Segment phase contrast image

[Laurent.guerard]

Hi all,

I've been trying to segment a phase contrast image with no real success these past weeks.
The images are time series in order to track the movement of these cells. Would any of you have ideas on how to it ?

Here is an example :


There are 2 main problems with the images though :
1. There are dead cells/debris in the images which are very bright and might cause problems in the segmentation
2. The cells are really confluent, making it also difficult to separate close cells.

Any help would be welcome :)

Thanks.

[Jan Eglinger]

Hi Laurent,

have you tried Weka Trainable Segmentation to do this?

http://imagej.net/Trainable_Weka_Segmentation

Here are a few threads on the forum discussing that:

http://forum.imagej.net/t/analysis-of-aggregates/303?u=imagejan
http://forum.imagej.net/t/segmentation-of-polymer-fibres-nonwovens-on-optical-microscopy-images/2313?u=imagejan
http://forum.imagej.net/t/image-segmentation/2359?u=imagejan

Hope that helps,
Jan


On 17.08.2016 15:55, Laurent.guerard wrote:

> Hi all,
>
> I've been trying to segment a phase contrast image with no real success
> these past weeks.
> The images are time series in order to track the movement of these cells.
> Would any of you have ideas on how to it ?
>
> Here is an example :
> <http://imagej.1557.x6.nabble.com/file/n5017034/TiledImage_T00_Trans.png>
>
> There are 2 main problems with the images though :
> 1. There are dead cells/debris in the images which are very bright and might
> cause problems in the segmentation
> 2. The cells are really confluent, making it also difficult to separate
> close cells.
>
> Any help would be welcome :)
>
> Thanks.
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Majurski, Michael Paul (Fed)]

Laurent,

In the past I have broken the problem into two parts, first create a Foreground-Background mask (simpler task), and then leverage the mask to split apart single cells.


To obtain a Foreground-Background mask for phase-contrast I would suggest thresholding the image gradient.

In Fiji the command is called "Find Edges" (I have no idea where it is in the menu structure, I just use the Command Finder (http://imagej.net/Using_the_Command_Launcher ) ).

Once you have the gradient you can apply the many thresholding methods Fiji has(http://imagej.net/Auto_Threshold ) and see which one comes closest to your intuition of a correct segmentation.

This paper highlights our tool for doing just that (but with a new threshold selection method).
http://onlinelibrary.wiley.com/doi/10.1111/jmi.12269/full 
There is a Fiji plugin implementation of this method available at our https://isg.nist.gov activities page (EGT Segmentation).



Once you have your Foreground-Background mask, you can use it to guide a watershed based on pixel intensity (phase contrast cells tend to have a dark core and a bright halo).

This paper discusses our tool for performing single cell segmentation based on that approach and we have had success with phase contrast before.  
https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-014-0431-x
This method is implemented in Matlab (with an executable that can be run without a Matlab license).


There is no guarantee that these tools will work with your images, but it is at least a place to start looking.
Good luck segmenting.
~Michael Majurski





-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Laurent.guerard
Sent: Wednesday, August 17, 2016 09:56
To: [hidden email]
Subject: Segment phase contrast image

Hi all,

I've been trying to segment a phase contrast image with no real success these past weeks.
The images are time series in order to track the movement of these cells.
Would any of you have ideas on how to it ?

Here is an example :
<http://imagej.1557.x6.nabble.com/file/n5017034/TiledImage_T00_Trans.png>

There are 2 main problems with the images though :
1. There are dead cells/debris in the images which are very bright and might cause problems in the segmentation 2. The cells are really confluent, making it also difficult to separate close cells.

Any help would be welcome :)

Thanks.




--
View this message in context: http://imagej.1557.x6.nabble.com/Segment-phase-contrast-image-tp5017034.html
Sent from the ImageJ mailing list archive at Nabble.com.

