flattening a layer from z-stacks of thick uneven tissue

[Mark Krebs]

Hi All,

I'm a relatively new ImageJ user.  I have confocal fluorescence microscopy
z-stacks of a 50-um thick epithelial tissue stained for nuclei (blue
channel.  The tissue is not completely flat; the z-position of its topmost
surface varies gradually within the xy plane by 5-10 um.   I would like to
analyze the cell nuclei of the topmost ~10-um thick cell layer (number,
area, volume, etc.), but I am unable to merge the z-slices without
fluorescence contributions from nuclei in other layers.  This makes it
much more difficult to identify nuclei and analyze them properly.

I am wondering whether there is any ImageJ macro or plugin that will
isolate the nuclei of the topmost cell layer from underlying nuclei by
calculating a topmost nuclear surface (S), which follows the uneven
surface of the tissue.  I imagine several steps to the analysis.  

First, there would be an algorithm to define S by threshold criteria,
which should be feasible since the top of the stack is black background.  
At each xy location of the image, the z-stack would be examined from the
top, defining the z-position of the surface as the first pixel encountered
that exceeds the threshold blue value.  Since nuclei are separated by ~20-
30 um, there would need to be a mechanism to ignore z positions calculated
at xy locations where there are no nuclei in the topmost cell layer.  This
might be achieved perhaps by analyzing a limited z-range that is provided
as input, or by "rolling a ball" of defined radius over the image surface
and selecting only the topmost positions.  Finally, S would be calculated
from the coordinates of the topmost nuclear positions.

Second, S would be used to collect a slab of z-pixels (with an input value
slightly exceeding the nuclear diameter) downwards from the topmost
nuclear surface.  This slab of z-pixels would be used to create a new,
flattened z-stack, which can be used as input for merging and analysis.

Does anyone know of an ImageJ application for this purpose?  Or is there a
simpler approach?  If there is no such application (we DON'T have an app
for that!), can someone offer advice on how to go about creating such an
application?

Thanks everyone!

Cheers,
Mark

[Glen MacDonald-2]

Have you tried the 'Extended Depth of field' plugin?

Regards,
Glen

Glen MacDonald
Core for Communication Research
Virginia Merrill Bloedel Hearing Research Center
Box 357923
University of Washington
Seattle, WA 98195-7923  USA
[phone]
[hidden email]

******************************************************************************
The box said "Requires WindowsXP or better", so I bought a Macintosh.
******************************************************************************


On Oct 12, 2009, at 7:26 AM, Mark Krebs wrote:

> Hi All,
>
> I'm a relatively new ImageJ user.  I have confocal fluorescence  
> microscopy
> z-stacks of a 50-um thick epithelial tissue stained for nuclei (blue
> channel.  The tissue is not completely flat; the z-position of its  
> topmost
> surface varies gradually within the xy plane by 5-10 um.   I would  
> like to
> analyze the cell nuclei of the topmost ~10-um thick cell layer  
> (number,
> area, volume, etc.), but I am unable to merge the z-slices without
> fluorescence contributions from nuclei in other layers.  This makes it
> much more difficult to identify nuclei and analyze them properly.
>
> I am wondering whether there is any ImageJ macro or plugin that will
> isolate the nuclei of the topmost cell layer from underlying nuclei by
> calculating a topmost nuclear surface (S), which follows the uneven
> surface of the tissue.  I imagine several steps to the analysis.
>
> First, there would be an algorithm to define S by threshold criteria,
> which should be feasible since the top of the stack is black  
> background.
> At each xy location of the image, the z-stack would be examined from  
> the
> top, defining the z-position of the surface as the first pixel  
> encountered
> that exceeds the threshold blue value.  Since nuclei are separated  
> by ~20-
> 30 um, there would need to be a mechanism to ignore z positions  
> calculated
> at xy locations where there are no nuclei in the topmost cell  
> layer.  This
> might be achieved perhaps by analyzing a limited z-range that is  
> provided
> as input, or by "rolling a ball" of defined radius over the image  
> surface
> and selecting only the topmost positions.  Finally, S would be  
> calculated
> from the coordinates of the topmost nuclear positions.
>
> Second, S would be used to collect a slab of z-pixels (with an input  
> value
> slightly exceeding the nuclear diameter) downwards from the topmost
> nuclear surface.  This slab of z-pixels would be used to create a new,
> flattened z-stack, which can be used as input for merging and  
> analysis.
>
> Does anyone know of an ImageJ application for this purpose?  Or is  
> there a
> simpler approach?  If there is no such application (we DON'T have an  
> app
> for that!), can someone offer advice on how to go about creating  
> such an
> application?
>
> Thanks everyone!
>
> Cheers,
> Mark

[Mark Krebs]

Thanks for the suggestion.

