Electron microscopy nuclei segmentation

[A A]

>
> Hello,
>
> Is there a straight-forward way to segment nuclei in electron microscopy
> images using fiji/imagej?
>
> Best,
> Albina
>

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

[Christian Goosmann-2]

Hi Albina,
I'm guessing, with straight forward you mean simple. And from my
experience, automatic segmentation in TEM pictures of Epon sections is
far from that. Usually chromatin comes in at least two typical densities
that again match with other organelles in the cell. I guess you could
try the trainable segmentation plugin and feed it an image where you
select the chromatin areas as one class and everything else as
background. Train the classifier repeatedly and it might become
efficient. The problem there is that it is time consuming especially on
large image formats of several k's of pixels (like TEM cameras often
produce). So you might want to start with cropped images.
Good luck
Christin

---
Christian Goosmann
Mikroskopie
Max-Planck-Institut für Infektionsbiologie
Campus Charité Mitte
Charitéplatz 1
10117 Berlin
Tel.: [phone]

A A wrote:

>> Hello,
>>
>> Is there a straight-forward way to segment nuclei in electron microscopy
>> images using fiji/imagej?
>>
>> Best,
>> Albina
>>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

[A A]

Hello Christin,


I'm guessing, with straight forward you mean simple. And from my

> experience, automatic segmentation in TEM pictures of Epon sections is far
> from that. Usually chromatin comes in at least two typical densities that
> again match with other organelles in the cell. I guess you could try the
> trainable segmentation plugin and feed it an image where you select the
> chromatin areas as one class and everything else as background. Train the
> classifier repeatedly and it might become efficient. The problem there is
> that it is time consuming especially on large image formats of several k's
> of pixels (like TEM cameras often produce). So you might want to start with
> cropped images.
>

I used simple thresholding (after smoothing) to detect chromatin patches
and filtered the thresholded objects by size.  This works pretty well for
our data. The problem is that patches belonging to one nucleus are
separate. I need to find a way to assign chromatin patches to their
corresponding nuclei.

Best,
Albina

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