background removal in fluorescence

[Martin Wessendorf]

Dear ImageJ folks--

Has anyone come up with a good method for background subtraction in
images obtained using fluorescence microscopy?  In my hands, the methods
that work for brightfield don't work for fluorescence.

Thanks!

Martin
--
Martin Wessendorf, Ph.D.                   office: [phone]
Assoc Prof, Dept Neuroscience                 lab: [phone]
University of Minnesota             Preferred FAX: [phone]
6-145 Jackson Hall, 321 Church St. SE    Dept Fax: [phone]
Minneapolis, MN  55455                E-mail: [hidden email]

[Pedro J CamelloDr Pedro J Camello]

In my opinion (my main experience is ca2+ signals in living cells) the
best method is to acquire a background image in an area of the coverslip
where there are no cells, using it to substract to the series of images.
Sometimes this is difficult if the preparation is too "dense", and if
you have to move too far from the area of interest the plane of focus
could be different. Now I´m removing to the entire image the average
fluorescence from an irregular ROI drawn in the "empty" area surrounding
the cells.

Martin Wessendorf wrote:

> Dear ImageJ folks--
>
> Has anyone come up with a good method for background subtraction in
> images obtained using fluorescence microscopy?  In my hands, the
> methods that work for brightfield don't work for fluorescence.
>
> Thanks!
>
> Martin

[W. Bryan Smith]

yes, here is what i was going to suggest:

While this is not the best method, a
low-tech solution is to acquire 'blank'
images of your sample and use this for
straight subtraction of the background.
If you are doing confocal or multiphoton
imaging, your 'blank' can simply be an
image z-slice in which there is no
relevant pixel data (only noise).  If you
have several such images, you can get an
average and subtract the average from each
frame in the z-series.  If you are using
epifluorescence it may be a bit tricky,
but you should be able to get a decent
'blank' image by acquiring an image
through the same coverslip/dish (whatever
you are using) without any specimen
present.  Again, take a few, get a mean,
and subtract the mean from your real data.
This low-tech solution can work
surprisingly well depending on the
statistics of the background you are
trying to subtract.

bryan

On Tue, 20 Sep 2005, Pedro J Camello wrote:

.' In my opinion (my main experience is ca2+ signals in living cells) the
.' best method is to acquire a background image in an area of the coverslip
.' where there are no cells, using it to substract to the series of images.
.' Sometimes this is difficult if the preparation is too "dense", and if
.' you have to move too far from the area of interest the plane of focus
.' could be different. Now I´m removing to the entire image the average
.' fluorescence from an irregular ROI drawn in the "empty" area surrounding
.' the cells.
.'
.' Martin Wessendorf wrote:
.'
.' > Dear ImageJ folks--
.' >
.' > Has anyone come up with a good method for background subtraction in
.' > images obtained using fluorescence microscopy?  In my hands, the
.' > methods that work for brightfield don't work for fluorescence.
.' >
.' > Thanks!
.' >
.' > Martin
.'
.'

Division of Biology
Broad Center
Caltech 114-96
Pasadena, CA, 91125
ph. [phone]

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The illusion that we are separate from one another
is an optical delusion of our consciousness.
                                -Albert Einstein
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[Tony Collins-2]

I agree with Pedro here. My only other comment is with some older
cameras, the background is not even so subtracting a single value for
the backround is not correct and a 'black-field' subtraction may be the
better way.

If it's a timecourse you may need to take a measure of background for
each timepoint too.

