deconvolution in MATLAB

[Shalin Mehta]

Search the CONFOCAL archive at http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal I would agree with Nathan,

 MATLAB has blind deconvolution which contrary to usual belief can work on N-dimensional data (http://blogs.mathworks.com/steve/2008/03/17/multidimensional-image-processing/#comment-20530).

It will require that you generate a 3D intensity distribution like PSF.  deconvolution algorithms are more sensitive to shape and size of PSF rather than actual values, so even specifying a 3D pattern of 1s should give a good start with blind deconvolution algorithm.

Best
shalin

On Tue, Apr 1, 2008 at 4:17 AM, Nathan <[hidden email]> wrote:

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

   If your university/department has a license, you might find Matlab's deconvblind, deconvwnr, deconvreg, and deconvlucy functions useful. They require basic Matlab skill and some knowledge about how to generate appropriate input PSFs, but I've used them successfully a few times.

Best,
Nate


Nathan O'Connor
Graduate Student
Physiology and Biophysics
Weill Medical College of Cornell University
NY, NY 10021

On Mon, Mar 31, 2008 at 3:40 PM, John Oreopoulos <[hidden email]> wrote:

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

Does anyone know of any freely available software that can deconvolve image data? I am only aware of one ImageJ deconvolution plugin that does a reasonable job under certain circumstances. I'd be interested to know if anyone has created any others.

John Oreopoulos, BSc,

PhD Candidate

University of Toronto

Institute For Biomaterials and Biomedical Engineering

Centre For Studies in Molecular Imaging

Tel: W:416-946-5022

On 31-Mar-08, at 3:02 PM, Mayandi Sivaguru wrote:

Search the CONFOCAL archive at http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
Valeria, my understanding is that you will be better off with deconvolving all your optical microscope data sets (widefield, confocal and etc) in general. With reference to colocalization analysis, you first sample the data following sequential scans (never simultaneous for the coloc analysis) Nyquist sampling in 3D (I would personally suggest a bit over sampling won't hurt, if you do not experience significant photobleaching), and then a deconvolution is a must with a plane by plane analysis.
Deconvolution will not change a "non-cocolalizing" data points in to "colocalizing" data points. But it can be otherwise, a colocaizing data points in raw data could become in fact not colocalizing anymore after deconvolution. But the parameters affecting your conlusion greatly is at much before you deconve the data i.e., the sample preparation, fixation, blocking, selection of antibodies, fluorophores, scan parameters and so on.
Shiv
   

At 10:05 AM 3/31/2008, you wrote:

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

Hi,

This question just fit in perfectly on what I am trying to find out about
colocalization.

When and why do I need do deconvolve pictures collected with a confocal in
order to be sure about my colocalization (or not colocalization) results?

To be specific: I am working on pre and post-synaptic proteins.

Thanks

Valeria



> Search the CONFOCAL archive at
> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>
> Colocalization based upon "yellow" could be accurate, if and only if,
> the intensities are comparable and pixel (voxel) quantities in the
> suspected colocalized volumes are in roughly equal.  .  Otherwise,
> the yellow is masked by the predominate channel.  Something small,
> like lysosomes, would need to be sampled properly. Colocalization
> could be masked by blur unless deconvolved, even if images are
> collected with a confocal.
> On Feb 7, 2007, at 1:05 PM, Marc Thibault wrote:
>
>> Search the CONFOCAL archive at
>> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>>
>> Hi all,
>>
>> It seems that in many papers from biologists or chemists, and i'm
>> talking
>> high impact factors journals,  colocalisation of two elements is is
>> often
>> assumed  by simple color superposition (ex: red and green fluoresce
>> yellow
>> when colocalising), while microscopists (many physisists I suppose)
>> seem to
>> need a more complex software-based confirmation.
>> Is it ok, when using high end equipment and corrected objectives
>> (apochromat
>> with high NA for ex.), to assume colocalisation by color
>> superposition,
>> especially when fluorophore are confined to small volume entities,
>> like
>> lysosomes ?
>>
>> Thanks
>>
>> Marc
>

