Segmentation of confluent cells

[PEARSON Matthew]

Hi all,

I've been asked whether its possible to measure the area of individual  
cells in a confluent monolayer autonomously.  As far as i'm aware this  
would be near impossible with brightfield techniques so i have  
suggested the cells are fixed and stained but i was wondering which  
structures would be best to label to give us the best chance of being  
able to do the analysis using ImageJ.  I was thinking the nucleus and  
some kind of cell junction marker would be best.  I'm not sure a  
cytoplasmic marker would add much if the cells are all touching.

As far as analysis goes, my knowledge of segmentation and separating  
objects goes as far as the basic thresholding and morphological  
operators erosion, dilation, watershed etc.  Are there any good  
specific tools/plugins for this kind of work?  And are there other  
more advanced methods of segmentation of separating touching objects  
available? Perhaps something like level sets would be good for this? I  
think the basis of my approach would be nucleus identification to  
pinpoint a single cell, cell junction enhancement and final  
segmentation of some kind.

Thanks for the advice,

Matt












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

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

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

[dscho]

Hi Matthew,

On Mon, 19 May 2014, Matthew Pearson wrote:

> I've been asked whether its possible to measure the area of individual cells
> in a confluent monolayer autonomously.

You will find that other scientists have a much easier time helping you if
you provide them with example images.

Ciao,
Johannes

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

[Richard Mort-2]

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

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

[PEARSON Matthew]

Hi Johannes,

I'm getting ahead of myself, the cells haven't been prepared for  
imaging yet.  I was asking pre-experiment in case there is anything we  
could do sample prep-wise to aid the analysis once the imaging is  
completed.  What i don't want is for the user to prepare the cells and  
then it turn out that we can't do anything with them, rubbish in,  
rubbish out as they say and its always good to think ahead.  I can  
post some images when we get to that stage.

Thanks,

Matt







On 19 May 2014, at 13:54, Johannes Schindelin wrote:

> Hi Matthew,
>
> On Mon, 19 May 2014, Matthew Pearson wrote:
>
>> I've been asked whether its possible to measure the area of  
>> individual cells
>> in a confluent monolayer autonomously.
>
> You will find that other scientists have a much easier time helping  
> you if
> you provide them with example images.
>
> Ciao,
> Johannes
>

--
Matt Pearson
Microscopy and Imaging Facility
MRC Human Genetics Unit
IGMM
University of Edinburgh
Crewe Road
Edinburgh
EH4 2XU

Tel: [phone]







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

[Cammer, Michael]

Depends 1. on what you want to do with the segmentation and 2.) whether it really has to be automated.  In the first case, maybe labeling the nuclei would be enough to mark which cell is which.  But if you want to see precise edges of cells, this won't work.  So 2.) sometimes labeled cells, such as constitutive GFP or CSFE or cells with the membranes labeled can be separated, or for confluent cells  maybe a protein at cell-cell contacts.  But in most cases, not.  Which brings up how important the experiment is and how lazy the researchers are.  Sometimes tracing is the way to go if you want the answer.   If tracing 100 cells is too difficult for someone, maybe research is the wrong field to be in...
===========================================================================
Michael Cammer, Microscopy Core & Dustin Lab , Skirball Institute, NYU Langone Medical Center
Cell:  914-309-3270   Lab: 212-263-3208
http://ocs.med.nyu.edu/microscopy & http://www.med.nyu.edu/skirball-lab/dustinlab/


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Matthew Pearson
Sent: Monday, May 19, 2014 11:11 AM
To: [hidden email]
Subject: Re: Segmentation of confluent cells

Hi Johannes,

I'm getting ahead of myself, the cells haven't been prepared for imaging yet.  I was asking pre-experiment in case there is anything we could do sample prep-wise to aid the analysis once the imaging is completed.  What i don't want is for the user to prepare the cells and then it turn out that we can't do anything with them, rubbish in, rubbish out as they say and its always good to think ahead.  I can post some images when we get to that stage.

