Help Counting Fluorescent Cells


#1

Hi,
I am trying to count fluorescent cells (green) but I’m having some issues selecting and adjusting an appropriate threshold method and strategy. CellProfiler seems to be picking up a lot of the background and counting it as a cell but disregarding actual cells for the green fluorescent images although the dapi images seem to be working fairly well.

Here’s the pipeline I have so far:
fj test.cpproj (552.2 KB)

and images:


#2

Hi, can you please re-upload the images, they seem not to be here yet.


#3

heres a link to the imaegs


#4

Hi,

Can you please grant us the access. The dropbox file is currently private.

Thanks.


#5

ok i’ve granted you access


#6

Hi,

Thanks for the image.

It’s in fact a difficult case when you have a noisy background and a low resolution/magnification at the same time.
Although I’ve tried a series of modules to reduce noise, it’s quite unworthy to quantify this set of image

I highly recommend you thoroughly clean the sample after incubation of antibodies, because non-specific binding can dramatically reduce the image quality.

Also if possible, please capture the image with higher magnification.

I share my pipeline here to demonstrate the effort of cleaning the background fj test.cpproj (863.2 KB)

Bests,


#7

Thank you! I’ve since taken new pictures which work much better. ( https://www.dropbox.com/sh/kzetn335zb5dhd4/AADMIXNaLsDIQTuRTcy-f6M0a?dl=0 )
I was also wondering if cellprofiler has a function which would allow me to count only colocalized cells other than the CalculateImageOverlap module?

TEST FJ PROJECT.cpproj (537.8 KB)


#8

Hi,

The new images are better, but you may have to avoid the crowded area, where cells are clustered and detached. You can choose this area by using IdentifyObjectManual and then “MaskImage” to include/exclude areas of images.
(By the way, are these pictures taken with 5x magnification?)

Other way to count only colocalized cells: you may want to relate them first using RelateObjects, and then use FilterObject to keep only the “parent” cells that have minimum 1 “child”.


#9

You can also use MaskObjects with the ‘Remove based on overlap’ setting to keep only cells that overlap in both channels by a certain fraction.