Confocal Microscopy List

background subtraction

Classic

List

Threaded

6 messages Options

mmodel

background subtraction

Hello,

Does anyone know of a background subtraction method that would act like this:

1) find the entire area occupied by a cell (automatically or manually)
2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.
3) subtract the restored background from the original

Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!

Mike

Julio Vazquez

Re: background subtraction

If you are segmenting your cells and defining a cell mask to measure intensities, you could add a second mask such as a ring or a few pixels width around the cell mask and measure the intensities in the ring mask, and subtract those from the values in the cell mask (using average intensities, or doing the extra math to compensate for differences in surface area). Our Cellomics high content analysis software does just that, and it should be possible to implement this technique with other image analysis software. If cells are clumped, however, this won't be accurate, since the ring mask will overlap with the cells next door...

Julio Vazquez

Fred Hutchinson Cancer Research Center

Seattle, WA 98109-1024

http://www.fhcrc.org

On Mar 20, 2010, at 5:13 PM, MODEL, MICHAEL wrote:

Hello,

Does anyone know of a background subtraction method that would act like this:

1) find the entire area occupied by a cell (automatically or manually)
2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.
3) subtract the restored background from the original

Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!

Mike

Julio Vazquez

Re: background subtraction

In reply to this post by mmodel

Hi Michael,

I guess what you meant was to create an image with the background subtracted.... not just get a background-subtracted measurement. That's a different beast...

I could imagine a method where you fit the background to points that you select across your image (in the background), and then you subtract that from your original image. If you pick enough background points, this should be relatively accurate. You could use a thresholded image as the source of the points.

Another approach would be to threshold your cells, keep background as is and set thresholded regions to the values of adjacent pixels, and use the resulting image as the background to subtract from the original image.

As I read myself, I realize the answer may not be tremendously useful, but someone with programming skills may pick on the idea......

Julio

On Mar 20, 2010, at 5:13 PM, MODEL, MICHAEL wrote:

Hello,

Does anyone know of a background subtraction method that would act like this:

1) find the entire area occupied by a cell (automatically or manually)
2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.
3) subtract the restored background from the original

Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!

Mike

Shalin Mehta

Re: background subtraction

In reply to this post by mmodel

Hi Michael,
Top-hat transform may work well. It is morphological filtering
operator, unlike linear filtering operator. MATLAB documentation shows
an example: http://www.mathworks.com/access/helpdesk/help/toolbox/images/imtophat.html

It works like this.
- morphological opening removes structures smaller than given mask
(structuring element) from the image. If you pick up a circular mask
slightly larger than cell, the opening operation will remove cells and
retain background.
- Subtracting the background should then leave cells on 'flat
background'. MATLAB example above removes background from a rice
image.
If background is smoother than cells of interest, above method works
almost always.
ImageJ has morphological opening function under Binary menu (I think).

best
Shalin
On Sun, Mar 21, 2010 at 8:13 AM, MODEL, MICHAEL <[hidden email]> wrote:

> Hello,
>
> Does anyone know of a background subtraction method that would act like this:
>
> 1) find the entire area occupied by a cell (automatically or manually)
> 2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.
> 3) subtract the restored background from the original
>
> Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!
>
> Mike

mmodel

Re: background subtraction

In reply to this post by Julio Vazquez

Thanks, Julio - That's exactly how I do it with ImageJ. It's probably good enough only when variations in the background are small on the scale of the analyzed cell or a group of cells.

From: Confocal Microscopy List [[hidden email]] On Behalf Of Julio Vazquez [[hidden email]]
Sent: Saturday, March 20, 2010 9:20 PM
To: [hidden email]
Subject: Re: background subtraction

Julio Vazquez

Fred Hutchinson Cancer Research Center

Seattle, WA 98109-1024

http://www.fhcrc.org

On Mar 20, 2010, at 5:13 PM, MODEL, MICHAEL wrote:

Hello,

Does anyone know of a background subtraction method that would act like this:

1) find the entire area occupied by a cell (automatically or manually)

2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.

3) subtract the restored background from the original

Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!

Mike

George McNamara

Re: background subtraction ... nested ring(s)

Hi Michael,

You did not explain what type of background you need to subtract. If camera noise is dominant, or if the background is not uniform. the average intensity of the "ring" may not be as good an estimator as averaging a series of dark reference images. Also, if fluorescence illumination shading is a problem (re: your 2001 Cytometry paper), ill-advised math operations could make the data worse.

If you use fluorescence to define the cell, you may need to use an additional channel(s), say Alexa Fluor 647-phalloidin or DiI (or DiD) or calcein(-AM) to define the cells (beta catenin immunofluorescence is useful for epithelial cell sheets). I am a big fan of confocal interference reflection microscopy to define the edges of cells (including filopodia) without having to rely on fluorescence. Hard to automate cIRM cell edge finding. Speaking of IRM, three nice recent papers are PubMed: 20013754, 19947618,
and 19816893.

That said, it is easy to make nested ring region(s) of interest in MetaMorph and other software.

If you don't have MetaMorph, but want it, see the "Amazing Image" free MetaMorph Basic offline offer in any of the last couple of issues of MetaMatters, http://www.moleculardevices.com/pages/software/meta_matters.html ... free in exchange for copyright and description of an "amazing image" that has been acquired, processed, or measured with any version of MetaMorph (send me the image at my UM address if you need to "MetaMorphize" your amazing image).

Also, http://www.moleculardevices.com/anc/ for other M.D. offers.

Sincerely,

George

At 07:44 AM 3/21/2010, you wrote:

Thanks, Julio - That's exactly how I do it with ImageJ. It's probably good enough only when variations in the background are small on the scale of the analyzed cell or a group of cells.

From: Confocal Microscopy List [[hidden email]] On Behalf Of Julio Vazquez [[hidden email]]
Sent: Saturday, March 20, 2010 9:20 PM
To: [hidden email]
Subject: Re: background subtraction

If you are segmenting your cells and defining a cell mask to measure intensities, you could add a second mask such as a ring or a few pixels width around the cell mask and measure the intensities in the ring mask, and subtract those from the values in the cell mask (using average intensities, or doing the extra math to compensate for differences in surface area). Our Cellomics high content analysis software does just that, and it should be possible to implement this technique with other image analysis software. If cells are clumped, however, this won't be accurate, since the ring mask will overlap with the cells next door...
--
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024

http://www.fhcrc.org

==

On Mar 20, 2010, at 5:13 PM, MODEL, MICHAEL wrote:

Hello,

Does anyone know of a background subtraction method that would act like this:

1) find the entire area occupied by a cell (automatically or manually)
2) reconstruct the background under a cell by interpolation from the outer boundary into the area occupied by a cell.
3) subtract the restored background from the original

Other and simpler background removal methods of which I know are based on filters and aren't precise enough. Thanks!

Mike

George McNamara, Ph.D.
Image Core Manager
Analytical Imaging Core Facility
University of Miami, Miller School of Medicine
Miami, FL 33136
[hidden email]
[hidden email]
305-243-8436 office
http://www.sylvester.org/AICF (Analytical Imaging Core Facility)
http://www.sylvester.org/AICF/pubspectra.zip (the entire 2000+ spectra .xlsx file is in the zip file)
http://home.earthlink.net/~geomcnamara