I've tried enhanced depth of field (EDF) plugins from http://bigwww.epfl.ch/demo/edf/, but so far I'm
unable to mask the fluorescence contributed by nuclei that are below the top layer.  EDF creates a
sharper image, but it appears to merge all of the z-stack, which is what I'm trying to avoid.  The
examples at the EDF website seem to be brightfield images of opaque objects (fly eye, mineral, laser
weld, mouse intestine) or of a transparent object with a single layer (retinal pigment epithelium).  My
images are of a transparent object with fluorescent nuclei in many layers.   How can EDF be used to
reduce fluorescence contributions from layers under the top layer of nuclei?

[Steffen Dietzel-2]

Dear list,

I came across a bug in the plugin Image5D and I am wondering if
somebody on the list would be able to look into it it.

The bug shows up with large data sets when I want to transform the
5D-Stack into an RGB-stack.
I have two channels, each e.g. 1000x1000 px, 92 slices. What I need
is an RGB-stack in which one channel has a pink-violett color, the
other yellowish-green (I want to make cyan-red 3D images (anaglyph
images) with another programm, thus I need two colors which both have
red and cyan components. I wouldn't know how to do such a 3D-stack
with two channels in ImageJ).

Opening the images and assigning colors in Imag5D works fine.
However, when I transform the 5D-stack into RGB, the resulting stack
shows problems.
- The first slice of the RGB stack shows the image that was last
shown in Image5D. A problem which of course can be avoided by going
to the top of the 5D-stack.
- More serious, several slices of the RGB-stack show the same
content, and some slices are even composed of different z-sections:
the upper part from one, the lower part from another slice.

My assumption is that maybe some parallel threads are not finished in
the expected order and that this occurs only under heavy load, i.e.
when large files are used. Unfortunately I do not know Java, so I
hope that there is somebody on this list who might fix this otherwise
so helpfull tool - or could provide other hints how I could reach my goal.

Best regards

Steffen

[Robert Baer]

> - More serious, several slices of the RGB-stack show the same content, and
> some slices are even composed of different z-sections: the upper part from
> one, the lower part from another slice.

It would be helpful if you provide your OS, Java version, etc.  The problem
could reside in the plugin or in Java.  Is it reproducible on more than one
OS?

I don't know if it is related, but on Windows Vista I was having problems
with frames of my stacks fractionating into scambled pieces.  This problem
seemed to resolve itself when I installed the latest Java rather than the
release candidate that is installed with the downloaded ImageJ.  Based on
myexperience, an unsophisticated suggestion would be to at least try the
most recent Java for your OS.  I leave it to others to provide more direct
advice.

> My assumption is that maybe some parallel threads are not finished in the
> expected order and that this occurs only under heavy load, i.e. when large
> files are used. Unfortunately I do not know Java, so I hope that there is
> somebody on this list who might fix this otherwise so helpfull tool - or
> could provide other hints how I could reach my goal.
>
> Best regards
>
> Steffen
>

[Mario Faretta]

Hi,
just my five cents to the discussion: I also noticed the problem with both
Windows (XP Service Pack3, last ImageJ release g Java 1.6.0 but noticed also
with previous version) and Mac OS X (10.4.11, ImageJ 1.43 g and previous
realeses Java 1.5.0_19).
I noticed the behaviour quite frequently when managing large amount of data.
Not Being a Java expert and/or investigated systematically I always bypassed
it by tricks (using color merge in the stack function after reconstructing the
channel stacks using the Hypervolume shuffling plugin).
Hope it helps
Mario

Robert Baer ([hidden email]) wrote:

>
> > - More serious, several slices of the RGB-stack show the same content, and
> > some slices are even composed of different z-sections: the upper part from
> > one, the lower part from another slice.
>
> It would be helpful if you provide your OS, Java version, etc.  The problem
> could reside in the plugin or in Java.  Is it reproducible on more than one
> OS?
>
> I don't know if it is related, but on Windows Vista I was having problems
> with frames of my stacks fractionating into scambled pieces.  This problem
> seemed to resolve itself when I installed the latest Java rather than the
> release candidate that is installed with the downloaded ImageJ.  Based on
> myexperience, an unsophisticated suggestion would be to at least try the
> most recent Java for your OS.  I leave it to others to provide more direct
> advice.
>
>
> > My assumption is that maybe some parallel threads are not finished in the
> > expected order and that this occurs only under heavy load, i.e. when large
> > files are used. Unfortunately I do not know Java, so I hope that there is
> > somebody on this list who might fix this otherwise so helpfull tool - or
> > could provide other hints how I could reach my goal.
> >
> > Best regards
> >
> > Steffen
> >
>

--
Mario Faretta
Department of Experimental Oncology
European Institute of Oncology
c/o IFOM-IEO Campus for Oncogenomics
via Adamello 16
20139 Milan
Italy
Phone: ++39-02574303054
email: [hidden email]
http://www.ifom-ieo-campus.it



[- Il Tuo 5 per Mille a favore della Ricerca dell'Istituto Europeo di
Oncologia

Tutti coloro che presentano il modello Unico, il modello 730 o più
semplicemente che ricevono dal proprio datore di lavoro il modello CUD, hanno
la facoltà di scegliere la destinazione del proprio 5 per mille.