Thanks,

Matt







On 19 May 2014, at 13:54, Johannes Schindelin wrote:

> Hi Matthew,
>
> On Mon, 19 May 2014, Matthew Pearson wrote:
>
>> I've been asked whether its possible to measure the area of
>> individual cells in a confluent monolayer autonomously.
>
> You will find that other scientists have a much easier time helping
> you if you provide them with example images.
>
> Ciao,
> Johannes
>

--
Matt Pearson
Microscopy and Imaging Facility
MRC Human Genetics Unit
IGMM
University of Edinburgh
Crewe Road
Edinburgh
EH4 2XU

Tel: [phone]







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

------------------------------------------------------------
This email message, including any attachments, is for the sole use of the intended recipient(s) and may contain information that is proprietary, confidential, and exempt from disclosure under applicable law. Any unauthorized review, use, disclosure, or distribution is prohibited. If you have received this email in error please notify the sender by return email and delete the original message. Please note, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.
=================================

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

[PEARSON Matthew]

Hi Michael,

I suspect what we'll find is that the segmentation of the cell  
boundary is not accurate enough because they'll all overlap therefore  
how do you distinguish one from the other in which case it will be a  
manual tracing task.  I'll leave that for the researcher to decide if  
its worth it or not.  I'll wait to see how the images come out first.

Thanks,

Matt
On 19 May 2014, at 16:27, Cammer, Michael wrote:

> Depends 1. on what you want to do with the segmentation and 2.)  
> whether it really has to be automated.  In the first case, maybe  
> labeling the nuclei would be enough to mark which cell is which.  
> But if you want to see precise edges of cells, this won't work.  So  
> 2.) sometimes labeled cells, such as constitutive GFP or CSFE or  
> cells with the membranes labeled can be separated, or for confluent  
> cells  maybe a protein at cell-cell contacts.  But in most cases,  
> not.  Which brings up how important the experiment is and how lazy  
> the researchers are.  Sometimes tracing is the way to go if you want  
> the answer.   If tracing 100 cells is too difficult for someone,  
> maybe research is the wrong field to be in...
> =
> =
> =
> =
> =
> ======================================================================
> Michael Cammer, Microscopy Core & Dustin Lab , Skirball Institute,  
> NYU Langone Medical Center
> Cell:  914-309-3270   Lab: 212-263-3208
> http://ocs.med.nyu.edu/microscopy & http://www.med.nyu.edu/skirball-lab/dustinlab/
>
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf  
> Of Matthew Pearson
> Sent: Monday, May 19, 2014 11:11 AM
> To: [hidden email]
> Subject: Re: Segmentation of confluent cells
>
> Hi Johannes,
>
> I'm getting ahead of myself, the cells haven't been prepared for  
> imaging yet.  I was asking pre-experiment in case there is anything  
> we could do sample prep-wise to aid the analysis once the imaging is  
> completed.  What i don't want is for the user to prepare the cells  
> and then it turn out that we can't do anything with them, rubbish  
> in, rubbish out as they say and its always good to think ahead.  I  
> can post some images when we get to that stage.
>
> Thanks,
>
> Matt
>
>
>
>
>
>
>
> On 19 May 2014, at 13:54, Johannes Schindelin wrote:
>
>> Hi Matthew,
>>
>> On Mon, 19 May 2014, Matthew Pearson wrote:
>>
>>> I've been asked whether its possible to measure the area of
>>> individual cells in a confluent monolayer autonomously.
>>
>> You will find that other scientists have a much easier time helping
>> you if you provide them with example images.
>>
>> Ciao,
>> Johannes
>>
>
>
>
>
>
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> ------------------------------------------------------------
> This email message, including any attachments, is for the sole use  
> of the intended recipient(s) and may contain information that is  
> proprietary, confidential, and exempt from disclosure under  
> applicable law. Any unauthorized review, use, disclosure, or  
> distribution is prohibited. If you have received this email in error  
> please notify the sender by return email and delete the original  
> message. Please note, the recipient should check this email and any  
> attachments for the presence of viruses. The organization accepts no  
> liability for any damage caused by any virus transmitted by this  
> email.
> =================================
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

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

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

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

[fabrice senger-2]

Hi,

I would try a membrane marker : wheat germ agglutinins or DiD lypophilic
stain (works with living cells...)

Fabrice.


