JACoP parameter and PSF

[gauri]

Hi,
I am looking at colocalization of two proteins on confocal microscope. I am using JACoP plugin to find Pearson
's coefficient and mander's coefficient. I had a few questions about setting parameters for JACoP. For the microscope parameter it asks about XY calibration and Z calibration. I am not sure what does that mean and what will be the correct value for it. The other problem i had was does JACoP calculate PSF. What PSF does it use. it never asks anythiong about PSF.
Thanks
Gauri

[Crowell Elizabeth]

Hello fellow ImageJ users,

Has anyone developed an algorithm in ImageJ for determining if an image
is out-of-focus?

I managed to flag images that are likely to be out-of-focus based on
intensity measurements of DAPI (nuclear) staining after running Find
Edges and thresholding, but any method based on intensity seems to fail
from one experiment to the next, since the global intensity varies
depending on staining success and lamp power, etc.  I did not find any
plugins or macros on the ImageJ site that present such a screening
procedure.

I'm sure someone out there must know of a way to do this well, since it
is a common problem!

Kind Regards,

 

Elizabeth CROWELL

----------------------------------------------------------------------
Membrane Traffic and Cell Division Research Group
Institut Pasteur
28 rue du Dr Roux
75015 PARIS, France

Tel :  [phone]
Fax : [phone]
----------------------------------------------------------------------

[Daniel James White]

Hi Gauri

On Mar 12, 2011, at 6:08 AM, IMAGEJ automatic digest system wrote:

>
> Date:    Fri, 11 Mar 2011 08:03:11 -0800
> From:    gauri <[hidden email]>
> Subject: JACoP parameter and PSF
>
> Hi,
> I am looking at colocalization of two proteins on confocal microscope. I am
> using JACoP plugin to find Pearson
> 's coefficient and mander's coefficient. I had a few questions about setting
> parameters for JACoP. For the microscope parameter it asks about XY
> calibration and Z calibration. I am not sure what does that mean and what
> will be the correct value for it. The other problem i had was does JACoP
> calculate PSF. What PSF does it use. it never asks anythiong about PSF.
> Thanks
> Gauri

Email the authors directly if you want fine details... but....

you will find more info here

http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:jacop_2.0:just_another_colocalization_plugin:start

and here

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2818.2006.01706.x/pdf.mpi-cbg.de/cgi-bin/gitweb.cgi?p=fiji.git;a=blob_plain;f=src-plugins/Colocalisation_Analysis/Colocalisation_Test.java;hb=HEAD

in a nutshell....

x y and z calibration is the pixel spacing of the image. How far apart are the pixels with respect to the sample.

PSF approximate size  is used for the image randomisation of the Costes significance test.
So it knows how big the chunks of image are to shuffle.

Excluding the object based coloc, all this stuff has now been redone... with fast image randomisation
standardised pdf result output and more,
in the upcoming Coloc_2 plugin for Fiji /  imageJ2 / imglib

I am writing the docs now.... and the new plugin will be released to the Fiji updater
as soon as we verify that recent changes to imglib dont break it.


Do you want to test it and give me feedback?

cheers

Dan



Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Processing and Analysis
Light Microscopy Facility
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
+49  (0)351 210 2627 (Work phone at MPI-CBG)
+49  (0)351 210 1078 (Fax MPI-CBG LMF)

http://www.bioimagexd.net                BioImageXD
http://pacific.mpi-cbg.de                Fiji (is just ImageJ - batteries included)
http://www.chalkie.org.uk
[hidden email]
( [hidden email] )

[Daniel James White]

Hi Elizabeth,

On Mar 12, 2011, at 6:08 AM, IMAGEJ automatic digest system wrote:

>
> Date:    Fri, 11 Mar 2011 19:01:24 +0100
> From:    Crowell Elizabeth <[hidden email]>
> Subject: automatic detection out-of-focus images
>
> Hello fellow ImageJ users,
>
> Has anyone developed an algorithm in ImageJ for determining if an image
> is out-of-focus?

yes, several.

