Yes
Enter the following information, separated by commas, if you would like to use the Laplacian of Gaussian method for identifying objects instead of using the above settings: Size of neighborhood(height,width),Sigma,Minimum Area,Size for Wiener Filter(height,width),Threshold /
What do you want to call the outlines of the identified objects (optional)? Do not save
Do you want to fill holes in identified objects? No
Do you want to run in test mode where each method for distinguishing clumped objects is compared? No

Im also attaching the illumination corrected images in next mail.
Kindly hav a look.
wishes
Mridul KK




#5

Illumination corrected images






#6

Thanks for the images. Another piece of information which I could use (which I should have asked for earlier) is what information/data do you want to obtain from the green and red images. What are you hoping to observe?

In the meantime, to identify the objects in the blue image (nuclei) try using MoG global and changing the typical diameter. The images you uploaded came out to be 756 x 756 pixels (not sure if this is the same as your originals) so the nuclei about 60-75 pixels in diameter. The diameter range you have now seem to be far too small, which is important since these are used to determine the automatic values for the smoothing and maxima parameters lower down the list.

You can check the length yourself by going to CellProfilerImageTools at the top of the figure window and selecting ShowOrHidePixelData. The pointer then turns into a crosshair which you can use to draw a line by clicking on the image, dragging a line and looking at the distance measurement which appears in the lower left.

Once you have these values and input them into IdentifyPrimAuto, the nuclei seem to be identified just fine.

Regards,
-Mark


#7

Hi Mark,
Thanks for yr suggestion. Let me see, if it works.

With the red & green channels, we want to see whether both proteins are getting activated at same time or different upon stimulation. Its basically, investigating the activation and translocation pattern of these two proteins for same stimulus.
I hope, CP will be able to give me this information.
I think, for total of three channels, the pipeline of object identification is sufficient with
CorrectIllumination_Apply (blue)
CorrectIllumination_Apply (green)
CorrectIllumination_Apply (red)
IdentifyPrimAutomatic (nucleus with diamater range + MoG global)
IdentifyPrimAutomatic (thresholded cells with diamater range + Otsu global)
IdentifySecondary (propcells)
IdentifySecondary (distance cells)
IdentifyTertiarySubregion (propcytoplasm)
IdentifyTertiarySubregion (distcytoplasm)

Just want to know, if I decided NOT to work with propcytoplams or its not giving gud result, I guess I can drop this module from the pipeline. Is it same with IdentifyPrimAutomatic (thresholded cells) too?

Thanks
Mridul KK


#8

Hi Mridul,

Re: Translocation - One issue to mention is that even with our example pipeline, image processing is difficult because you are relying on a florescence signal that is not stationary (i.e, the translocating protein) to determine where the cell boundaries are. If you look at the output for the IdentifySecondary module for some of the images where the protein has completely translocated to nucleus, you can see what i mean.

If you have another cell membrane stain altogether that is not translocating, the problem becomes easier because you have a non-changing image for IdentifySecondary to work from.

Looking at your images, it seems that image3_2 gives a more reliable sense of the cell borders than image3_3. But if it is translocating as you say, it won’t remain that way over the course of the experiment, correct?

However, it seems that the blue channel (image3_1) has some amount of fluorescence at the borders. So try using IdentifySecondary on the blue channel. I did some work with it and it looks like Propagation as the identification method and Otsu global or MoG global as the thresholding method, and perhaps adjust the threshold correction factor. Or you can set the threshold to an absolute number between 0 and 1 if none of the thresholding methods work better.

Re: Whether to work with propcytoplams or not - Yes, if it is not working for you, you can drop it and the modules that use it. Remember, the pipeline is meant to be used according to your specifications; if you find something in an example pipeline that is not useful or applicable for your needs, feel free to omit it. Additionally, the pipeline will run to completion quicker if it is not performing unneeded (or unwanted) calculations.

Regards,
-Mark


#9

Thanks Mark.
I did played with diameter range as well as threshold values. Its better though im getting +3 cells per image.

Mridul KK