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Laurent.guerard]

Hi Jan,

Thanks for the answer. I tried some machine learning algorithms (Weka,
one in ilastik etc) and can't find anything good enough for these images ..

I'll keep trying though and I'll read the link you sent.

Thanks.

*------------------
*

*Laurent Guerard*


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Laurent.guerard]

Hi Michael,

Thanks for the answer, I'll play with it and come back to you.

*------------------
*

*Laurent Guerard*


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Laurent.guerard]

Hi Michael,

Thanks for the segmentation link it worked very nicely. I now have delimited what is background and what is foreground.

Do you have any guidance on that watershed based on pixel intensity ?
I tried installing and running the matlab code but i'm having problems with compiler (My matlab seems to be too recent and it doesn't work with version 9 of the compiler)

Thanks again for your help.
Laurent
--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Majurski, Michael Paul (Fed)]

Laurent,

Did you try to use the compiled version of Fogbank running with an MCR? Or did you run from source?

Both the executable and the source should be available at:
https://isg.nist.gov/deepzoomweb/resources/csmet/pages/fogbank_segmentation/fogbank_segmentation.html

I you have not tried the source version it is more likely to work. I usually refer people to the compiled version since not everyone has a matlab license.

If you are having matlab compiler problems w.r.t. the mex files, none of those are required. They are only run if the mex64 file exists. The equivalent basic matlab code is included and it checks whether the mex file exists before trying to launch it.

Given that mex file versions are very finiky, I would suggest either recompiling the mex64 files from the .c source files, or simply deleting them and letting it run basic matlab code. There is basic matlab code included and it only runs the mex file if it can find it. If the mex files appear to be the problem, and you are not comfortable recompiling them from the included .c files, simply delete the .mex64 files and the code will run using basic matlab (albeit slower).

As for how to segment, there is a basic guide in the doc folder of the source code which is a good place to start for using the tool.

~Michael Majurski



-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Laurent Guerard
Sent: Thursday, August 18, 2016 07:27
To: [hidden email]
Subject: Re: Segment phase contrast image

Hi Michael,

Thanks for the segmentation link it worked very nicely. I now have delimited what is background and what is foreground.

Do you have any guidance on that watershed based on pixel intensity ?
I tried installing and running the matlab code but i'm having problems with compiler (My matlab seems to be too recent and it doesn't work with version 9 of the compiler)

Thanks again for your help.
Laurent
--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Laurent.guerard]

Hi Michael,

I only tried the executable so far but I double click on it I get an
error message saying

"Could not find version 8.1 of the MCR. Attempting to load
mclmcrrt8_1.dll. Please install the correct version of the MCR. Contact
your vendor if you do not have an installer for the MCR."

I tried downloading MCR 8.1 from Matlab's site but I get the same error
everytime and when I check the version that Matlab is using it says 9
which corresponds to Matlab 2015b that I'm using.

I'll also try with the source code to see if it possible to make it work
but I'm looking at possibly building a pipeline in Fiji to do batch
processing so I'm also looking at possible plugins and/or macros.

Thanks for your help !

*------------------
*

*Laurent Guerard*


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[jmutterer]

Hi, you could also try the following software by Fabrice Cordelieres that
does tracking of cells in phase contrast videos.
https://sites.google.com/site/itrack4usoftware/home
Sincerely
Jerome.

Le 18 août 2016 6:54 AM, "Laurent Guerard" <[hidden email]> a écrit :

> Hi Michael,
>
> I only tried the executable so far but I double click on it I get an error
> message saying
>
> "Could not find version 8.1 of the MCR. Attempting to load
> mclmcrrt8_1.dll. Please install the correct version of the MCR. Contact
> your vendor if you do not have an installer for the MCR."
>
> I tried downloading MCR 8.1 from Matlab's site but I get the same error
> everytime and when I check the version that Matlab is using it says 9 which
> corresponds to Matlab 2015b that I'm using.
>
> I'll also try with the source code to see if it possible to make it work
> but I'm looking at possibly building a pipeline in Fiji to do batch
> processing so I'm also looking at possible plugins and/or macros.
>
> Thanks for your help !
>
> *------------------
> *
>
> *Laurent Guerard*
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Majurski, Michael Paul (Fed)]

Hmm.