2014-05-19 13:14 GMT+02:00 Matthew Pearson <[hidden email]>:

> Hi all,
>
> I've been asked whether its possible to measure the area of individual
> cells in a confluent monolayer autonomously.  As far as i'm aware this
> would be near impossible with brightfield techniques so i have suggested
> the cells are fixed and stained but i was wondering which structures would
> be best to label to give us the best chance of being able to do the
> analysis using ImageJ.  I was thinking the nucleus and some kind of cell
> junction marker would be best.  I'm not sure a cytoplasmic marker would add
> much if the cells are all touching.
>
> As far as analysis goes, my knowledge of segmentation and separating
> objects goes as far as the basic thresholding and morphological operators
> erosion, dilation, watershed etc.  Are there any good specific
> tools/plugins for this kind of work?  And are there other more advanced
> methods of segmentation of separating touching objects available? Perhaps
> something like level sets would be good for this? I think the basis of my
> approach would be nucleus identification to pinpoint a single cell, cell
> junction enhancement and final segmentation of some kind.
>
> Thanks for the advice,
>
> Matt
>
>
>
>
>
>
>
>
>
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
>

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

[Mike Esterman]

If you stain the nucleus and the cell membrane - there are a number of fluorescent stains for the cell membrane- then locate the nucleus of each cell and use a region growing algorithm until you encounter the cell membrane of that cell.  Since the cells are varied in shape you will have to do this in various directions from the nucleus.  There is a program that does this but is very expensive for a license - it is eCognition. If you or someone in your group understands image processing and analysis you could probably find references to region growing algorithms and write one within ImageJ.
 
Mike


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Fabrice Senger
Sent: Tuesday, May 20, 2014 11:52 AM
To: [hidden email]
Subject: Re: Segmentation of confluent cells

Hi,

I would try a membrane marker : wheat germ agglutinins or DiD lypophilic stain (works with living cells...)

Fabrice.


2014-05-19 13:14 GMT+02:00 Matthew Pearson <[hidden email]>:

> Hi all,
>
> I've been asked whether its possible to measure the area of individual
> cells in a confluent monolayer autonomously.  As far as i'm aware this
> would be near impossible with brightfield techniques so i have
> suggested the cells are fixed and stained but i was wondering which
> structures would be best to label to give us the best chance of being
> able to do the analysis using ImageJ.  I was thinking the nucleus and
> some kind of cell junction marker would be best.  I'm not sure a
> cytoplasmic marker would add much if the cells are all touching.
>
> As far as analysis goes, my knowledge of segmentation and separating
> objects goes as far as the basic thresholding and morphological
> operators erosion, dilation, watershed etc.  Are there any good
> specific tools/plugins for this kind of work?  And are there other
> more advanced methods of segmentation of separating touching objects
> available? Perhaps something like level sets would be good for this? I
> think the basis of my approach would be nucleus identification to
> pinpoint a single cell, cell junction enhancement and final segmentation of some kind.
>
> Thanks for the advice,
>
> Matt
>
>
>
>
>
>
>
>
>
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
>

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

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

[Kenneth Sloan]

The devil is in the details.  IF the cell walls are absolutely clean, then this has a chance of working.  Even then, you may need a scene analysis layer to apply knowledge well beyond simple nucleus detection and region growing.  Especially if some of the images (always the crucial ones) are "atypical".

The bottom line is that every preparation and every experimental condition will require one of:
A) very sophisticated custom computer vision techniques, or
B) a human trained observer assisted by a custom program to do the bookkeeping and provide support for editing, or
C) low standards ("any cockamamie scheme can get 85% right answers; 90% is cutting edge work; 95% is a career, and 98% is a pipe dream")

When the customer's research depends on 99% correct measurements, spend your time on ease of editing and don't bet the farm on a fully automatic system.

-Kenneth Sloan
(von meinem iPhone4S gesendet)

