ImageJ

(no subject)

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

7 messages Options

Ash Dunne

(no subject)

Hi all,

Apologies if this is very simple to those in the know.

I am trying to threshold the attached image so that I can attempt to segment and count the nuclei but I am struggling with it. Can anybody help with a method that might provide some success?! I have dozens more like this that I'm hoping to process, so I'm very keen to find out how it is done in a more complicated scenario than a nicely separated handful of cells.

Many thanks,

Ash
--------------------------------------------------------
This email is confidential and intended solely for the use of the person(s) ('the intended recipient') to whom it was addressed. Any views or opinions presented are solely those of the author and do not necessarily represent those of the Paterson Institute for Cancer Research or the University of Manchester. It may contain information that is privileged & confidential within the meaning of applicable law. Accordingly any dissemination, distribution, copying, or other use of this message, or any of its contents, by any person other than the intended recipient may constitute a breach of civil or criminal law and is strictly prohibited. If you are NOT the intended recipient please contact the sender and dispose of this e-mail as soon as possible.

multiple cell count.tif (366K) Download Attachment

Cammer, Michael

(no subject)

The overall intensity varies at a low frequency across the field but the objects you want to isolate are approx 14 pixels in diameter and locally stand out from their background. Try FFT bandpass filter of 15-0 or 16-1 (or 15-1 or 16-0), maybe two iterations, and then try to threshold with Huang or IsoData.

________________________________________________________
Michael Cammer, Assistant Research Scientist
Skirball Institute of Biomolecular Medicine
Lab: (212) 263-3208 Cell: (914) 309-3270

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Achille Dunne
Sent: Thursday, February 02, 2012 12:39 PM
To: [hidden email]
Subject:

Hi all,

Apologies if this is very simple to those in the know.

I am trying to threshold the attached image so that I can attempt to segment and count the nuclei but I am struggling with it. Can anybody help with a method that might provide some success?! I have dozens more like this that I'm hoping to process, so I'm very keen to find out how it is done in a more complicated scenario than a nicely separated handful of cells.

Many thanks,

Ash
--------------------------------------------------------
This email is confidential and intended solely for the use of the person(s) ('the intended recipient') to whom it was addressed. Any views or opinions presented are solely those of the author and do not necessarily represent those of the Paterson Institute for Cancer Research or the University of Manchester. It may contain information that is privileged & confidential within the meaning of applicable law. Accordingly any dissemination, distribution, copying, or other use of this message, or any of its contents, by any person other than the intended recipient may constitute a breach of civil or criminal law and is strictly prohibited. If you are NOT the intended recipient please contact the sender and dispose of this e-mail as soon as possible.

------------------------------------------------------------
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.
=================================

Citron, Bruce

Re: nuclei counting

In reply to this post by Ash Dunne

That image may be impossible. The problem is that in the central area,
between the nuclei, the intensity is higher than the max intensity of nuclei
near the periphery. If it was more uniform, then watershed segmentation
could be tried or if your nuclear areas are constant, as they appear to be,
you could measure the thresholded area and divide by the area per nucleus.
But neither seem workable for this, but hopefully someone else will have a
better idea.
-Bruce
--
Bruce A. Citron, Ph.D., Bay Pines VA

> On 2/2/12 12:38 , "Achille Dunne" <[hidden email]> wrote:
> Hi all,
>
> Apologies if this is very simple to those in the know.
>
> I am trying to threshold the attached image so that I can attempt to segment
> and count the nuclei but I am struggling with it. Can anybody help with a
> method that might provide some success?! I have dozens more like this that I'm
> hoping to process, so I'm very keen to find out how it is done in a more
> complicated scenario than a nicely separated handful of cells.
>
> Many thanks,
>
> Ash

Olivier Burri

Re: nuclei counting

Dear Ash,

With images like these, you're always going to have some level of inaccuracy due to the blur in very dense areas.
One solution would be to use a stereology tool to count certain regions by hand. Unfortunately, I am not aware of stereology plugins for ImageJ.