Tony

Pedro J Camello wrote:

> In my opinion (my main experience is ca2+ signals in living cells) the
> best method is to acquire a background image in an area of the coverslip
> where there are no cells, using it to substract to the series of images.
> Sometimes this is difficult if the preparation is too "dense", and if
> you have to move too far from the area of interest the plane of focus
> could be different. Now I´m removing to the entire image the average
> fluorescence from an irregular ROI drawn in the "empty" area surrounding
> the cells.
>
> Martin Wessendorf wrote:
>
>> Dear ImageJ folks--
>>
>> Has anyone come up with a good method for background subtraction in
>> images obtained using fluorescence microscopy?  In my hands, the
>> methods that work for brightfield don't work for fluorescence.
>>
>> Thanks!
>>
>> Martin

--
Tony Collins, Ph.D.
Facility Manager
Wright Cell Imaging Facility
Toronto Western Research Institute
13-407 McLaughlin Pavilion
399 Bathurst Street
Toronto, ON. M5T 2S8
tel. [phone] fax: [phone]
http://www.uhnresearch.ca/wcif

This e-mail may contain confidential and/or privileged information for the sole use of the intended recipient. Any review or distribution by anyone other than the person for whom it was originally intended is strictly prohibited. If you have received this e-mail in error, please contact the sender and delete all copies. Opinions, conclusions or other information contained in this e-mail may not be that of the organization.

[Joel Sheffield]

I would imagine that part of the problem is that you are trying to
remove background signal which is mostly dark.  This means that the
pixel values are close to zero or so.  Image subtraction might cause
the values to drop below zero, which wouldn't be helpful.  Some times
I have found it useful to divide the data image by the background
image, using a 16 or 32 bit resultant image, and then readjust.    

Joel

> Dear ImageJ folks--
>
> Has anyone come up with a good method for background subtraction in
> images obtained using fluorescence microscopy?  In my hands, the
> methods that work for brightfield don't work for fluorescence.
>
> Thanks!
>
> Martin
> --
> Martin Wessendorf, Ph.D.                   office: [phone]
> Assoc Prof, Dept Neuroscience                 lab: [phone]
> University of Minnesota             Preferred FAX: [phone]
> 6-145 Jackson Hall, 321 Church St. SE    Dept Fax: [phone]
> Minneapolis, MN  55455                E-mail: [hidden email]

Joel B. Sheffield, Ph.D.
Biology Department, Temple University
1900 North 12th Street
Philadelphia, PA 19122
[hidden email]  
[phone], fax [phone]
http://astro.temple.edu/~jbs

[Martin Wessendorf]

Dear ImageJ-ers--

I realize that I didn't articulate the problem particularly well.  --We
want to make some montages of images.  Our illumination is not perfectly
even and so parts of the field are illuminated more brightly than
others.  Thus when we try to make the montage, there are abrupt
inconsistencies in the lighting across it.

I had thought that the following *should* work:

1) Take an image of a uniform fluorescent field (image A)

2) Take an image of the specimen (image B)

3) Divide B by A; multiply by the mean intensity of A.

(--I think that this is what Wayne's plugin does.)

However, when we use that method, the inconsistencies of the
illumination are greatly accentuated rather than being reduced.  My
guess is that this is due to non-linearity of response in our CCD camera.

I'd be delighted if anyone has a work-around for this problem!

Martin
--
Martin Wessendorf, Ph.D.                   office: [phone]
Assoc Prof, Dept Neuroscience                 lab: [phone]
University of Minnesota             Preferred FAX: [phone]
6-145 Jackson Hall, 321 Church St. SE    Dept Fax: [phone]
Minneapolis, MN  55455                E-mail: [hidden email]

[Bevic, Jordan]

Have you tried running the Background Subtractor on each separate image
before making the montage?  Although I think there are some minor bugs
in it (the bottom few rows show artifacts on contrast stretching) it
works quite well at flattening out illumination differences.

Jordan Bevic
QuadTech Advanced Vision Systems

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
Martin Wessendorf
Sent: Tuesday, September 20, 2005 1:12 PM
To: [hidden email]
Subject: Re: background removal in fluorescence

Dear ImageJ-ers--

I realize that I didn't articulate the problem particularly well.  --We
want to make some montages of images.  Our illumination is not perfectly

even and so parts of the field are illuminated more brightly than
others.  Thus when we try to make the montage, there are abrupt
inconsistencies in the lighting across it.