Nella casella riservata al Finanziamento agli Enti della Ricerca Sanitaria
inserisci il codice fiscale dello IEO (08 69 14 40 153) ed apponi la Tua
firma.
Il Tuo 5 per Mille verrà destinato alla ricerca contro il cancro dell'Istituto
Europeo di Oncologia

NON COSTA NULLA E NON COMPORTA ALCUN AUMENTO DELLE IMPOSTE DA VERSARE
NON È UN'ALTERNATIVA ALL'8 PER MILLE
E' UN GESTO CONCRETO E DI GRANDE VALORE

Per saperne di più vai al sito dello IEO www.ieo.it < http://www.ieo.it >
oppure scrivi a [hidden email] < mailto:[hidden email]>

Segnala ad un amico questa opportunità

 -]

[Gabriel Landini]

On Tuesday 13 October 2009  15:21:14 Mark Krebs wrote:
> My images are of a transparent object with fluorescent nuclei in many
> layers.

Sorry to ask the obvious. Have you tried thresholding the stack?
Cheers,
G.

[Steffen Dietzel-2]

Sorry, should have done that right away:
OS is WinXP 32 bit.
I've tried it with ImageJ 1.43g - Java1.6.0_10
and 1.37s - Java 1.3.1_13.

The Image5D-plugin was reinstalled from the Website after the problem
occured, to make sure it is the latest version and not corrupted.

I think I had the same problem under Linux Fedora 64 bit - can't look
up the details of this system right now.

Concerning the Java-Versions, there are so many around (JDK, JRE, in
different flavors and versions that confused me) that I stuck to the
one that came with ImageJ. I also seem to remember that some time ago
some versions were incompatible with imageJ, and I just thought that
won't be any of those that come with the bundle...
Still, it's a possibility of course.

Steffen

At 16:02 13.10.2009, you wrote:

>>- More serious, several slices of the RGB-stack show the same
>>content, and some slices are even composed of different z-sections:
>>the upper part from one, the lower part from another slice.
>
>It would be helpful if you provide your OS, Java version, etc.  The
>problem could reside in the plugin or in Java.  Is it reproducible
>on more than one OS?
>
>I don't know if it is related, but on Windows Vista I was having
>problems with frames of my stacks fractionating into scambled
>pieces.  This problem seemed to resolve itself when I installed the
>latest Java rather than the release candidate that is installed with
>the downloaded ImageJ.  Based on myexperience, an unsophisticated
>suggestion would be to at least try the most recent Java for your
>OS.  I leave it to others to provide more direct advice.
>
>
>>My assumption is that maybe some parallel threads are not finished
>>in the expected order and that this occurs only under heavy load,
>>i.e. when large files are used. Unfortunately I do not know Java,
>>so I hope that there is somebody on this list who might fix this
>>otherwise so helpfull tool - or could provide other hints how I
>>could reach my goal.
>>
>>Best regards
>>
>>Steffen

[Glen MacDonald-2]

As mentioned by Gabriel, thresholding may help.  Deconvolution or  
imaging with higher NA will also reject more of the out of focus light.

Glen
On Oct 13, 2009, at 2:02 AM, Mark Krebs wrote:

> Thanks for the suggestion.
>
> I've tried enhanced depth of field (EDF) plugins from http://bigwww.epfl.ch/demo/edf/ 
> , but so far I'm
> unable to mask the fluorescence contributed by nuclei that are below  
> the top layer.  EDF creates a
> sharper image, but it appears to merge all of the z-stack, which is  
> what I'm trying to avoid.  The
> examples at the EDF website seem to be brightfield images of opaque  
> objects (fly eye, mineral, laser
> weld, mouse intestine) or of a transparent object with a single  
> layer (retinal pigment epithelium).  My
> images are of a transparent object with fluorescent nuclei in many  
> layers.   How can EDF be used to
> reduce fluorescence contributions from layers under the top layer of  
> nuclei?

[dpoburko]

Sorry to beat a dead horse, but did you try the 3D-modeling version of
Extended depth of field? I will also reduce your stack to single plain,
but it does a pretty solid job of removing out of focus fuzz. I'm using
this to gather fluorescent puncta on multiple plain into a single image.
Deconvolution before had might also be helpfull, as per Glen's suggestion.