If you have installed the correct version of the MCR, per my understanding it should work. Maybe its finding a newer version of Matlab and trying to help you out by defaulting to the newer one?

If you do have a matlab license try running from source.

Porting fogbank to a Fiji plugin has been on my todo list (fairly far down) for awhile. If I do get it ported to Java as a Fiji plug I will make an announcement here which you might see.

I am not really familiar with any other automatic phase-contrast single cell segmentation Fiji plugin tools. Others on this list might have other ideas.

~Michael



-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Laurent Guerard
Sent: Thursday, August 18, 2016 09:54
To: [hidden email]
Subject: Re: Segment phase contrast image

Hi Michael,

I only tried the executable so far but I double click on it I get an error message saying

"Could not find version 8.1 of the MCR. Attempting to load mclmcrrt8_1.dll. Please install the correct version of the MCR. Contact your vendor if you do not have an installer for the MCR."

I tried downloading MCR 8.1 from Matlab's site but I get the same error everytime and when I check the version that Matlab is using it says 9 which corresponds to Matlab 2015b that I'm using.

I'll also try with the source code to see if it possible to make it work but I'm looking at possibly building a pipeline in Fiji to do batch processing so I'm also looking at possible plugins and/or macros.

Thanks for your help !

*------------------
*

*Laurent Guerard*


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Ignacio Arganda-Carreras-2]

Another option would be the Morphological Segmentation plugin available
with MorphoLibJ: http://imagej.net/Morphological_Segmentation

On Thu, Aug 18, 2016 at 4:30 PM, Majurski, Michael Paul (Fed) <
[hidden email]> wrote:

> Hmm.
>
> If you have installed the correct version of the MCR, per my understanding
> it should work. Maybe its finding a newer version of Matlab and trying to
> help you out by defaulting to the newer one?
>
> If you do have a matlab license try running from source.
>
> Porting fogbank to a Fiji plugin has been on my todo list (fairly far
> down) for awhile. If I do get it ported to Java as a Fiji plug I will make
> an announcement here which you might see.
>
> I am not really familiar with any other automatic phase-contrast single
> cell segmentation Fiji plugin tools. Others on this list might have other
> ideas.
>
> ~Michael
>
>
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
> Laurent Guerard
> Sent: Thursday, August 18, 2016 09:54
> To: [hidden email]
> Subject: Re: Segment phase contrast image
>
> Hi Michael,
>
> I only tried the executable so far but I double click on it I get an error
> message saying
>
> "Could not find version 8.1 of the MCR. Attempting to load
> mclmcrrt8_1.dll. Please install the correct version of the MCR. Contact
> your vendor if you do not have an installer for the MCR."
>
> I tried downloading MCR 8.1 from Matlab's site but I get the same error
> everytime and when I check the version that Matlab is using it says 9 which
> corresponds to Matlab 2015b that I'm using.
>
> I'll also try with the source code to see if it possible to make it work
> but I'm looking at possibly building a pipeline in Fiji to do batch
> processing so I'm also looking at possible plugins and/or macros.
>
> Thanks for your help !
>
> *------------------
> *
>
> *Laurent Guerard*
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Ignacio Arganda-Carreras, Ph.D.
Ikerbasque Research Fellow
Departamento de Ciencia de la Computacion e Inteligencia Artificial
Facultad de Informatica, Universidad del Pais Vasco
Paseo de Manuel Lardizabal, 1
20018 Donostia-San Sebastian
Guipuzcoa, Spain

Phone : [phone]
Website: http://sites.google.com/site/iargandacarreras/
<http://biocomp.cnb.csic.es/~iarganda/index_EN.html>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html