Mayandi Sivaguru, PhD, PhD
Microscopy Facility Manager
8, Institute for Genomic Biology
University of Illinois at Urbana-Champaign
1206 West Gregory Dr.
Urbana, IL 61801 USA

Office: 217.333.1214
Fax: [phone]
[hidden email]
http://core.igb.uiuc.edu

--
~~~~~~~~~~~~~~~~~~~~~~~~~
Shalin Mehta
mobile: [phone]
blog: shalin.wordpress.com
~~~~~~~~~~~~~~~~~~~~~~~~~~
Bioimaging Lab, Block-E3A, #7-10
Div of Bioengineering, NUS Singapore 117574
website: http://www.bioeng.nus.edu.sg/optbioimaging/colin/index.html

Liver Cancer Functional Genomics Lab, #6-05
National Cancer Centre, Singapore 169610
~~~~~~~~~~~~~~~~~~~~~~~~~~~

[M. van de corput]

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

I am not too enthousiastic about blind deconvolution. When the psf is
rather distorted your object <PSF look like the actual PSF = distorted, so
not a nice american football. The centroids will be shifted likewise and
can influence the co-localization analysis (when it's based on centroids
that is). Also it can generate some very weird deconvolution artefacts.

As you are scanning you perfect images of your sample it is not too much
more trouble to do few more scans: to determine the psf and the chromatic
shift. Invitrogen (no commerical interest) used to sell bead-samples ready
for use (don't know if they still sell them). But it is easy to make your
own bead-samples.

Mariette

Dr. Kemner-van de Corput
MGC - Dept. of Cell Biology & Genetics
Erasmus Medical Center
Dr. Molewaterplein 50, 3015 GE Rotterdam
POB 2040, 3000 CA Rotterdam, The Netherlands



Op Di, 1 april, 2008 3:10 am, schreef Shalin Mehta:

> Search the CONFOCAL archive at
> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>
> I would agree with Nathan,
>
>  MATLAB has blind deconvolution which contrary to usual belief can work on
> N-dimensional data (
> http://blogs.mathworks.com/steve/2008/03/17/multidimensional-image-processing/#comment-20530).
>
>
> It will require that you generate a 3D intensity distribution like PSF.
> deconvolution algorithms are more sensitive to shape and size of PSF
> rather
> than actual values, so even specifying a 3D pattern of 1s should give a
> good
> start with blind deconvolution algorithm.
>
> Best
> shalin
>
> On Tue, Apr 1, 2008 at 4:17 AM, Nathan <[hidden email]> wrote:
>
>> Search the CONFOCAL archive at
>> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal Hi Jon,
>>
>>    If your university/department has a license, you might find Matlab's
>> deconvblind, deconvwnr, deconvreg, and deconvlucy functions useful. They
>> require basic Matlab skill and some knowledge about how to generate
>> appropriate input PSFs, but I've used them successfully a few times.
>>
>> Best,
>> Nate
>>
>>
>> Nathan O'Connor
>> Graduate Student
>> Physiology and Biophysics
>> Weill Medical College of Cornell University
>> NY, NY 10021
>>
>>
>> On Mon, Mar 31, 2008 at 3:40 PM, John Oreopoulos <
>> [hidden email]> wrote:
>>
>> > Search the CONFOCAL archive at
>> > http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>> >
>> > Does anyone know of any freely available software that can deconvolve
>> > image data? I am only aware of one ImageJ deconvolution plugin that
>> does a
>> > reasonable job under certain circumstances. I'd be interested to know
>> if
>> > anyone has created any others.
>> >
>> >
>> > John Oreopoulos, BSc,
>> > PhD Candidate
>> > University of Toronto
>> > Institute For Biomaterials and Biomedical Engineering
>> > Centre For Studies in Molecular Imaging
>> >
>> > Tel: W:416-946-5022
>> >
>> >
>> > On 31-Mar-08, at 3:02 PM, Mayandi Sivaguru wrote:
>> >
>> > Search the CONFOCAL archive at
>> > http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>> > Valeria, my understanding is that you will be better off with
>> > deconvolving all your optical microscope data sets (widefield,
