Deconvolution on MacBook?

[Mark Adelman (Work)]

Is it realistic to attempt to do a limited number of wide field  
deconvolutions using ImageJ on a MacBook?  My MacBook has a 2.13 GHz  
Intel Core 2 Duo Processor and Memory of 2 GB 800 MHz DDR2 SDRAM.  I  
am running Leopard (OS 10.5.8).  I have the ability to capture Z  
series fluorescence microscopy data sets (Zeiss Photomicroscope III  
with a Focus Controller and a digital camera), under the control of an  
iMac (running under 10.4.11).  I already know the iMac does not have  
enough computing power.  If anyone has suggestions regarding  
possibilities using the MacBook, please contact me offline (unless you  
think others would be interested).

Thanks in advance!!

Mark R. Adelman
[hidden email]

[Daniel James White]

Hi Mark,

On Mar 9, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:

> Date:    Tue, 8 Mar 2011 11:42:17 -0500
> From:    "Mark Adelman (Work)" <[hidden email]>
> Subject: Deconvolution on MacBook?
>
> Is it realistic to attempt to do a limited number of wide field  
> deconvolutions using ImageJ on a MacBook?

Yes, so long as the images are small enough to fit in your RAM with a copy or two also in RAM...
So RAM needs to be a few times larger than the size of the image.

>  My MacBook has a 2.13 GHz  
> Intel Core 2 Duo Processor and Memory of 2 GB 800 MHz DDR2 SDRAM.

so the biggest image you could probably try and deconvolve in imageJ is limited by the
2GB ram, less whatever the MacOSX is usiung for itself...in the ideal case.

Close all other applications and reboot the computer,
and only open imageJ or Fiji,
so the java VM can get that whole nearly 2GB  RAM in a single continuous chunk.

If the memory is already fragmented in to smaller bits, you will struggle.

Try DeconvolutionLab plugin from the BIG group in lausanne.
Its very good.
http://bigwww.epfl.ch/algorithms/deconvolutionlab/

Use a measured PSF (image og a 100 nm bead z stack)
or use Bob's Diffraction PSF 3D plugin to calculate a pretend 3D widefield PSF
and use that in the deconvolution.

http://www.optinav.com/Diffraction-PSF-3D.htm


>  I  
> am running Leopard (OS 10.5.8).  I have the ability to capture Z  
> series fluorescence microscopy data sets (Zeiss Photomicroscope III  

thats a very cool old microscope!
http://www.microscopy-uk.org.uk/mag//artnov07/dw-pm3.html

How are you doing fluorescence  on it? II dint see where the filter sets go? Do you have the epi accessories?

Deconvolution doesnt really work on transmitted light images... there are approaches for that... but
the tools above are strictly for fluorescence.


> with a Focus Controller and a digital camera), under the control of an  
> iMac (running under 10.4.11).  I already know the iMac does not have  
> enough computing power.  If anyone has suggestions regarding  
> possibilities using the MacBook, please contact me offline (unless you  
> think others would be interested).

the mac book will work fine... so long as the java VM doesnt run out of RAM.

so for small images (say 256x256x20) you might be fine,
but large images, eg full camera chip image and many z slices you will be in trouble...

then the only solution is a powerful 64 bit workstation with tens of GB or ram and multiple processors.

you need to make sure that the magnification is matched to the camera xy pixel spacing so that you
fulfill the Nyquist  sampling criterion, and that you z slice separation is also fulfilling that criterion, and its precise.

What objective are you using?

you can calculate the xyz pixel sizes needed to get data suitable doe deconvolution here:
http://www.svi.nl/NyquistCalculator
and read up more on that wiki, it is excellent for understanding how to do deconvolution microscopy.


>
> Thanks in advance!!

but I didn't help you before I wrote this email...

if you like you can retract the premature felicitations, and thank me now instead!!

cheers

Dan

>
> Mark R. Adelman
> [hidden email]

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 )

[Daniel James White]

Dear Elizabeth,

I thought I would cc this to the list, to share the words... hope that's ok.

