Re: Addition to post on "confocal detectors and deconvolution"

Posted by Brian Northan on
URL: http://confocal-microscopy-list.275.s1.nabble.com/confocal-detectors-and-deconvolution-tp7588223p7588256.html

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Allison

You raise some really good points.  I've been developing image analysis
algorithms for 15+ years, and I'm always suspect of marketing, that has a
lot of vague, "feel good", and subjective terms.  The description of an
image processing algorithm, should convey enough information, for example
algorithms names and references, so that someone with a background in
signal processing, can do a little research and understand what the
algorithm does.

As we all know, some deconvolution softwares are more quantitative than
> others.


As far as I am aware, little work has been done in this area.  Especially
in terms of evaluating morphology of results, and PSF models.  What is
needed is for a group of people, to evaluate the properties of the
fundamental implementations of deconvolution and PSF generation algorithms
(reference implementations are available from a variety of open sources
such as DeconvolutionLab2,
http://bigwww.epfl.ch/deconvolution/deconvolutionlab2/) and then evaluate
whether Vendor implementations have the same properties.

Going back to the "Lightning" product, using descriptive algorithm
terminology and references, to describe the product, would go a long way to
helping people understand and evaluate it properly.  According to an answer
on Leica's facebook page, it seems Lightning does use deconvolution,(
https://www.facebook.com/LeicaMicrosystems/photos/a.145939838807420.
34521.135295379871866/1848698365198217/?type=3)

As an algorithm developer I would be curious, exactly what they are doing.
Are they doing multiple overlapping deconvolutions, with different
parameters and blending them??  Are they using an spatially varying PSF??
Are they perhaps using recent techniques in machine learning (deep learning
is a hot topic these days) to further enhance the image??  I don't really
know, and can't begin to even try to understand what they are doing, based
on the vague marketing terminology used on their web site.

I think promoting a culture, where more importance is placed on the
technical names, details and properties of algorithms would help in
addressing some of these issues.  If users of image processing software,
can be educated on more of the nuts of bolts of image processing, it would
help them ask vendors the right questions, and put pressure on vendors to
give descriptive documentation