>> confocal and
>> > etc) in general. With reference to colocalization analysis, you first
>> sample
>> > the data following sequential scans (never simultaneous for the coloc
>> > analysis) Nyquist sampling in 3D (I would personally suggest a bit
>> over
>> > sampling won't hurt, if you do not experience significant
>> photobleaching),
>> > and then a deconvolution is a must with a plane by plane analysis.
>> > Deconvolution will not change a "non-cocolalizing" data points in to
>> > "colocalizing" data points. But it can be otherwise, a colocaizing
>> data
>> > points in raw data could become in fact not colocalizing anymore after
>> > deconvolution. But the parameters affecting your conlusion greatly is
>> at
>> > much before you deconve the data i.e., the sample preparation,
>> fixation,
>> > blocking, selection of antibodies, fluorophores, scan parameters and
>> so on.
>> > Shiv
>> >
>> >
>> > At 10:05 AM 3/31/2008, you wrote:
>> >
>> > Search the CONFOCAL archive at
>> >  http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>> >
>> > Hi,
>> >
>> > This question just fit in perfectly on what I am trying to find out
>> > about
>> > colocalization.
>> >
>> > When and why do I need do deconvolve pictures collected with a
>> confocal
>> > in
>> > order to be sure about my colocalization (or not colocalization)
>> > results?
>> >
>> > To be specific: I am working on pre and post-synaptic proteins.
>> >
>> > Thanks
>> >
>> > Valeria
>> >
>> >
>> >
>> > > Search the CONFOCAL archive at
>> > > http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>> > >
>> > > Colocalization based upon "yellow" could be accurate, if and only
>> if,
>> > > the intensities are comparable and pixel (voxel) quantities in the
>> > > suspected colocalized volumes are in roughly equal.  .  Otherwise,
>> > > the yellow is masked by the predominate channel.  Something small,
>> > > like lysosomes, would need to be sampled properly. Colocalization
>> > > could be masked by blur unless deconvolved, even if images are
>> > > collected with a confocal.
>> > > On Feb 7, 2007, at 1:05 PM, Marc Thibault wrote:
>> > >
>> > >> Search the CONFOCAL archive at
>> > >> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>> > >>
>> > >> Hi all,
>> > >>
>> > >> It seems that in many papers from biologists or chemists, and i'm
>> > >> talking
>> > >> high impact factors journals,  colocalisation of two elements is is
>> > >> often
>> > >> assumed  by simple color superposition (ex: red and green fluoresce
>> > >> yellow
>> > >> when colocalising), while microscopists (many physisists I suppose)
>> > >> seem to
>> > >> need a more complex software-based confirmation.
>> > >> Is it ok, when using high end equipment and corrected objectives
>> > >> (apochromat
>> > >> with high NA for ex.), to assume colocalisation by color
>> > >> superposition,
>> > >> especially when fluorophore are confined to small volume entities,
>> > >> like
>> > >> lysosomes ?
>> > >>
>> > >> Thanks
>> > >>
>> > >> Marc
>> > >
>> >
>> >  Mayandi Sivaguru, PhD, PhD
>> > Microscopy Facility Manager
>> > 8, Institute for Genomic Biology
>> > University of Illinois at Urbana-Champaign
>> > 1206 West Gregory Dr.
>> > Urbana, IL 61801 USA
>> >
>> > Office: 217.333.1214
>> > Fax: [phone]
>> > [hidden email]
>> >  http://core.igb.uiuc.edu
>> >
>> >
>> >
>> >
>>
>>
>
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~
> Shalin Mehta
> mobile: [phone]
> blog: shalin.wordpress.com
> ~~~~~~~~~~~~~~~~~~~~~~~~~~
> Bioimaging Lab, Block-E3A, #7-10
> Div of Bioengineering, NUS Singapore 117574
> website: http://www.bioeng.nus.edu.sg/optbioimaging/colin/index.html
>
> Liver Cancer Functional Genomics Lab, #6-05
> National Cancer Centre, Singapore 169610
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~
>