On Mar 9, 2011, at 8:45 PM, Crowell Elizabeth wrote:

> Dear Dr. White,
>
> Thank you very much for sharing some very useful information on the ImageJ mailing list about deconvolution.
> I was happy to have your recommendation for using Deconvolution Lab.  It is indeed often difficult to choose a plugin from the many ones available!
> Although Deconvolution Lab looks promising for my purposes, the webpage gives minimal information on use.  I am not experienced enough to understand everything, I'm afraid!  Would you mind giving me a little advice?

Sure, why not contact the authors of the plugin directly... that are human and will respond to well formed questions.
I have met them, and they are good folks as well as experts in the methods.
>
> I would like to try calculating a theoretical PSF.  There is some ambiguity in the parameters to input into the PSF generator from the BIG lab.  For example, is "refractive index" the r.i. of the lens objective medium, or the sample mounting medium?  

which plugin exactly do you mean?
Indeed there are 2  RIs to take care of, and you correctly query which it asks for... better ask them or look for clarification in the docs.

> How do I calculate the spatial and axial resolution?  

this is the pixel separation in x y and z no?
You can only roughly calculate it from the camera pixel size and the nominal magnification of the system. This might be wrong by 10% or more.
Best to get a stage micrometer and measure it!

point scanning confocals are spatially calibrated and record the pixel separation in the meta data of the files they save, but scanners should also be calibrated,

> Is the "wavelength" the excitation or emission wavelength?

emission.
>
> I also have stacks of subresolution size beads that I would like to use to measure my PSF.  How can I obtain this measurement?  

the image of 1 bead is pretty much almost-ish  the PSF. So long as its really a sub resolution bead. You can align multiple images of beads and average them.
But if you get a single good bead with very good signal : noise, that will work.

Huygens software has a tool to make a psf from a bead image by removing the size of the bead and cleaning up the psf image.
API softWoRx does it too (OTF in this case)
Not sure i saw a tool like this for imageJ... if you did then please let me know!

For use as a psf, the bead image must be centered on the centre of the bead for it to work as a psf.

> I found information on a PSF 3D measurement plugin by Janick Cardinale, but it does not appear to be available for download in the list of ImageJ plugins, and I find no similar plugins.

can you share that info?
Also see the mosaic plugins from ETH (i haven't tested them yet though)

>
> I realize these must sound like very naive questions,

not at all...

> but it is the first time I try to calculate a PSF and I would like to understand what I am doing.  I hope you will be kind enough to answer, since I have some trouble finding help so far!  I have spent quite some time reading the excellent website of SVI, but my questions remain unfortunately unanswered.

feel free to ask more on the imagej list...