I had thought that the following *should* work:

1) Take an image of a uniform fluorescent field (image A)

2) Take an image of the specimen (image B)

3) Divide B by A; multiply by the mean intensity of A.

(--I think that this is what Wayne's plugin does.)

However, when we use that method, the inconsistencies of the
illumination are greatly accentuated rather than being reduced.  My
guess is that this is due to non-linearity of response in our CCD
camera.

I'd be delighted if anyone has a work-around for this problem!

Martin
--
Martin Wessendorf, Ph.D.                   office: [phone]
Assoc Prof, Dept Neuroscience                 lab: [phone]
University of Minnesota             Preferred FAX: [phone]
6-145 Jackson Hall, 321 Church St. SE    Dept Fax: [phone]
Minneapolis, MN  55455                E-mail: [hidden email]

[Prodanov, D. (FYS)]

Dear Martin,

Actually, what the background subtraction does is to use the "rolling ball"

Image is sampled in a ratio and then degraded by a Structuring Element.
then the degraded image is subtracted from the initial image (i.e. "top hat" transform).

Anyway, I think you're on the right track. May be you should do
stack averaging? Or may be you should track the intensity of a praticular structure
and readjust the histograms according to it.

best regards

Dimiter

Date:    Tue, 20 Sep 2005 13:11:52 -0500
From:    Martin Wessendorf <[hidden email]>
Subject: Re: background removal in fluorescence

Dear ImageJ-ers--

I realize that I didn't articulate the problem particularly well.  --We
want to make some montages of images.  Our illumination is not perfectly
even and so parts of the field are illuminated more brightly than
others.  Thus when we try to make the montage, there are abrupt
inconsistencies in the lighting across it.

I had thought that the following *should* work:

1) Take an image of a uniform fluorescent field (image A)

2) Take an image of the specimen (image B)

3) Divide B by A; multiply by the mean intensity of A.

(--I think that this is what Wayne's plugin does.)

However, when we use that method, the inconsistencies of the
illumination are greatly accentuated rather than being reduced.  My
guess is that this is due to non-linearity of response in our CCD camera.

I'd be delighted if anyone has a work-around for this problem!

Martin

[Ryan J Deaton]

Wouldn't you want to just subtract the background image from the target
image?  'B - A' in your example below?  

Ryan Deaton
Data Management Specialist
Preventive Medicine
Northwestern University
ph: 312-503-1979
fax: [phone]

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
Prodanov, D. (FYS)
Sent: Wednesday, September 21, 2005 3:58 AM
To: [hidden email]
Subject: Re: background removal in fluorescence

Dear Martin,

Actually, what the background subtraction does is to use the "rolling ball"

Image is sampled in a ratio and then degraded by a Structuring Element.
then the degraded image is subtracted from the initial image (i.e. "top hat"
transform).

Anyway, I think you're on the right track. May be you should do
stack averaging? Or may be you should track the intensity of a praticular
structure
and readjust the histograms according to it.

best regards

Dimiter

Date:    Tue, 20 Sep 2005 13:11:52 -0500
From:    Martin Wessendorf <[hidden email]>
Subject: Re: background removal in fluorescence

Dear ImageJ-ers--

I realize that I didn't articulate the problem particularly well.  --We
want to make some montages of images.  Our illumination is not perfectly
even and so parts of the field are illuminated more brightly than
others.  Thus when we try to make the montage, there are abrupt
inconsistencies in the lighting across it.

I had thought that the following *should* work:

1) Take an image of a uniform fluorescent field (image A)

2) Take an image of the specimen (image B)

3) Divide B by A; multiply by the mean intensity of A.

(--I think that this is what Wayne's plugin does.)

However, when we use that method, the inconsistencies of the
illumination are greatly accentuated rather than being reduced.  My
guess is that this is due to non-linearity of response in our CCD camera.

I'd be delighted if anyone has a work-around for this problem!

Martin