> On May 20, 2014, at 13:29, Mike Esterman <[hidden email]> wrote:
>
> If you stain the nucleus and the cell membrane - there are a number of fluorescent stains for the cell membrane- then locate the nucleus of each cell and use a region growing algorithm until you encounter the cell membrane of that cell.  Since the cells are varied in shape you will have to do this in various directions from the nucleus.  There is a program that does this but is very expensive for a license - it is eCognition. If you or someone in your group understands image processing and analysis you could probably find references to region growing algorithms and write one within ImageJ.
>
> Mike
>
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Fabrice Senger
> Sent: Tuesday, May 20, 2014 11:52 AM
> To: [hidden email]
> Subject: Re: Segmentation of confluent cells
>
> Hi,
>
> I would try a membrane marker : wheat germ agglutinins or DiD lypophilic stain (works with living cells...)
>
> Fabrice.
>
>
> 2014-05-19 13:14 GMT+02:00 Matthew Pearson <[hidden email]>:
>
>> Hi all,
>>
>> I've been asked whether its possible to measure the area of individual
>> cells in a confluent monolayer autonomously.  As far as i'm aware this
>> would be near impossible with brightfield techniques so i have
>> suggested the cells are fixed and stained but i was wondering which
>> structures would be best to label to give us the best chance of being
>> able to do the analysis using ImageJ.  I was thinking the nucleus and
>> some kind of cell junction marker would be best.  I'm not sure a
>> cytoplasmic marker would add much if the cells are all touching.
>>
>> As far as analysis goes, my knowledge of segmentation and separating
>> objects goes as far as the basic thresholding and morphological
>> operators erosion, dilation, watershed etc.  Are there any good
>> specific tools/plugins for this kind of work?  And are there other
>> more advanced methods of segmentation of separating touching objects
>> available? Perhaps something like level sets would be good for this? I
>> think the basis of my approach would be nucleus identification to
>> pinpoint a single cell, cell junction enhancement and final segmentation of some kind.
>>
>> Thanks for the advice,
>>
>> Matt
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>> The University of Edinburgh is a charitable body, registered in
>> Scotland, with registration number SC005336.
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

[PEARSON Matthew]

Hi guys,

Thanks for your responses.  I'll look into region growing algorithms  
used in ImageJ.  I now a least know that this isn't particularly  
straight forward.  I'll make sure i'm there for the image capture so  
the images can be as good as possible and we'll take it from there.

cheers,

Matt


On 20 May 2014, at 20:12, Kenneth Sloan wrote:

> The devil is in the details.  IF the cell walls are absolutely  
> clean, then this has a chance of working.  Even then, you may need a  
> scene analysis layer to apply knowledge well beyond simple nucleus  
> detection and region growing.  Especially if some of the images  
> (always the crucial ones) are "atypical".
>
> The bottom line is that every preparation and every experimental  
> condition will require one of:
> A) very sophisticated custom computer vision techniques, or
> B) a human trained observer assisted by a custom program to do the  
> bookkeeping and provide support for editing, or
> C) low standards ("any cockamamie scheme can get 85% right answers;  
> 90% is cutting edge work; 95% is a career, and 98% is a pipe dream")
>
> When the customer's research depends on 99% correct measurements,  
> spend your time on ease of editing and don't bet the farm on a fully  
> automatic system.
>
> -Kenneth Sloan
> (von meinem iPhone4S gesendet)
>
>> On May 20, 2014, at 13:29, Mike Esterman <[hidden email]> wrote:
>>
>> If you stain the nucleus and the cell membrane - there are a number  
>> of fluorescent stains for the cell membrane- then locate the  
>> nucleus of each cell and use a region growing algorithm until you  
>> encounter the cell membrane of that cell.  Since the cells are  
>> varied in shape you will have to do this in various directions from  
>> the nucleus.  There is a program that does this but is very  
>> expensive for a license - it is eCognition. If you or someone in  
>> your group understands image processing and analysis you could  
>> probably find references to region growing algorithms and write one  
>> within ImageJ.
>>
>> Mike
>>
>>
>> -----Original Message-----
>> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf  
>> Of Fabrice Senger
>> Sent: Tuesday, May 20, 2014 11:52 AM
>> To: [hidden email]
>> Subject: Re: Segmentation of confluent cells
>>
>> Hi,
>>
>> I would try a membrane marker : wheat germ agglutinins or DiD  
>> lypophilic stain (works with living cells...)
>>
>> Fabrice.
>>
>>
>> 2014-05-19 13:14 GMT+02:00 Matthew Pearson <[hidden email]
>> >:
>>
>>> Hi all,
>>>
>>> I've been asked whether its possible to measure the area of  
>>> individual
>>> cells in a confluent monolayer autonomously.  As far as i'm aware  
>>> this
>>> would be near impossible with brightfield techniques so i have
>>> suggested the cells are fixed and stained but i was wondering which
>>> structures would be best to label to give us the best chance of  
>>> being
>>> able to do the analysis using ImageJ.  I was thinking the nucleus  
>>> and
>>> some kind of cell junction marker would be best.  I'm not sure a
>>> cytoplasmic marker would add much if the cells are all touching.
>>>
>>> As far as analysis goes, my knowledge of segmentation and separating
>>> objects goes as far as the basic thresholding and morphological
>>> operators erosion, dilation, watershed etc.  Are there any good
>>> specific tools/plugins for this kind of work?  And are there other
>>> more advanced methods of segmentation of separating touching objects
>>> available? Perhaps something like level sets would be good for  
>>> this? I
>>> think the basis of my approach would be nucleus identification to
>>> pinpoint a single cell, cell junction enhancement and final  
>>> segmentation of some kind.
>>>
>>> Thanks for the advice,
>>>
>>> Matt
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>>
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