cheers


Dan

>
> Kind Regards,
>
> Elizabeth Crowell
>
>
>
>
>
> Daniel James White a écrit :
>> Hi Mark,
>>
>> On Mar 9, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:
>>
>>  
>>
>>> Date:    Tue, 8 Mar 2011 11:42:17 -0500
>>> From:    "Mark Adelman (Work)"
>>> <[hidden email]>
>>>
>>> Subject: Deconvolution on MacBook?
>>>
>>> Is it realistic to attempt to do a limited number of wide field  
>>> deconvolutions using ImageJ on a MacBook?
>>>    
>>>
>>
>> Yes, so long as the images are small enough to fit in your RAM with a copy or two also in RAM...
>> So RAM needs to be a few times larger than the size of the image.
>>
>>  
>>
>>>  My MacBook has a 2.13 GHz  
>>> Intel Core 2 Duo Processor and Memory of 2 GB 800 MHz DDR2 SDRAM.
>>>    
>>>
>>
>> so the biggest image you could probably try and deconvolve in imageJ is limited by the
>> 2GB ram, less whatever the MacOSX is usiung for itself...in the ideal case.
>>
>> Close all other applications and reboot the computer,
>> and only open imageJ or Fiji,
>> so the java VM can get that whole nearly 2GB  RAM in a single continuous chunk.
>>
>> If the memory is already fragmented in to smaller bits, you will struggle.
>>
>> Try DeconvolutionLab plugin from the BIG group in lausanne.
>> Its very good.
>>
>> http://bigwww.epfl.ch/algorithms/deconvolutionlab/
>>
>>
>> Use a measured PSF (image og a 100 nm bead z stack)
>> or use Bob's Diffraction PSF 3D plugin to calculate a pretend 3D widefield PSF
>> and use that in the deconvolution.
>>
>>
>> http://www.optinav.com/Diffraction-PSF-3D.htm
>>
>>
>>
>>  
>>
>>>  I  
>>> am running Leopard (OS 10.5.8).  I have the ability to capture Z  
>>> series fluorescence microscopy data sets (Zeiss Photomicroscope III  
>>>    
>>>
>>
>> thats a very cool old microscope!
>>
>> http://www.microscopy-uk.org.uk/mag//artnov07/dw-pm3.html
>>
>>
>> How are you doing fluorescence  on it? II dint see where the filter sets go? Do you have the epi accessories?
>>
>> Deconvolution doesnt really work on transmitted light images... there are approaches for that... but
>> the tools above are strictly for fluorescence.
>>
>>
>>  
>>
>>> with a Focus Controller and a digital camera), under the control of an  
>>> iMac (running under 10.4.11).  I already know the iMac does not have  
>>> enough computing power.  If anyone has suggestions regarding  
>>> possibilities using the MacBook, please contact me offline (unless you  
>>> think others would be interested).
>>>    
>>>
>>
>> the mac book will work fine... so long as the java VM doesnt run out of RAM.
>>
>> so for small images (say 256x256x20) you might be fine,
>> but large images, eg full camera chip image and many z slices you will be in trouble...
>>
>> then the only solution is a powerful 64 bit workstation with tens of GB or ram and multiple processors.
>>
>> you need to make sure that the magnification is matched to the camera xy pixel spacing so that you
>> fulfill the Nyquist  sampling criterion, and that you z slice separation is also fulfilling that criterion, and its precise.
>>
>> What objective are you using?
>>
>> you can calculate the xyz pixel sizes needed to get data suitable doe deconvolution here:
>>
>> http://www.svi.nl/NyquistCalculator
>>
>> and read up more on that wiki, it is excellent for understanding how to do deconvolution microscopy.
>>
>>
>>  
>>
>>> Thanks in advance!!
>>>    
>>>
>>
>> but I didn't help you before I wrote this email...
>>
>> if you like you can retract the premature felicitations, and thank me now instead!!
>>
>> cheers
>>
>> Dan
>>
>>  
>>
>>> Mark R. Adelman
>>>
>>> [hidden email]
>>>
>>>    
>>>
>>
>> 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 )
>>  
>>
>
>
> --
>
> 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 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]

Dan White a écrit :
> Dear Elizabeth,
> Sure, why not contact the authors of the plugin directly... that are human and will respond to well formed questions.
> I have met them, and they are good folks as well as experts in the methods.
>  
Yes, I have contacted them as well, but it is always interesting to get
input from users as well as the developers!

>> I would like to try calculating a theoretical PSF.  There is some ambiguity in the parameters to input into the PSF generator from the BIG lab.  For example, is "refractive index" the r.i. of the lens objective medium, or the sample mounting medium?  
>>    
>
> which plugin exactly do you mean?
> Indeed there are 2  RIs to take care of, and you correctly query which it asks for... better ask them or look for clarification in the docs.
>  
I am referring to the PSF generator plugin at
http://bigwww.epfl.ch/deconvolution//?p=plugins
The documentation amounts to: Select the parameters and click on
"Generate PSF".  I am not joking!  So I contacted the author with my
specific questions.

> this is the pixel separation in x y and z no?
> You can only roughly calculate it from the camera pixel size and the nominal magnification of the system. This might be wrong by 10% or more.
> Best to get a stage micrometer and measure it!
>  
Okay, I have done this already.

>> Is the "wavelength" the excitation or emission wavelength?
>>    
> emission.
>  
>> I also have stacks of subresolution size beads that I would like to use to measure my PSF.  How can I obtain this measurement?  
>>    
>
> the image of 1 bead is pretty much almost-ish  the PSF. So long as its really a sub resolution bead. You can align multiple images of beads and average them.
> But if you get a single good bead with very good signal : noise, that will work.
>  

This is a strange concept for me, that a "function" could be equated to
an image stack!  But I suppose the mathematical function itself is
calculated from the image stack by the software?  After all the
information I have read, PSF still remains a vague term for me,
sometimes used to describe an image, sometimes used to describe the
diffraction pattern of light around an object, sometimes used to
describe the mathematical function that will be applied to the raw image
to deconvolve it.

