deconvolution software

[David Stuss]

Hello all,

Our lab is considering purchasing some deconvolution software for cleaning up confocal image stacks. I'm soliciting any recommendations, pro or con, for particular software packages, based on quality of results, ease of use, cost etc. If you prefer to send your comments directly rather than post them, please reach me at the email below. 

Thanks very much for your consideration!

Cheers, 

David Stuss

[hidden email]

[Stanislav Vitha]

Hallo,
I have used three software packages:

Media Cybernetics (fromerly AutoQuant) AutoDeblur X - I used it mostly for
wide-field deconvolution. I like the Blind deconvolution option, since the
average user does not need to know all the details about the optical setup
and the sample, and also use the option to estimate spherical aberration
from the image stack.  You can also use the theoretical or experimental PSF.
In my experience the software works well if you have just few datasets and
want to process individual stacks one by one. The actual 3D deconvolution
can be done in a batch, but the preprocessing steps, like attenuation
correction, and estimation of spherical aberration cannot be scripted and
must be done interactively for each stack. I had about forty widefield
z-stacks to process from one afternoon of imaging and the lack of scripting
was a big hindrance, it took two days to babysit the preprocessing and
deconvolution.  I would not want to do it again this way.  I only have the
deconvolution portion of the software, without the visualization tools
(AutoVisualize), or stack alignment tools, so I cannot comment on those
features.

SVI Huygens Essentials/Huygens Pro - I used the 1-month evaluation license
and tried it on relatively large confocal datasets (Vitha et al, 3D Confocal
Imaging of Pollen. Microscopy Today,  18(2)  Mar 2010,  pp 26 - 28)
 Overall I like the user interface better than the AutoDeblur's; I prefer
the Huygens Pro, but the Essentials version may be easier for new users. I
love the visualization/measurement options, but I did not directly compare
them with Autodeblur/Autovisualize suite mentioned above.  I believe Huygens
allows full scripting, so processing large number of stacks without
intervention should be possible. On the other hand, there is no blind
deconvolution algorithm, you use either a theoretical or experimental point
spread function. Also, I do not remember if you can estimate spherical
aberration from the dataset itself, or if it is only calculated from your
input parameters (RI of the mounting medium, RI of the immersion fluid,
depth of imaging).

XCOSM  (freeware, runs on Linux; also can run the EM algorithm in command
line mode in Windows) - works; is scriptable, but unless you are familiar
with Linux, you may need somebody to install and compile it. It requires the
input data stack to be converted to a specific format, using command line
routines, and byte swapping for 16-bpp datasets (from big Endian to Little
Endian byte order). The EM algorithms provide high quality output, but the
convergence is very slow, it requires hundreds or thousands of iterations.
The results were comparable or subjectively somewhat better than those from
AutoDeblur's (I only tried wide-field data sets).
 
I have briefly tested 3D deconvolution in Amira, but there is definitely a
learning curve for Amira's way of linking the data with their processing. So
I got discouraged quickly  - I think you actually have to read the user
manual to know what to do.
 
I have not tried (yet) the new DECONV freeware package (Sun Y, Davis P,
Kosmacek EA, Ianzini F, Mackey MA., An open-source deconvolution software
package for 3-D quantitative fluorescence microscopy imaging. J Microsc.
2009 Dec;236(3):180-93.)

AutoDeblur and Huygens are modular, so the cost will depend on exactly what
options you get. I would expect to pay between $6k and 10k for a
confocal-only package that does not include the scripting or surface
rendering options.


Sincerely,

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

On Mon, 29 Mar 2010 21:53:16 -0700, David Stuss <[hidden email]> wrote:

>Hello all,
>
>Our lab is considering purchasing some deconvolution software for cleaning
>up confocal image stacks. I'm soliciting any recommendations, pro or con,
>for particular software packages, based on quality of results, ease of use,
>cost etc. If you prefer to send your comments directly rather than post
>them, please reach me at the email below.
>
>Thanks very much for your consideration!
>
>Cheers,
>
>David Stuss
>
>[hidden email]
>

[Vincent Schoonderwoert]

*commercial interest*

Hello,

Thank you Stan for your nice overview.
I would like to add the following to what you wrote about our Huygens software.

I believe Huygens allows full scripting, so processing large number of stacks without
intervention should be possible. 

Both the Huygens Essential and Professional have a batch processor to schedule and automate large numbers of deconvolution tasks .

The Huygens Professional, like the Scripting (a package that was not mentioned), has indeed full Tcl-based scripting possibilities for processing large file series.

Then, we have also a web-based version which is the Huygens Core. The Core can work with the Huygens Remote Manager (HRM; open source) as the web-interface to schedule deconvolution tasks.

On the other hand, there is no blind deconvolution algorithm, you use either a theoretical or experimental point
spread function. Also, I do not remember if you can estimate spherical
aberration from the dataset itself, or if it is only calculated from your
input parameters (RI of the mounting medium, RI of the immersion fluid,
depth of imaging).
  

Like the theoretical PSF, the spherical aberration (SA) is not estimated from the image but based on the known microscopical and sample parameters.

We have recently added the possibility to set the distance of the first image plane to the coverslip, since the SA progressively worsens when imaging further from the coverslip.

AutoDeblur and Huygens are modular, so the cost will depend on exactly what
options you get. I would expect to pay between $6k and 10k for a
confocal-only package that does not include the scripting or surface
rendering options.
  

The Huygens Essential (32bit/64bit) with a confocal deconvolution option and the batch processor is available below this range.

Best regards,

Vincent Schoonderwoert

Hello all,

Our lab is considering purchasing some deconvolution software for cleaning
up confocal image stacks. I'm soliciting any recommendations, pro or con,
for particular software packages, based on quality of results, ease of use,
cost etc. If you prefer to send your comments directly rather than post
them, please reach me at the email below.

Thanks very much for your consideration!

Cheers,

David Stuss

[hidden email]