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

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

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

[Cammer, Michael]

When I mentioned cell-cell junction proteins, I was thinking of something like e-cadherin.
The red in https://www.flickr.com/photos/mcammer/2234624761/

===========================================================================
Michael Cammer, Microscopy Core & Dustin Lab , Skirball Institute, NYU Langone Medical Center
Cell:  914-309-3270   Lab: 212-263-3208
http://ocs.med.nyu.edu/microscopy & http://www.med.nyu.edu/skirball-lab/dustinlab/

From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Matthew Pearson
Sent: Wednesday, May 21, 2014 7:40 AM
To: [hidden email]
Subject: Re: Segmentation of confluent cells

Hi guys,

Thanks for your responses.  I'll look into region growing algorithms
used in ImageJ.  I now a least know that this isn't particularly
straight forward.  I'll make sure i'm there for the image capture so
the images can be as good as possible and we'll take it from there.

cheers,

Matt

------------------------------------------------------------
This email message, including any attachments, is for the sole use of the intended recipient(s) and may contain information that is proprietary, confidential, and exempt from disclosure under applicable law. Any unauthorized review, use, disclosure, or distribution is prohibited. If you have received this email in error please notify the sender by return email and delete the original message. Please note, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.
=================================


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

[Unruh, Jay-2]

Note that for many confluent cell layers, the cells extend beneath one another to a fairly significant extent.  The "boundary" will underestimate the area by a fairly significant amount.  Given this uncertainty and the difficulty in segmenting the boundaries anyway, you are probably better off simply using the distance to the neighboring nuclei as an estimate of cell size.  Voronoi analysis will give you a slightly more sophisticated version of this.

Jay

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Cammer, Michael
Sent: Wednesday, May 21, 2014 8:49 AM
To: [hidden email]
Subject: Re: Segmentation of confluent cells

When I mentioned cell-cell junction proteins, I was thinking of something like e-cadherin.
The red in https://www.flickr.com/photos/mcammer/2234624761/

===========================================================================
Michael Cammer, Microscopy Core & Dustin Lab , Skirball Institute, NYU Langone Medical Center
Cell:  914-309-3270   Lab: 212-263-3208
http://ocs.med.nyu.edu/microscopy & http://www.med.nyu.edu/skirball-lab/dustinlab/

From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Matthew Pearson
Sent: Wednesday, May 21, 2014 7:40 AM
To: [hidden email]
Subject: Re: Segmentation of confluent cells

Hi guys,

Thanks for your responses.  I'll look into region growing algorithms used in ImageJ.  I now a least know that this isn't particularly straight forward.  I'll make sure i'm there for the image capture so the images can be as good as possible and we'll take it from there.

cheers,

Matt

------------------------------------------------------------
This email message, including any attachments, is for the sole use of the intended recipient(s) and may contain information that is proprietary, confidential, and exempt from disclosure under applicable law. Any unauthorized review, use, disclosure, or distribution is prohibited. If you have received this email in error please notify the sender by return email and delete the original message. Please note, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.
=================================


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

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

[Kenneth Sloan]

Voronoi analysis is very powerful (it’s what I have used almost exclusively in
analysis of en face images of human retina; we’ve been doing this for .. awhile...).

The one caveat is that you must understand that the Voronoi polygons are not IDENTICAL
to the cell boundaries, and if your subsequent analysis depends critically on
what is happening near the boundaries, you may be introducing an extra
source of variability.

And…especially with the “interesting” cases, the cells may not be convex…

And…the nuclei may not be particularly close to the “center” of the cell…

But, again - that’s what we do with (for example) RPE cells in human retina.
We stain the cytoskeleton, semi-automatically identify the “center” of each cell,
allow a trained observer to correct, extend, or completely re-do the “findCenters” step,
and then construct a Voronoi Diagram.  The Voronoi polygons are then used as surrogates for
the cells.