> Huygens software has a tool to make a psf from a bead image by removing the size of the bead and cleaning up the psf image.
> API softWoRx does it too (OTF in this case)
> Not sure i saw a tool like this for imageJ... if you did then please let me know!
>
>  
Yes, it is the plugin by Janick Cardinale from the Mosaic lab.  You can
find the pdf manual at
www.mosaic.ethz.ch/Downloads/ParticleTracker/PSF_measurement_3D.pdf
But I cannot find the plugin anywhere!  I wrote to the author.

To be continued...  thanks again,
Elizabeth

> For use as a psf, the bead image must be centered on the centre of the bead for it to work as a psf.
>
>  
>> I found information on a PSF 3D measurement plugin by Janick Cardinale, but it does not appear to be available for download in the list of ImageJ plugins, and I find no similar plugins.
>>    
>
> can you share that info?
> Also see the mosaic plugins from ETH (i haven't tested them yet though)
> feel free to ask more on the imagej list...
>
> cheers
>
>
> Dan
>
>
>  

[Daniel James White]

Hi Elizabeth,
>>
>> which plugin exactly do you mean?
>> Indeed there are 2  RIs to take care of, and you correctly query which it asks for... better ask them or look for clarification in the docs.
>>  
>>
> I am referring to the PSF generator plugin at http://bigwww.epfl.ch/deconvolution//?p=plugins
> The documentation amounts to: Select the parameters and click on "Generate PSF".  I am not joking!  So I contacted the author with my specific questions.

OK, so there is that one, and also Bob's Diffraction PSD 3D plugin.

They give similar but slightly different PSFs for the same input parameters,
as they must use a different model to generate the image....?
>>
>>> I also have stacks of subresolution size beads that I would like to use to measure my PSF.  How can I obtain this measurement?  
>>>    
>> the image of 1 bead is pretty much almost-ish  the PSF. So long as its really a sub resolution bead. You can align multiple images of beads and average them.
>> But if you get a single good bead with very good signal : noise, that will work.
>>  
>>
> This is a strange concept for me, that a "function" could be equated to an image stack!  But I suppose the mathematical function itself is calculated from the image stack by the software?  After all the information I have read, PSF still remains a vague term for me, sometimes used to describe an image, sometimes used to describe the diffraction pattern of light around an object, sometimes used to describe the mathematical function that will be applied to the raw image to deconvolve it.

you are right, it is odd to talk about a discretized image ( a set of point values) as a function... its just the jargon of the field.
Physically, it is a smooth function, but we have to deal with a discrete representation of it... an "image", in order to do the deconvolution.
I understand your confusion, and it is well grounded
>
>> Huygens software has a tool to make a psf from a bead image by removing the size of the bead and cleaning up the psf image.
>> API softWoRx does it too (OTF in this case)
>> Not sure i saw a tool like this for imageJ... if you did then please let me know!
>>
> Yes, it is the plugin by Janick Cardinale from the Mosaic lab.  You can find the pdf manual at www.mosaic.ethz.ch/Downloads/ParticleTracker/PSF_measurement_3D.pdf
> But I cannot find the plugin anywhere!  I wrote to the author.

I has a quick look... this is not what you need, as you want to measure the PSF in 3D,
but this one takes bead images and averaved them axially, to make a 2D PSF that is a rotational projection around the axis. You dont want that,

What you want is a gadget likje the Huygens PSF distiller that makes 3D PSFs from real bead images....

but you can use an image of a bead that is 100nm AS the PSF.


cheers

Dan

>
> To be continued...  thanks again,
> Elizabeth
>
>
>
>> For use as a psf, the bead image must be centered on the centre of the bead for it to work as a psf.
>>
>>  
>>
>>> I found information on a PSF 3D measurement plugin by Janick Cardinale, but it does not appear to be available for download in the list of ImageJ plugins, and I find no similar plugins.
>>>    
>>>
>>
>> can you share that info?
>> Also see the mosaic plugins from ETH (i haven't tested them yet though)
>> feel free to ask more on the imagej list...
>>
>> cheers
>>
>>
>> Dan
>>
>>
>>  
>>
>

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 )

[lechristophe]

Hi,

What I did once (I was using Volocity at the time) was to generate a virtual
PSF (you can do that with Bob Dougherty's plugin or the EPFL one), but use a
real empirical PSF from my setup to infer the real NA to feed the PSF
generation.