I've tried using BG subtraction, CLAHE, Laplacian of Gaussian, and then detecting the local maxima from the image, with some success. You'll detect fewer nuclei than through a manual count and you should quantify this difference (By comparing it to a manual count) to see whether this accuracy is enough for you.

http://documents.epfl.ch/users/o/ob/oburri/public/Multi-Cell-Count.tif

Steps are as follows:
1. BG Subtraction, rolling ball radius 100px
2. Enhance Local Contrast, std settings
3. Feature Extraction -> Laplacian, Sigma = 3.0
4. Find Maxima: White BG, tolerance = 0.75 output segmented particles.

Let me know if this helped.

Best

Oli

Olivier Burri
Engineer, Development in Image Processing
BioImaging and Optics platform (PTBIOP)
Tel: [+4121 69] 39629

________________________________________
From: ImageJ Interest Group [[hidden email]] on behalf of Bruce Citron [[hidden email]]
Sent: Thursday, February 02, 2012 9:32 PM
To: [hidden email]
Subject: Re: nuclei counting

That image may be impossible. The problem is that in the central area,
between the nuclei, the intensity is higher than the max intensity of nuclei
near the periphery. If it was more uniform, then watershed segmentation
could be tried or if your nuclear areas are constant, as they appear to be,
you could measure the thresholded area and divide by the area per nucleus.
But neither seem workable for this, but hopefully someone else will have a
better idea.
-Bruce
--
Bruce A. Citron, Ph.D., Bay Pines VA

Michael Schmid

Re: nuclei counting

Hi Ash,

after playing a bit, I found a method to level the image, but it is far too crowded even for manual counting (image processing in our brains is the result of millions of years of a massively parallel evolutionary optimization algorithm, probably better than the best image processing program on a computer ;-)

run("32-bit");
run("Square Root"); //dark parts of the images have less contrast; this compensates for it a bit
run("Enhance Contrast", "saturated=0.35"); //only to see what's going on
run("Subtract Background...", "rolling=200 sliding");
run("Duplicate...", "title=[multiple cell count-1.tif]"); //create a copy for normalization
run("Median...", "radius=2");
run("Maximum...", "radius=2");
run("Minimum...", "radius=1");
run("Gaussian Blur...", "sigma=2");
run("Subtract Background...", "rolling=200 light create sliding");
run("Gaussian Blur...", "sigma=5");
imageCalculator("Divide create 32-bit", "multiple cell count.tif","multiple cell count-1.tif");

you can try to threshold the result and run watershed on the binary image, but the crowded areas won't be split correctly.

Michael
________________________________________________________________
On Feb 3, 2012, at 09:36, Burri Olivier wrote:

> Dear Ash,
>
> With images like these, you're always going to have some level of inaccuracy due to the blur in very dense areas.
> One solution would be to use a stereology tool to count certain regions by hand. Unfortunately, I am not aware of stereology plugins for ImageJ.
>
> I've tried using BG subtraction, CLAHE, Laplacian of Gaussian, and then detecting the local maxima from the image, with some success. You'll detect fewer nuclei than through a manual count and you should quantify this difference (By comparing it to a manual count) to see whether this accuracy is enough for you.
>
> http://documents.epfl.ch/users/o/ob/oburri/public/Multi-Cell-Count.tif
>
> Steps are as follows:
> 1. BG Subtraction, rolling ball radius 100px
> 2. Enhance Local Contrast, std settings
> 3. Feature Extraction -> Laplacian, Sigma = 3.0
> 4. Find Maxima: White BG, tolerance = 0.75 output segmented particles.
>
> Let me know if this helped.
>
> Best
>
> Oli
>
> Olivier Burri
> Engineer, Development in Image Processing
> BioImaging and Optics platform (PTBIOP)
> Tel: [+4121 69] 39629
>
> ________________________________________
> From: ImageJ Interest Group [[hidden email]] on behalf of Bruce Citron [[hidden email]]
> Sent: Thursday, February 02, 2012 9:32 PM
> To: [hidden email]
> Subject: Re: nuclei counting
>
> That image may be impossible. The problem is that in the central area,
> between the nuclei, the intensity is higher than the max intensity of nuclei
> near the periphery. If it was more uniform, then watershed segmentation
> could be tried or if your nuclear areas are constant, as they appear to be,
> you could measure the thresholded area and divide by the area per nucleus.
> But neither seem workable for this, but hopefully someone else will have a
> better idea.
> -Bruce
> --
> Bruce A. Citron, Ph.D., Bay Pines VA
>
>
>> On 2/2/12 12:38 , "Achille Dunne" <[hidden email]> wrote:
>> Hi all,
>>
>> Apologies if this is very simple to those in the know.
>>
>> I am trying to threshold the attached image so that I can attempt to segment
>> and count the nuclei but I am struggling with it. Can anybody help with a
>> method that might provide some success?! I have dozens more like this that I'm
>> hoping to process, so I'm very keen to find out how it is done in a more
>> complicated scenario than a nicely separated handful of cells.
>>
>> Many thanks,
>>
>> Ash

