Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

Claire,

DIC images are typically difficult to separate from background. To my way 
of thinking more so because of uneven background illumination than contrast 
(since contrast can often be increased).

A really useful way of "evening out" background is through the use of the 
high pass filter in Photoshop (under Filter>Other>High Pass). Iteratively 
set the radius in this dialogue box at a point in which the DIC features 
become dark against an evenly lit background. If the images suffer from 
lack of contrast, these might be just visible above background.

The high pass filter in Photoshop operates differently than in many 
quantization programs as it shifts the histogram to center on 128 (the 
middle point of an 8-bit image), and so I'd recommend this filter over 
those in many scientific quantization programs.

The edges may still be darker if there is a vignetting problem, or if the 
DIC effects cause a large degree of uneven illumination from one end of the 
image to the other. In that case, you might consider using a consistent 
part of the image versus the whole thing.

If noise is an issue (it probably isn't with a brighfield image), you may 
want to also use a median filter to remove noise (Filter>Noise>Median).

The image can now be thresholded (Image>Adjust>Threshold) and saved for 
subsequent analysis in MetaMorph. If thresholding is a problem because of 
lack of contrast (probably not), increase contrast first using Curves or 
Levels (Image>Adjust>Levels).

If the features are used only for counting, the entire cell need not be 
thresholded to black: just enough to mark it as 1. If a scattering of dots 
is introduced at random areas in the image before all cells are thresholded 
to black, then these can be eliminated using a cut off in MetaMorph, or by 
contracting and expanding the selection in Photoshop (1. Under Select, 
choose Color Range and then Shadows from the drop down list. That will 
select all the black thresholded areas. 2. Under Select, choose Modify and 
then Contract. Generally 1 pixel is enough, but this is iterative. 3. Under 
Select, choose Modify and then Expand. Choose the pixel value for 
Contract).

This whole procedure can be automated by recording the steps in the Actions 
dialogue box, and then applied to all images in a folder using 
Automate>Batch, but this is another lot of sentences to describe.

As far as brightfield goes, it all depends on the stain that is used (if a 
stain was used at all). If you still can't get separation, you can send an 
image to me at the email below, or write me if you have any further issues.

Jerry Sedgewick
Director, Biomedical Processing Lab (BIPL)
University of Minnesota, Department of Neuroscience
www.bipl.umn.edu
[log in to unmask]
 

Search the CONFOCAL archive at 
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal I have a user 
interested in using metamorph to count cells in DIC/brightfield. There is 
not a lot of contrast between the cells and the background. How do people 
manage to count with these kinds of images?
 
Sincerely,
 
Claire
 
 
_________________________________________________________________

Claire M. Brown, PhD

Life Sciences Complex Imaging Facility Director

McGill University Department of Biochemistry

http://www.lifesciencescomplex.mcgill.ca/imaging