To do that, just measure the angle of the PSF cone in XZ, like explained
here : http://www.svi.nl/NumericalAperture

So you get the advantages of a calculated PSF (noise-free), but with an
overall shape that corresponds to your setup. Surpirisngly, the objective
was 1.3 but the measured NA from the cone anlgle was around 1.1 !

Anyway, the svi website is a great resource for all things deconvolution.

Cheers,

Christophe


On Thu, Mar 10, 2011 at 15:43, Daniel James White <[hidden email]> wrote:

> Hi Elizabeth,
> >>
> >> which plugin exactly do you mean?
> >> Indeed there are 2  RIs to take care of, and you correctly query which
> it asks for... better ask them or look for clarification in the docs.
> >>
> >>
> > I am referring to the PSF generator plugin at
> http://bigwww.epfl.ch/deconvolution//?p=plugins
> > The documentation amounts to: Select the parameters and click on
> "Generate PSF".  I am not joking!  So I contacted the author with my
> specific questions.
>
> OK, so there is that one, and also Bob's Diffraction PSD 3D plugin.
>
> They give similar but slightly different PSFs for the same input
> parameters,
> as they must use a different model to generate the image....?
> >>
> >>> I also have stacks of subresolution size beads that I would like to use
> to measure my PSF.  How can I obtain this measurement?
> >>>
> >> the image of 1 bead is pretty much almost-ish  the PSF. So long as its
> really a sub resolution bead. You can align multiple images of beads and
> average them.
> >> But if you get a single good bead with very good signal : noise, that
> will work.
> >>
> >>
> > This is a strange concept for me, that a "function" could be equated to
> an image stack!  But I suppose the mathematical function itself is
> calculated from the image stack by the software?  After all the information
> I have read, PSF still remains a vague term for me, sometimes used to
> describe an image, sometimes used to describe the diffraction pattern of
> light around an object, sometimes used to describe the mathematical function
> that will be applied to the raw image to deconvolve it.
>
> you are right, it is odd to talk about a discretized image ( a set of point
> values) as a function... its just the jargon of the field.
> Physically, it is a smooth function, but we have to deal with a discrete
> representation of it... an "image", in order to do the deconvolution.
> I understand your confusion, and it is well grounded
> >
> >> Huygens software has a tool to make a psf from a bead image by removing
> the size of the bead and cleaning up the psf image.
> >> API softWoRx does it too (OTF in this case)
> >> Not sure i saw a tool like this for imageJ... if you did then please let
> me know!
> >>
> > Yes, it is the plugin by Janick Cardinale from the Mosaic lab.  You can
> find the pdf manual at
> www.mosaic.ethz.ch/Downloads/ParticleTracker/PSF_measurement_3D.pdf
> > But I cannot find the plugin anywhere!  I wrote to the author.
>
> I has a quick look... this is not what you need, as you want to measure the
> PSF in 3D,
> but this one takes bead images and averaved them axially, to make a 2D PSF
> that is a rotational projection around the axis. You dont want that,
>
> What you want is a gadget likje the Huygens PSF distiller that makes 3D
> PSFs from real bead images....
>
> but you can use an image of a bead that is 100nm AS the PSF.
>
>
> cheers
>
> Dan
>
>
> >
> > To be continued...  thanks again,
> > Elizabeth
> >
> >
> >
> >> For use as a psf, the bead image must be centered on the centre of the
> bead for it to work as a psf.
> >>
> >>
> >>
> >>> I found information on a PSF 3D measurement plugin by Janick Cardinale,
> but it does not appear to be available for download in the list of ImageJ
> plugins, and I find no similar plugins.
> >>>
> >>>
> >>
> >> can you share that info?
> >> Also see the mosaic plugins from ETH (i haven't tested them yet though)
> >> feel free to ask more on the imagej list...
> >>
> >> cheers
> >>
> >>
> >> Dan
> >>
> >>
> >>
> >>
> >
>
> 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 )
>