Help designing a pipeline

cellprofiler

#1

Hi,

i am trying to design a pipeline for several weeks without success. My main problem is i am not able to outline the cells.

I have a set of images of cells that are staining for 3 different components: DAPI (blue), Prot1 (green immunostaining), Prot2 (red immunostaining).

I have a set images of control cells and a few sets of cells with treatments.

What we want to do is measure the effect of those treaments on the level of Prot2 (red).

Prot1 is our reference signal; it covers the whole cell.

Prot2 give a cortical signal, mostly on the edge of the cells.

Can you suggest a simple pipeline?

What i have tried so far:

Load images
IdentifyPrimAutomatic
Identify secondary
MeasureObjectIntensity.

It seems i can’t outline the cells properly.
Then, when i do measurements of intensity, the results are the same for control cells and positive controls, on which the level of Prot2 should be a lot decreased.
Can you give me any advice on threshold settings, etc.?

I have uploaded a set of 3 pics (DAPI, Prot1, Prot2).

Any help would be appreciated!!!
Controls.zip (220 KB)


#2

So, I have a few suggestions for you that may help your analysis in the future.
-Try not to save your images from the scope as JPGs, instead you should use TIF, PNG or another lossless image format. JPGs will be compressed and lose resolution.
-The images are oversaturated and false colored. You can tell an image is oversaturated because your pixel values max out at 1. To check this, load your images in CellProfiler, double-click on the image filename, then click CellProfiler Image Tools , then ShowOrHidePixelData. When you scroll over the a bright area of an object, the maximum value is 1. If you do not know how to turn off the false coloring on the microscope software, you can load your images into CellProfiler and then use ColorToGray to change the colored images to grayscale. Why is this important? Well, IdentifyPrimAutomatic works with grayscaled images.

Things to consider:
-Have you corrected the illumination? In all of your images, it looks like the lower half of the images are brighter than the upper left corner. To correct the illumination, you will need 2 modules: CorrectIllumination_Calculate and CorrectIllumination Apply. The help for these modules can explain the settings for you. (To grab the help, click on the ‘?’ button in the CellProfiler window or the ‘? Module Help’ in the Add Modules window). You should do this for each channel, separately.
-There seems to be a ‘halo’ around each object, regardless of the channel. Do you want to identify the ‘halo’ around the object, or just the bright, center portion of each object? If the former is true, then you want to set the lower threshold to the background-level (ie black portion of the image). If the latter is true, then you want to set the lower threshold to just above the halo values. Again, you can decide this using the ShowOrHidePixelValues tool.
-You may also want to try one of the ‘Adaptive’ algorithms in IdentifyPrimAutomatic. (ie: Otsu Adaptive). This may help deal with the varying halo or background.
A sample pipeline for you might look like this (my comments on which channels/what to measure are in brackets):
LoadImages [load all images]
CorrectIllumination_Calculate [calc for DNA]
CorrectIllumination_Calculate [calc for prot1]
CorrectIllumination_Calculate [calc prot2]
CorrectIllumination_Apply [apply DNA]
CorrectIllumination_Apply [apply prot1]
CorrectIllumination_Apply [apply prot2]
IdentifyPrimAutomatic [identify nuclei in corrected DNA image]
IdentifySecondary [identify cell boundaries using corrected Prot1]
MeasureObjectIntensity [use Cell boundary objects from IdSecondary, use image for Prot2]
…and other Measure modules
OverlayOutlines [if you want to save the outlines of what you identified]
SaveImages [to save the images]
ExportToDatabase [or ExportToExcel module…depending on how much data you have to export]

You can calculate ratios using Excel (or another program, ie: R) post-CellProfiler or you can use the CalculateRatios module. Again, be sure to check the help for detailed information on the specific settings.
Hope this helps.


#3

Thanks, this should really help.
My orginal files are from Volocity software, i can export them as .tif easily.

I will try all of your recommendations and come back with results, hopefully!

-SL


#4

Oh, i was wondering:

when i load colored .tiff images with the LoadImages module, they are loaded as grayscale automatically. Is this normal?

SL


#5

Are they REALLY colored or is it false-coloring from the software (so it only appears colored in the software)? To check this, just open up the .tiffs in CellProfiler. CellProfiler will display as a colored image (if it is color) or grayscale (if it is saved as a grayscale image).