See
http://pacific.mpi-cbg.de/wiki/index.php/Extended_Depth_of_Field

>
> I managed to flag images that are likely to be out-of-focus based on
> intensity measurements of DAPI (nuclear) staining after running Find
> Edges and thresholding, but any method based on intensity seems to fail
> from one experiment to the next, since the global intensity varies
> depending on staining success and lamp power, etc.

Yes, thats the case.

>  I did not find any
> plugins or macros on the ImageJ site that present such a screening
> procedure.

http://pacific.mpi-cbg.de/wiki/index.php/Extended_Depth_of_Field

>
> I'm sure someone out there must know of a way to do this well, since it
> is a common problem!

its been worked on a lots...

so you should be able to use  ideas from
http://pacific.mpi-cbg.de/wiki/index.php/Extended_Depth_of_Field
http://en.wikipedia.org/wiki/Focus_stacking

cheers

Dan



>
> Kind Regards,
>
>
>
> Elizabeth CROWELL

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Processing and Analysis
Light Microscopy Facility
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
+49  (0)351 210 2627 (Work phone at MPI-CBG)
+49  (0)351 210 1078 (Fax MPI-CBG LMF)

http://www.bioimagexd.net                BioImageXD
http://pacific.mpi-cbg.de                Fiji (is just ImageJ - batteries included)
http://www.chalkie.org.uk
[hidden email]
( [hidden email] )

[Gabriel Landini]

On Friday 11 Mar 2011, you wrote:
> Has anyone developed an algorithm in ImageJ for determining if an image
> is out-of-focus?

In addition to the links that Dan provided, be aware that there is no method
to say with certainty if one single arbitrary photo is in focus unless you
know in advance the image contents.
For example, a photo of a field that has no high frequency features might be
impossible to tell apart from another which is out of focus.
That is why most autofocusing methods use a search or scan over the z axis to
find the frame or region that contains the highest contrast or high frequency
components.

In microscopy the problem is worse than standard photography because we deal
with semi transparent objects in different z planes, so one might not be
focusing exactly the same thing at various z distances. In such instances,
defining depth from focus fails (that is there might be highly focused regions
in more than one z plane).

Cheers

Gabriel

[gauri]

Thanks Dan
I had another question about analysis of colocalization using JACoP.
i make substack of 4-5 slices for both the channels. Do i just use those
substack as such for colocalization or do i merge these 4-5 slices in one
image before i use them in JACoP.
If i use 4-5 slices for each substack, how does JACoP calculate the r value?
Does it look at r value of each slice and then average it, or does it
average out the pixel. i am so confused about how it works.

--
View this message in context: http://imagej.588099.n2.nabble.com/JACoP-parameter-and-PSF-tp6161933p6165172.html
Sent from the ImageJ mailing list archive at Nabble.com.

[Daniel James White]

Hi Gauri,

dont forget to cc me directly... or i might miss message to addressed to me but only sent to the mailing list.

On Mar 13, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:

> Date:    Sat, 12 Mar 2011 14:42:51 -0800
> From:    gauri <[hidden email]>
> Subject: Re: colocalization analysis
>
> Thanks Dan
> I had another question about analysis of colocalization using JACoP.
> i make substack of 4-5 slices for both the channels. Do i just use those
> substack as such for colocalization or do i merge these 4-5 slices in one
> image before i use them in JACoP.

if you merge z slices from the same xy field of view,
you lose the z resolution, and can then get false coloc of objects in different z planes.
Don't do that!

However, if the slices are from different x,y areas of the sample, then its ok.

> If i use 4-5 slices for each substack, how does JACoP calculate the r value?

Pearsons r is a global statistic calculated from all the pixels included in the calculation.
You would get the same r if the same pixels were in many z slices or only one 2D image...
or even in a different order. Its a pixel by pixel calculation, and the result is summed over all pixels counted.

> Does it look at r value of each slice and then average it, or does it
> average out the pixel. i am so confused about how it works.