Joel Sheffield

Re: nuclei counting

There are, indeed, many problems with this images. Obviously, the uneven
background is a major problem. So is the degree of overlap of the nuclei,
and variations in depth. Furthermore, the nuclei are not themselves
uniform structures. They have a dark periphery and a central nucleolus,
with a light matrix. This adds to the complexity of thresholding.

I too have played around with flattening the field. I find that a
combination of FFT Bandwidth and CLAHE (with a small box size) does bring
most of the nuclei into a threshold range, but then, as others have noted,
there is too much overlap of individual nuclei. The image that I get has
too much bleeding between adjacent nuclei, and then the watershed algorithm
breaks them into pieces that are too small.

So, the obvious recommendation is to get a better preparation, and use
better illumination.

Joel

On Fri, Feb 3, 2012 at 12:48 PM, Michael Schmid <[hidden email]>wrote:

> Hi Ash,
>
> after playing a bit, I found a method to level the image, but it is far
> too crowded even for manual counting (image processing in our brains is the
> result of millions of years of a massively parallel evolutionary
> optimization algorithm, probably better than the best image processing
> program on a computer ;-)
>
>
> run("32-bit");
> run("Square Root"); //dark parts of the images have less contrast; this
> compensates for it a bit
> run("Enhance Contrast", "saturated=0.35"); //only to see what's going on
> run("Subtract Background...", "rolling=200 sliding");
> run("Duplicate...", "title=[multiple cell count-1.tif]"); //create a copy
> for normalization
> run("Median...", "radius=2");
> run("Maximum...", "radius=2");
> run("Minimum...", "radius=1");
> run("Gaussian Blur...", "sigma=2");
> run("Subtract Background...", "rolling=200 light create sliding");
> run("Gaussian Blur...", "sigma=5");
> imageCalculator("Divide create 32-bit", "multiple cell
> count.tif","multiple cell count-1.tif");
>
> you can try to threshold the result and run watershed on the binary image,
> but the crowded areas won't be split correctly.
>
> Michael
> ________________________________________________________________
> On Feb 3, 2012, at 09:36, Burri Olivier wrote:
>
> > Dear Ash,
> >
> > With images like these, you're always going to have some level of
> inaccuracy due to the blur in very dense areas.
> > One solution would be to use a stereology tool to count certain regions
> by hand. Unfortunately, I am not aware of stereology plugins for ImageJ.
> >
> > I've tried using BG subtraction, CLAHE, Laplacian of Gaussian, and then
> detecting the local maxima from the image, with some success. You'll detect
> fewer nuclei than through a manual count and you should quantify this
> difference (By comparing it to a manual count) to see whether this accuracy
> is enough for you.
> >
> > http://documents.epfl.ch/users/o/ob/oburri/public/Multi-Cell-Count.tif
> >
> > Steps are as follows:
> > 1. BG Subtraction, rolling ball radius 100px
> > 2. Enhance Local Contrast, std settings
> > 3. Feature Extraction -> Laplacian, Sigma = 3.0
> > 4. Find Maxima: White BG, tolerance = 0.75 output segmented particles.
> >
> > Let me know if this helped.
> >
> > Best
> >
> > Oli
> >
> > Olivier Burri
> > Engineer, Development in Image Processing
> > BioImaging and Optics platform (PTBIOP)
> > Tel: [+4121 69] 39629
> >
> > ________________________________________
> > From: ImageJ Interest Group [[hidden email]] on behalf of Bruce
> Citron [[hidden email]]
> > Sent: Thursday, February 02, 2012 9:32 PM
> > To: [hidden email]
> > Subject: Re: nuclei counting
> >
> > That image may be impossible. The problem is that in the central area,
> > between the nuclei, the intensity is higher than the max intensity of
> nuclei
> > near the periphery. If it was more uniform, then watershed segmentation
> > could be tried or if your nuclear areas are constant, as they appear to
> be,
> > you could measure the thresholded area and divide by the area per
> nucleus.
> > But neither seem workable for this, but hopefully someone else will have
> a
> > better idea.
> > -Bruce
> > --
> > Bruce A. Citron, Ph.D., Bay Pines VA
> >
> >
> >> On 2/2/12 12:38 , "Achille Dunne" <[hidden email]> wrote:
> >> Hi all,
> >>
> >> Apologies if this is very simple to those in the know.
> >>
> >> I am trying to threshold the attached image so that I can attempt to
> segment
> >> and count the nuclei but I am struggling with it. Can anybody help with
> a
> >> method that might provide some success?! I have dozens more like this
> that I'm
> >> hoping to process, so I'm very keen to find out how it is done in a more
> >> complicated scenario than a nicely separated handful of cells.
> >>
> >> Many thanks,
> >>
> >> Ash
>