Brian


On Fri, May 11, 2018 at 9:35 AM, Alison North <[hidden email]>
wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Dear all,
>
> I am a big fan of carefully performed deconvolution, and we have used
> Huygens, Autoquant and GE's SoftWoRx deconvolution packages on many types
> of microscope images (widefield, confocal, spinning disk, light sheet etc.)
> with great success.  However, I'm not sure I agree fully with George's
> comment that every confocal microscope should have deconvolution running
> immediately downstream, unless the manufacturers are going to do a better
> job of educating every single user about what goes on in their black box of
> deconvolution.  As we all know, some deconvolution softwares are more
> quantitative than others.  I am particularly concerned after a talk I just
> heard about the new Leica LIGHTNING software, which uses different
> parameters for each voxel in the image.  Here is a direct quote from
> Leica's web page:
>
> "Maximize the information you extract from your specimen and get in-depth
> answers to scientific questions with the LIGHTNING detection package for
> image information extraction.
>
> LIGHTNING fully automatically detects the finest structures and details,
> which are otherwise simply not visible. The key technology of LIGHTNING is
> an adaptive process for extraction of hidden information in the image.
> Unlike traditional technologies, that use a global set of parameters for
> the full image, *LIGHTNING calculates an appropriate set of parameters for
> each voxel *to uncover every detail with the highest fidelity."
>
> I think we need to impress upon the microscope manufacturers very strongly
> that we are in the business of collecting quantitative scientific data, not
> just pretty pictures.  At their presentation of this new software at the
> ABRF meeting last week, I asked whether it would be possible for them to
> flash up a big red warning on the screen - "Pretty picture ONLY!" - every
> time somebody uses this operation, to warn the user that this is a
> non-quantitative operation and therefore the user will never be able to
> perform any kind of quantitative image analysis on those images.  I'm not
> sure they took my comment seriously, but I was indeed being perfectly
> serious!  I also think the same should be implemented for any kind of
> operation that turns a quantitative raw image into one that is not, such as
> the use of the High Dynamic Range button on some of the confocal systems.
> Since many journals require authors to declare any nonlinear operations to
> images, including gamma adjustments, it's critical for every researcher to
> be aware of any such operations.  It may be acceptable to perform this kind
> of operation after the fact, as long as one understands what has been done
> and states it clearly in the methods or figure legend, but I am very
> nervous about acquiring RAW data that has already been changed in a
> nonlinear way.
>
> I'm not meaning to pick on Leica - they are not the only ones with evil
> buttons and options in their software! - but I do believe we need to point
> out how worrying this trend is.
>
> Best,
> Alison
>
>
>
> On 5/11/2018 8:55 AM, Vincent Schoonderwoert wrote:
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.um
>> n.edu_cgi-2Dbin_wa-3FA0-3Dconfocalmicroscopy&d=DwIDaQ&c=JeTk
>> UgVztGMmhKYjxsy2rfoWYibK1YmxXez1G3oNStg&r=RBx0-WJrAO5vwSOLNm
>> FbqYvikvIZS5ns3-USwvMOuLo&m=otc_bgF6TO_iCl4lyz-BEkOTqW9EsR
>> GuiTXpZHmRB3U&s=AR4nk9u2v55TucTfbtDFWmLZpC8zt70fMkhWmC9TtIQ&e=
>> Post images on https://urldefense.proofpoint.
>> com/v2/url?u=http-3A__www.imgur.com&d=DwIDaQ&c=JeTkUgVztGMmh
>> KYjxsy2rfoWYibK1YmxXez1G3oNStg&r=RBx0-WJrAO5vwSOLNmFbqYvikvI
>> ZS5ns3-USwvMOuLo&m=otc_bgF6TO_iCl4lyz-BEkOTqW9EsRGuiTXpZHmRB
>> 3U&s=9p6_6YMvYBgHgIKwAUK46-MDCP_Rsm-V8WV0zaFIGE4&e= and include the link
>> in your posting.
>> *****
>>
>> ***Vendor message***
>>
>>
>> Dear all,
>>
>>
>> Without getting involved in a discussion about the need for deconvolution
>> (although we are tempted), we feel a profound need to inform you about the
>> status of the collaboration between SVI-Huygens and Leica to prevent any
>> misunderstanding if you are considering a new confocal microscope.
>>
>> Per December 1, 2017,  despite its success, Leica stopped selling
>> Hyvolution2 with Huygens as its computing engine. As George kindly
>> mentioned, Hyvolution2 includes "the SVI Huygens software under the hood".
>> New Leica customers are now offered another option in LASX that does NO
>> LONGER involve the Huygens software and SVI.
>>
>> We will maintain our high level of support to customers who already
>> purchased a Huygens license via Hyvolution or directly with their SP8. Like
>> before, we also keep offering Huygens for the latest microscope types and
>> file formats.  Also the new and fully automated on-the-fly deconvolution
>> tool "Batch Express" will offer this functionality.
>>
>> We welcome any further questions and will respond offline.
>>
>> Apologies to Mike for using your interesting post.
>>
>> With kind regards,
>> Vincent
>>
>>
>> Get the best out of your microscopy images with Huygens Batch Express:
>>
>> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.svi.
>> nl_BatchExpress&d=DwIDaQ&c=JeTkUgVztGMmhKYjxsy2rfoWYibK1YmxX
>> ez1G3oNStg&r=RBx0-WJrAO5vwSOLNmFbqYvikvIZS5ns3-USwvMOuLo&m=
>> otc_bgF6TO_iCl4lyz-BEkOTqW9EsRGuiTXpZHmRB3U&s=EAEoN_
>> a12grsO5QrEbCKFIa1T8P_LjFnR_rzGasiUpk&e=
>>
>> ***********************************************
>> Vincent Schoonderwoert, PhD
>> Senior Imaging Specialist/Account Manager
>> Scientific Volume Imaging
>> www.svi.nl
>> [phone]
>> ***********************************************
>>
>
> --
> Alison J. North, Ph.D.,
> Research Associate Professor and
> Senior Director of the Bio-Imaging Resource Ce
> <https://maps.google.com/?q=irector+of+the+Bio-Imaging+Resource+Ce&entry=gmail&source=g>
> nter,
> The Rockefeller University,
> 1230 York Avenue,
> New York,
> NY 10065.
> Tel: office     ++ 212 327 7488
> Tel: lab        ++ 212 327 7486
> Fax:            ++ 212 327 7489
>