Look at the maths on the Manders and Costes articles.
You will be able to convince yourself by understanding the
simple equation used to calculate Pearson's r :

http://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient

https://ifn.mpi-cbg.de/wiki/ifn/images/0/0d/QuantitativeColocAnalysis-12-2010.pdf

r =  ( sum of (channel1 pixel intensity - average ch1 intensity) x (channel2 pixel intensity - average ch1 intensity) ) / sqrt of square of top half of eqn

...that either way gives the same result or not....

In practice, one typically takes all pixels from all slices to be analysed and uses the mean of those...
and does not do it slice by slice then average the results.... there is no point in doing the latter...? Right?

There are 2 (or maybe more) ways to implement Pearsons r in computer code.
1) The "fast" way used in the colocalization thershold plugin, using a rearrangement of the equation
2) the "classic" way using the equation as written.

The new Coloc_2 plugin can do both... and we notice that the classic way is not really slower and less susceptible to numerical problems... eg when the means of both images are the same.

Not sure which JaCOP uses without looking at the code...

cheers

Dan






Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Processing and Analysis
Light Microscopy Facility
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
+49  (0)351 210 2627 (Work phone at MPI-CBG)
+49  (0)351 210 1078 (Fax MPI-CBG LMF)

http://www.bioimagexd.net                BioImageXD
http://pacific.mpi-cbg.de                Fiji (is just ImageJ - batteries included)
http://www.chalkie.org.uk
[hidden email]
( [hidden email] )

[Crowell Elizabeth]

Hello ImageJ list members, Dan, Gabriel,

Gabriel Landini a écrit :
>> Has anyone developed an algorithm in ImageJ for determining if an image
>> is out-of-focus?
>>    
>
> In addition to the links that Dan provided, be aware that there is no method
> to say with certainty if one single arbitrary photo is in focus unless you
> know in advance the image contents.
> For example, a photo of a field that has no high frequency features might be
> impossible to tell apart from another which is out of focus
I think I understand the problem: if my image does not contain any real
objects, then it cannot be determined if it is in-focus.  The first step
in my screening procedure is to eliminate "empty" images, so hopefully
this should not be an issue?
However, I am not working with stacks, actually, but with a set of 2D
images!
I apologize for not having specified this in my first email.  In the
image set, I simply want to eliminate the out-of-focus images (i.e.
blurry enough to confound analysis) and keep only the sharp ones for
analysis.

I have tried the extended depth-of-field plugin and it is very easy to
use, and gave me beautiful results on the first try.  This is far
superior to generating simple projections, evidently.  I wish I were
erudite enough to understand how it works, so that I could apply this
principle to my current problem.  My problem is not to sort blurry
pixels from sharp pixels in the same image, but to sort blurry images
from sharp images.

Is there a way to apply the wavelet transform concept to sort my images?

Thanks again,
Elizabeth

> In microscopy the problem is worse than standard photography because we deal
> with semi transparent objects in different z planes, so one might not be
> focusing exactly the same thing at various z distances. In such instances,
> defining depth from focus fails (that is there might be highly focused regions
> in more than one z plane).
>
> Cheers
>
> Gabriel
>  

--

Elizabeth CROWELL

----------------------------------------------------------------------
Membrane Traffic and Cell Division Research Group
Institut Pasteur
28 rue du Dr Roux
75015 PARIS, France

Tel :  [phone]
Fax : [phone]
----------------------------------------------------------------------

[Daniel James White]

Hi Elizabeth,


On Mar 14, 2011, at 4:37 PM, Crowell Elizabeth wrote:

> Hello ImageJ list members, Dan, Gabriel,
>
> Gabriel Landini a écrit :
>>
>>> Has anyone developed an algorithm in ImageJ for determining if an image
>>> is out-of-focus?
>>>    
>>
>> In addition to the links that Dan provided, be aware that there is no method
>> to say with certainty if one single arbitrary photo is in focus unless you
>> know in advance the image contents.
>> For example, a photo of a field that has no high frequency features might be
>> impossible to tell apart from another which is out of focus
> I think I understand the problem: if my image does not contain any real objects, then it cannot be determined if it is in-focus.  The first step in my screening procedure is to eliminate "empty" images, so hopefully this should not be an issue?
> However, I am not working with stacks, actually, but with a set of 2D images!
> I apologize for not having specified this in my first email.  In the image set, I simply want to eliminate the out-of-focus images (i.e. blurry enough to confound analysis) and keep only the sharp ones for analysis.
>
> I have tried the extended depth-of-field plugin and it is very easy to use, and gave me beautiful results on the first try.  This is far superior to generating simple projections, evidently.  I wish I were erudite enough to understand how it works, so that I could apply this principle to my current problem.  My problem is not to sort blurry pixels from sharp pixels in the same image, but to sort blurry images from sharp images.
>
> Is there a way to apply the wavelet transform concept to sort my images?

Where there is a will there is a way. It would probably require some scripting, and resources to get that done.
Out of the box, well there might be... but I'm not sure myself.

maybe there is something in CellProfiler that looks for blurry images to throw them out....
indeed maybe there is...

http://cellprofiler.org/forum/viewtopic.php?f=2&t=233&hilit=blur

cheers

Dan

>
> Thanks again,
> Elizabeth
>
>
>> In microscopy the problem is worse than standard photography because we deal
>> with semi transparent objects in different z planes, so one might not be
>> focusing exactly the same thing at various z distances. In such instances,
>> defining depth from focus fails (that is there might be highly focused regions
>> in more than one z plane).
>>
>> Cheers
>>
>> Gabriel
>>  
>
>
> --
>
> Elizabeth CROWELL
>
> ----------------------------------------------------------------------
> Membrane Traffic and Cell Division Research Group
> Institut Pasteur
> 28 rue du Dr Roux
> 75015 PARIS, France
>
> Tel :  [phone]
> Fax : [phone]
> ----------------------------------------------------------------------

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Visualisation, Processing and Analysis
Light Microscopy and Image Processing Facilities
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
[phone] (Work phone at MPI-CBG)
[phone] (Fax MPI-CBG LMF)

http://www.bioimagexd.net  BioImageXD
http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
http://www.chalkie.org.uk                Dan's Homepages
https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
dan (at) chalkie.org.uk
( white (at) mpi-cbg.de )

[Baron, Ken (US SSA)]

Just compress your images with jpeg or probably any image compressor. The file that is the largest is probably the one with best focus.

Ken Baron

>
>Is there a way to apply the wavelet transform concept to sort my images?
>

[Gabriel Landini]

On Monday 14 Mar 2011 15:49:25  Baron, Ken (US SSA) wrote:
> Just compress your images with jpeg or probably any image compressor. The
> file that is the largest is probably the one with best focus.

That would only work if all the images were of the same scene which is not the
case here.  A "busy" image out of focus may well take more space to store
compressed than one with very few objects but well in focus.
This would not work either when parts of the image are in focus and others are
not.
Rather than compressing the image, it would be better to look at the high
frequencies (for example in the Fourier space), but even so, this makes sense
only when you compare images of the same (or very similar) scene. This would
fail when the image is originally lacking high frequency components.

Anyway, finding the best focused image requires looking how much focus can be
achieved at other focal distances. The eye does it and photo cameras do it
that way too.
Have a look at various papers on autofocus techniques (which search for the
best focused image) to get an idea how these things are implemented.
Understanding the technical problem might help realising what can and cannot
be achieved under certain constraints.

Roca X, Binefa X, Vitria J. New autofocusing algorithm for cytological tissue
in a microscopy environment. Optical Engineering 1998;37(2):635-641.

Firestone L, Cook K, Culp K, Talsania N, Preston K. Comparison of autofocus
methods for automated microscopy. Cytometry 1991;12:195-206.

Santos A, Ortiz de Solorzano C, Vaquero JJ, Pena JM, Malpica N, Del Pozo F.
Evaluation of autofocus functions in molecular cytogenetic analysis. Journal
of Microscopy 1997;188(3):264-272.