--

Joel B. Sheffield, Ph.D
Department of Biology
Temple University
Philadelphia, PA 19122
Voice: 215 204 8839
e-mail: [hidden email]
URL: http://astro.temple.edu/~jbs

Cammer, Michael

Re: nuclei counting

In reply to this post by Michael Schmid

I posted this yesterday but was surprised there were no comments. But I did leave out the last step...
The overall intensity varies at a low frequency across the field but the objects to be isolated are approx 14 pixels in diameter and locally are contrasted from their background. FFT bandpass filter of 15-0 or 16-1, two iterations, threshold with Huang or IsoData. Follow by a watershed, this appears to work well. Choose an appropriate minimum size when analyzing particles.

________________________________________________________
Michael Cammer, Assistant Research Scientist Skirball Institute of Biomolecular Medicine
Lab: (212) 263-3208 Cell: (914) 309-3270

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Michael Schmid
Sent: Friday, February 03, 2012 12:48 PM
To: [hidden email]
Subject: Re: nuclei counting

Hi Ash,

after playing a bit, I found a method to level the image, but it is far too crowded even for manual counting (image processing in our brains is the result of millions of years of a massively parallel evolutionary optimization algorithm, probably better than the best image processing program on a computer ;-)

run("32-bit");
run("Square Root"); //dark parts of the images have less contrast; this compensates for it a bit
run("Enhance Contrast", "saturated=0.35"); //only to see what's going on
run("Subtract Background...", "rolling=200 sliding");
run("Duplicate...", "title=[multiple cell count-1.tif]"); //create a copy for normalization
run("Median...", "radius=2");
run("Maximum...", "radius=2");
run("Minimum...", "radius=1");
run("Gaussian Blur...", "sigma=2");
run("Subtract Background...", "rolling=200 light create sliding");
run("Gaussian Blur...", "sigma=5");
imageCalculator("Divide create 32-bit", "multiple cell count.tif","multiple cell count-1.tif");

you can try to threshold the result and run watershed on the binary image, but the crowded areas won't be split correctly.

Michael
________________________________________________________________
On Feb 3, 2012, at 09:36, Burri Olivier wrote:

------------------------------------------------------------
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.
=================================