Damon

Mark Krebs wrote:

> Hi All,
>
> I'm a relatively new ImageJ user.  I have confocal fluorescence microscopy
> z-stacks of a 50-um thick epithelial tissue stained for nuclei (blue
> channel.  The tissue is not completely flat; the z-position of its topmost
> surface varies gradually within the xy plane by 5-10 um.   I would like to
> analyze the cell nuclei of the topmost ~10-um thick cell layer (number,
> area, volume, etc.), but I am unable to merge the z-slices without
> fluorescence contributions from nuclei in other layers.  This makes it
> much more difficult to identify nuclei and analyze them properly.
>
> I am wondering whether there is any ImageJ macro or plugin that will
> isolate the nuclei of the topmost cell layer from underlying nuclei by
> calculating a topmost nuclear surface (S), which follows the uneven
> surface of the tissue.  I imagine several steps to the analysis.  
>
> First, there would be an algorithm to define S by threshold criteria,
> which should be feasible since the top of the stack is black background.  
> At each xy location of the image, the z-stack would be examined from the
> top, defining the z-position of the surface as the first pixel encountered
> that exceeds the threshold blue value.  Since nuclei are separated by ~20-
> 30 um, there would need to be a mechanism to ignore z positions calculated
> at xy locations where there are no nuclei in the topmost cell layer.  This
> might be achieved perhaps by analyzing a limited z-range that is provided
> as input, or by "rolling a ball" of defined radius over the image surface
> and selecting only the topmost positions.  Finally, S would be calculated
> from the coordinates of the topmost nuclear positions.
>
> Second, S would be used to collect a slab of z-pixels (with an input value
> slightly exceeding the nuclear diameter) downwards from the topmost
> nuclear surface.  This slab of z-pixels would be used to create a new,
> flattened z-stack, which can be used as input for merging and analysis.
>
> Does anyone know of an ImageJ application for this purpose?  Or is there a
> simpler approach?  If there is no such application (we DON'T have an app
> for that!), can someone offer advice on how to go about creating such an
> application?
>
> Thanks everyone!
>
> Cheers,
> Mark
>  

[Mark Krebs]

Thanks for the suggestions.  I will try them.

However, I suspect thresholding and enhanced depth of field may not be sufficient to overcome the
UNEVEN surface of the tissue, which results in target nuclei from the topmost epithelial cell layer and
undesirable nuclei from lower levels in the same z-slice.  All are in focus and have similar size and
intensity.  Perhaps I can differentiate them with shape-based thresholding.  Any ideas?

Cheers,
Mark

[Joachim Walter]

Dear all,

the last update for Image5D has been quite some time ago, now, and all
incoming requests for bugfixes or new features I have to answer with
"sorry, no time". So I finally decided to look for someone to take over
the project. If there is anyone willing to take over the Image5D
project, do bugfixes, add features as he/she wishes, please contact me
on- or offlist.

Best,
Joachim

[ctrueden]

Hi Joachim,

LOCI would be willing to take custody of the Image5D project. I am currently
traveling, but we can discuss further off-list once I return on Wednesday.

Regards,
Curtis

On Sun, Oct 18, 2009 at 9:02 AM, Joachim Walter
<[hidden email]>wrote:

> Dear all,
>
> the last update for Image5D has been quite some time ago, now, and all
> incoming requests for bugfixes or new features I have to answer with "sorry,
> no time". So I finally decided to look for someone to take over the project.
> If there is anyone willing to take over the Image5D project, do bugfixes,
> add features as he/she wishes, please contact me on- or offlist.
>
> Best,
> Joachim
>

[Przemko]

Damon Poburko wrote:
> Sorry to beat a dead horse, but did you try the 3D-modeling version of
> Extended depth of field? I will also reduce your stack to single plain,
> but it does a pretty solid job of removing out of focus fuzz. I'm using
> this to gather fluorescent puncta on multiple plain into a single image.
> Deconvolution before had might also be helpfull, as per Glen's suggestion.
>
> Damon
>
>  

I tried that. I also tried extended depth of field in Adobe Photoshop
CS4. I must say InageJ failed. I have photos of zebrafish and got to
stitch a few of them together because of the thickness of the embryo,
not all planes are in focus. I tried different versions and settings of
Extended depth of field in ImageJ and got terrible results. Then I did
CS4 and presto. 4 clicks and I was there. I guess the main thing is that
CS4 aligns the sections. When you change focus to get different parts of
fish sharp the entire specimen shifts a bit (normal). ImageJ is however
unable to adjust for that. CS4 first aligns then does the rest. The
results are great. Any idea how to achieve that in ImajgeJ
Przemko

--
Przemko Tylzanowski PhD
Dept. Musculoskeletal Sciences
LSD & Joints
ON1 box 813

Univeristy of Leuven
Herestraat 49
3000 Leuven
Belgium

phone: [phone]
fax  : [phone]