Pieper RJ, Korpel A. Image processing for extended depth of field. Applied
Optics 1983;22(10):1449-1453.

I hope this is useful.
Regards

Gabriel

[dscho]

Hi,

On Mon, 14 Mar 2011, Crowell Elizabeth wrote:

> Hello ImageJ list members, Dan, Gabriel,
>
> Gabriel Landini a écrit :
> > > Has anyone developed an algorithm in ImageJ for determining if an
> > > image is out-of-focus?
> >
> > In addition to the links that Dan provided, be aware that there is no
> > method to say with certainty if one single arbitrary photo is in focus
> > unless you know in advance the image contents. For example, a photo of
> > a field that has no high frequency features might be impossible to
> > tell apart from another which is out of focus
>
> I think I understand the problem: if my image does not contain any real
> objects, then it cannot be determined if it is in-focus.

Even if it contains any real objects, it is hard to tell whether fuzziness
stems from out-of-focus effects or from the real optical properties of
your specimen; there are many reasons for "blurred" images in addition to
non-optimal focus. In general, therefore, your problem is not solvable.

Having said that, if you know what your specimen should look like, you may
be able to define a measure that describes "out-of-focusness".

The plugins mentioned so far typically do a good job when the optimal
focus results in sharp edges (or at least sharper ones than in
out-of-focus images), but do keep in mind that all of those methods work
on multiple planes of the _same_ specimen.

Ciao,
Johannes

[vtali]

We had the same problem with waterborne pathogen detection using FTIC/DAPI florescense channels and for Whole Slide Imaging of histology samples (H&E).

The key issue is that it is not the whole Image (Field of View) that is out of focus but some parts of it, because sample of "thick" for the objective focus depth .

We have tried avaialbe ImageJ plugings from here but none of tham

Now we are tying to find approach and write Java code for deconvolution that would work adaptevely.
The idea is to use Z-stack on representative VOF with known Z movement; reconstruct PSF and use it adaptevely across the image.

I will send update if you come up with something.

One Tip that we found: Adaptive background removal (Rolling ball in ImageJ)  + Unsharp helps with segmentation.

I too would be very interested if somebody finds a way to adaptively deconvolve focus blur on "thick" samples.

Vitali
Smart Imaging Technologies
713-589-3500
live.simagis.com


 

[dscho]

Hi,

On Thu, 7 Apr 2011, vtali wrote:

> We had the same problem with waterborne pathogen detection using
> FTIC/DAPI florescense channels and for Whole Slide Imaging of histology
> samples (H&E).
>
> The key issue is that it is not the whole Image (Field of View) that is
> out of focus but some parts of it, because sample of "thick" for the
> objective focus depth .
>
> We have tried avaialbe ImageJ plugings from here but none of tham

... yes?

The bigger issue is: how do _you_ differentiate between out-of-focus and
naturally blurry parts of an image?

Hth,
Johannes

[Gabriel Landini]

On Friday 08 Apr 2011 01:40:05 you wrote:

> > The key issue is that it is not the whole Image (Field of View) that is
> > out of focus but some parts of it, because sample of "thick" for the
> > objective focus depth .
> >
> > We have tried avaialbe ImageJ plugings from here but none of tham
>
> ... yes?
>
> The bigger issue is: how do _you_ differentiate between out-of-focus and
> naturally blurry parts of an image?

That is usually resolved by comparing the features across the Z axis for each
pixel. The real problem is that autofocus implies finding "the" most in focus
pixel in the z axis and this works fine for opaque objects (like in real world
photograhps) however in (well, brightfield at least) microscopy, objects are
not opaque but semi transparent, and therefore is it possible (for a fixed
focal length objective travelling along the z axis) to find more than one area
with "in focus" objects.

So the problem is not the same as with other types of imagery. There is no
guarantee (with autofocus algorithms) that the final image is the best one.
One needs a different type of autofocusing that takes partial/semi-transparent
occlusion into account.

Of course, with thicker sections this issue becomes more relevant.


Cheers

Gabriel