Need help to analyze an image

[Nariman]

Hello,

I'm new to imageJ and I need your help please !!
I have attached my image and I want to separate this two cells, I found that analyze particle plugin can count the cells as i want but how i can separate them before using it?

Thank you for any help or suggestions ..

[Michael Schmid]

Hi Nariman,

Probably the easiest is Process>Binary>Watershed on the thresholded  
image. Before this, it may help to run a slight background  
subtraction (Process>Subtract Background, sliding paraboloid, no  
light background, radius a few hundred).

If you want to do measurements where the binary (thresholded) image  
is insufficient, use 'Redirect to' to the original data.

Michael
________________________________________________________________

On 14 May 2010, at 12:16, Nariman wrote:

> Hello,
>
> I'm new to imageJ and I need your help please !!
> I have attached my image and I want to separate this two cells, I  
> found that
> analyze particle plugin can count the cells as i want but how i can  
> separate
> them before using it?
>
> Thank you for any help or suggestions ..
> http://n2.nabble.com/file/n5050304/cells.jpg

[Hoeppner, Daniel (NIH/NINDS) [E]]

Hi Nariman,

The attached macro will separate your nuclei.

I recommend using a local threshold approach as you have a range of signals from the different nuclei.  You'll have to download G. Landini's Auto_Local_Threshold .jar file, place it in the plugins folder of ImageJ then restart ImageJ before running the macro.  To run the macro, select plugins>macros>run.  Browse for the macro and select open.

http://www.dentistry.bham.ac.uk/landinig/software/software.html

Text of the attached macro:

run("8-bit");
run("Auto Local Threshold", "method=MidGrey radius=5 parameter_1=0 parameter_2=0 white");
run("Analyze Particles...", "size=5000-Infinity circularity=0.00-1.00 show=Nothing clear summarize add");
run("Revert");

Best Wishes,

Dan Hoeppner
NIH



On 5/14/10 8:43 AM, "Michael Schmid" <[hidden email]> wrote:

Hi Nariman,

Probably the easiest is Process>Binary>Watershed on the thresholded
image. Before this, it may help to run a slight background
subtraction (Process>Subtract Background, sliding paraboloid, no
light background, radius a few hundred).

If you want to do measurements where the binary (thresholded) image
is insufficient, use 'Redirect to' to the original data.

Michael
________________________________________________________________

On 14 May 2010, at 12:16, Nariman wrote:

> Hello,
>
> I'm new to imageJ and I need your help please !!
> I have attached my image and I want to separate this two cells, I
> found that
> analyze particle plugin can count the cells as i want but how i can
> separate
> them before using it?
>
> Thank you for any help or suggestions ..
> http://n2.nabble.com/file/n5050304/cells.jpg

[Nariman]

Hi Daniel and Michael,

Indeed when I tried the same macro, I got 11 particles and they It doesn't encircle the nuclei !!
(and when I applied it to my real image which is .tif I didn't get any result !!)

Then I applied the same instruction that Michael told me about:

run("Subtract Background...", "rolling=300 sliding");

Then I should apply Watershed on the thresholded image, I tried to threshold the image after applying "contrast enhance" and this didn't help me too. This is the macro that I apply it

run("Enhance Contrast", "saturated=0.4");
run("Subtract Background...", "rolling=300 sliding");
run("MultiThresholder", "isodata");
run("8-bit");
run("Watershed");

Watershed didn't segment my cells well!!

Is there any preprocessing could help me before that?
Thank you so much for your help..

[Rasband, Wayne (NIH/NIMH) [E]]

On May 14, 2010, at 6:16 AM, Nariman wrote:

> Hello,
>
> I'm new to imageJ and I need your help please !!
> I have attached my image and I want to separate this two cells, I found that
> analyze particle plugin can count the cells as i want but how i can separate
> them before using it?
>
> Thank you for any help or suggestions ..
> http://n2.nabble.com/file/n5050304/cells.jpg 

You can separate the cells using Process>Binary>Watershed but some preprocessing (blurring) is required. Here is an example macro:

  rename("RGB"); // known name needed for redirect
  run("Duplicate...", "title=Mask");
  run("8-bit");
  run("Gaussian Blur...", "sigma=2");
  setAutoThreshold("Li dark");
  run("Convert to Mask");
  run("Watershed");
  run("Set Measurements...", "area mean min limit redirect=RGB");
  run("Analyze Particles...", "size=1000-Infinity display clear");

-wayne

[Nariman]

Hello,
I'm sorry for asking again but I don't know why watershed separated the particles in this way (I'v attached my image) !!
Maybe I should detect the edges before or something else? or there is another solution??
(I applied the same macro that Wayne wrote it, I added only "Subtract Background" before it and I tried to enhance the contrast but it was always the same results!!)
Thank you so much for your help..

[David Webster]

I'm confused, as usual. Are you trying to do someting like partition the two
touching cells into two separate cells using this image?

David

On Sat, May 15, 2010 at 12:06 AM, Nariman <[hidden email]> wrote:

> Hello,
> I'm sorry for asking again but I don't know why watershed separated the
> particles in this way (I'v attached my image) !!
> Maybe I should detect the edges before or something else? or there is
> another solution??
> (I applied the same macro that Wayne wrote it, I added only "Subtract
> Background" before it and I tried to enhance the contrast but it was always
> the same results!!)
> Thank you so much for your help..
> http://n2.nabble.com/file/n5058431/Mask.png
> --
> View this message in context:
> http://imagej.588099.n2.nabble.com/Need-help-to-analyze-an-image-tp5050304p5058431.html
>  Sent from the ImageJ mailing list archive at Nabble.com.
>

[Michael Schmid]

> I'm sorry for asking again but I don't know why watershed separated the
> particles in this way (I'v attached my image) !!
> Maybe I should detect the edges before or something else? or there is
> another solution??
> http://n2.nabble.com/file/n5058431/Mask.png

It is hard to see in the png image (it seems to be downscaled) - there
seems to be a segmentation at a constriction in one of the cells, so the
standard Watershed command is too sensitive. See this post for a macro
that gives you control over the sensitivity of watershed:
https://list.nih.gov/cgi-bin/wa.exe?A2=ind0909&L=IMAGEJ&F=&S=&P=214248

Michael