[beware of edge effects - we ask a trained observer to draw a large polygon (not necessarily convex)
which encloses the “good” portion of a field of view.  The Voronoi Diagram involves ALL cell centers found,
but subsequent analysis uses only the Voronoi polygons which lie entirely inside the “good” region.  In simpler
cases, all you need to do is to shrink the field of view and analyze only the central region.]

In normal tissue, this works great.  In diseased tissue, there are issues that are typically resolved by
having someone closely examine the results to be sure that the Voronoi Diagram is truly representative of
the actual cells.  In some cases (we haven’t gotten there, yet) it *might* be necessary to actually
trace the cell boundaries (and use a somewhat more complicated data structure to hold the result).

If you do this, take care to identify the CENTERS (defined in a fairly loose way) of the cells - this is not
necessarily the same as the positions of the nuclei.  Be especially careful of measures such as “number of neighbors”,
a measurement which is very sensitive to the precise locations of the Voronoi “sites”.  Do periodic, systematic reviews
of the Voronoi Diagrams and the original images to be sure that the VD is capturing what you think it is.
Cell density is no problem.  Mean values of cell area or distance to neighbors are no problem, but there may
be some noise in variance.  # of neighbors requires extra care (but is so powerful that it’s worth the effort).
If all you need is cell density, you can be much more tolerant of low-precision localization of the cell centers, and
simply detecting nuclei will be fine.  Only…beware of making this decision and THEN LATER having someone
decide to use the VD for more sensitive measurements.  Sometimes, the second level of analysis conveniently forgets, or ignores, or is simply unaware of, the limitations imposed by the initial data collection.  Don’t over-interpret!
--
Kenneth Sloan
[hidden email]
"La lutte elle-même vers les sommets suffit à remplir un coeur d'homme; il faut imaginer Sisyphe heureux."


On May 21, 2014, at 08:54 , Unruh, Jay <[hidden email]> wrote:

> Note that for many confluent cell layers, the cells extend beneath one another to a fairly significant extent.  The "boundary" will underestimate the area by a fairly significant amount.  Given this uncertainty and the difficulty in segmenting the boundaries anyway, you are probably better off simply using the distance to the neighboring nuclei as an estimate of cell size.  Voronoi analysis will give you a slightly more sophisticated version of this.
>
> Jay
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Cammer, Michael
> Sent: Wednesday, May 21, 2014 8:49 AM
> To: [hidden email]
> Subject: Re: Segmentation of confluent cells
>
> When I mentioned cell-cell junction proteins, I was thinking of something like e-cadherin.
> The red in https://www.flickr.com/photos/mcammer/2234624761/
>
> ===========================================================================
> Michael Cammer, Microscopy Core & Dustin Lab , Skirball Institute, NYU Langone Medical Center
> Cell:  914-309-3270   Lab: 212-263-3208
> http://ocs.med.nyu.edu/microscopy & http://www.med.nyu.edu/skirball-lab/dustinlab/
>
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Matthew Pearson
> Sent: Wednesday, May 21, 2014 7:40 AM
> To: [hidden email]
> Subject: Re: Segmentation of confluent cells
>
> Hi guys,
>
> Thanks for your responses.  I'll look into region growing algorithms used in ImageJ.  I now a least know that this isn't particularly straight forward.  I'll make sure i'm there for the image capture so the images can be as good as possible and we'll take it from there.
>
> cheers,
>
> Matt
>
> ------------------------------------------------------------
> This email message, including any attachments, is for the sole use of the intended recipient(s) and may contain information that is proprietary, confidential, and exempt from disclosure under applicable law. Any unauthorized review, use, disclosure, or distribution is prohibited. If you have received this email in error please notify the sender by return email and delete the original message. Please note, the recipient should check this email and any attachments for the presence of viruses. The organization accepts no liability for any damage caused by any virus transmitted by this email.
> =================================
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

[Dimiter Prodanov (imec)]

Hi,

I agree with Mr. Unruh. Segmentation of confluent cells is not trivial.
So If you want only area measure Voronoi analysis is a good alternative.
If you need only cell number/density then nuclear identification is sufficient.

Best regards,

Dimiter Prodanov

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