*****
To join or leave the confocal microscopy listserv or to change your email address, go to:
https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
Post images on http://www.imgur.com and include the link in your posting.
*****
Following this thread for a while, I feel compelled to question how a random study of the performance of various PSF generation- and deconvolution algorithms would be helpful? Randomized studies are relevant in life science to find statistical distributions of otherwise indeterminate behavior. Except for machine learning, algorithms are developed to reconstruct sample representations based on strictly deterministic mathematical models. If I understand correctly, this "reproducibility determination study" is based on how a number of available software implementations perform without considering the developmental background among them, notwithstanding even if they are buggy or contain systemic faults. I do understand however that there is a need for scientists using tools from other scientific areas and communities in knowing which implementations perform best. In that case, the random part should not be the implementations of algorithms because they are within themselves at least somewhat deterministic...
Regards
Lutz
-----Original Message-----
From: Confocal Microscopy List <[log in to unmask]> On Behalf Of Thomas Pengo
Sent: Friday, April 22, 2022 9:45 PM
To: [log in to unmask]
Subject: Re: 3D Image Analysis Tools and Reproducibility Event - April 27th via Zoom
*****
To join or leave the confocal microscopy listserv or to change your email address, go to:
https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi Brian
Thanks for your comments so far and your work on the study’s images.
I have been helping with the generation of the study’s images and wanted to add some background. We chose not to reinvent the wheel and use an existing open source software, Cytopacq, which has also been used before in other studies. Rather than use the online generator, we preferred to use the command line version that would allow us to record a configuration file and use a custom PSF file. We generated a PSF using another open source software, PSF Generator. After you raised your concerns, we started to look into it but didn’t get very far at the time. My availability was limited and I have only recently been able to get back to the issue. At the time, we deemed it was not invalidating the study as the point of it was, unlike a traditional challenge, to investigate how different researchers would get different results given the same task. I agree reproducibility in the narrow sense implies that were looking for a perfect ability to replay a given approach step by step, but I do believe this variability study does fall under the same broader problem of reporting results completely and correctly, applicability of a given method to a different domain, generalizability, …
To your point, we’re trying another approach that hopefully will serve your purpose. We are trying to generate an “experimental PSF”, by using a microsphere phantom. One one hand it would allow you and others to perform hopefully a more successful deconvolution, and on the other to study the mismatch between expected and actual results.
I hope to have some more to report early next week.
Thomas
On Fri, Apr 22, 2022 at 2:56 PM Brian Northan <[log in to unmask]> wrote:
> *****
> To join or leave the confocal microscopy listserv or to change your
> email address, go to:
> https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Hi Kris
>
> It seems from what you are saying, that you knew the PSF spacings were
> wrong in November but wouldn't confirm it, as not to disrupt the
> study? I spent many hours in November trouble shooting the ImageJ and
> Python PSF generators, to see if the issue was with them. It would
> have been nice to know then that I wasn't going crazy and there
> weren't bugs in the open source libraries I was using. As a long time
> contributor to imagej it would have been nice to have reliable test images to verify code with.
>
> Anyway, why would you have needed to stop the study? All you would
> have had to do was announce there was a mistake in the PSF and you
> were working to fix it. In fact it would be interesting to see if you
> got better or different answers after fixing the PSF.
>
> The branding of the study as a reproducibility study is what I am
> really having an issue with. Reproducibility is a very powerful and
> personal word to many people, especially people who have experienced
> real reproducibility controversies and all the horrible politics that go with it.
> Reproducibility is when you try to verify a claim such as 'this method
> has the best segmentation accuracy", and step by step try to reproduce
> someone else's work. What you are doing seems to be a variability
> study. It is certainly valuable but should not be branded as a reproducibility study.
>
> Brian
>
> On Fri, Apr 22, 2022 at 2:46 PM Kubow, Kristopher E - kubowke <
> [log in to unmask]> wrote:
>
> > *****
> > To join or leave the confocal microscopy listserv or to change your
> > email address, go to:
> > https://lists.umn.edu/cgi-bin/wa?SUBED1=confocalmicroscopy&A=1
> > Post images on http://www.imgur.com and include the link in your
> posting.
> > *****
> >
> > Hi Brian,
> >
> > I agree with the core issue: that there is some kind of mismatch
> > between the acquisition parameters reported for the nuclei sets and
> > the PSF apparent in the images and that this may interfere with
> > deconvolving the image. And I agree that this is an important issue
> > to resolve. However, I respectfully disagree on the dire
> > implications of this for our study and event.
> >
> > Our study was never intended nor designed to find or rank the best
> > segmentation procedures. Our goal was simply to assess variability
> > in the segmentation results produced by a variety of different
> > people (with different backgrounds/expertise/etc.) using a wide variety of approaches.
> > In this way, our study is very different from the segmentation
> > challenges that are conducted to crowd-source new and better
> > algorithms. Although we do hope to make some general conclusions
> > about what procedures lead to
> more
> > or less variability, these conclusions will be very coarse-grained
> because
> > they will depend on a sufficient number of people having done the
> > same operation (which we can't control). For example, it's likely
> > that many people used some kind of threshold in their procedure, so
> > we might be
> able
> > to make some broad comments on that. Indeed, it would be better to
> > have a highly controlled experiment (changing just one variable at a
> > time) to
> test
> > the effects of specific techniques - in particular because
> > segmentation procedures involve multiple components. Even if we were
> > to identify everyone who performed deconvolution as an initial step,
> > they would all likely have different subsequent steps - all of which
> > might introduce significant variability. Our primary goal is to
> > assess the current state
> of
> > variability (what's currently going on out there in the real world?)
> > similar to, e.g., the recent study of microscopy methods reporting
> > by Marques et al. (https://elifesciences.org/articles/55133).
> >
> > When you brought the PSF issue to our attention several months ago,
> > we
> had
> > to make a decision. At that point, many people had already completed
> > our study so, in essence, the "damage was done." We could stop the
> > study,
> omit
> > all the previously collected data, identify and fix the problem, and
> > then start the study from scratch. Or we could continue the study
> > and accept
> the
> > fact that it would be limited in that we would not have submissions
> > involving deconvolution. Because, to our knowledge, the PSF problem
> > only affects one potential image operation (deconvolution) typically
> > involved
> in
> > segmentation, and because our study was focused on measuring overall
> > variability and not the effects of specific techniques, we decided
> > to continue the study. Now, you might argue whether one should
> > segment a widefield image without first deconvolving it, but that's
> > beside the
> point,
> > since we are examining not what people ought to do, but what they
> actually
> > do. So our study is limited in that people who would have performed
> > deconvolution before segmenting nuclei may not be represented in our
> > dataset. We fully admit this limitation, but we feel that it will
> > not severely impact our broad assessment of person-to-person
> > variability in image segmentation.
> >
> > Likewise, our event next week is not meant to compare the
> > performance of different programs. It is purely educational: an
> > opportunity for
> different
> > image analysis platforms to demonstrate potential workflows for
> performing
> > segmentation and to show the ways that the platform can promote
> > image analysis reproducibility (e.g. in standardizing and
> > documenting
> workflows).
> > Again, we acknowledge that there is a PSF/deconvolution limitation,
> > but since that is only one potential element of a segmentation
> > workflow, we feel that the event will still be fruitful even if it
> > deconvolution is
> not
> > possible.
> >
> > Again, we are working with our colleagues at Cytopacq to understand
> > the problem and will post updates to image.sc. If there have been
> > any miscommunications or misunderstandings regarding the study or
> > event's goals, I take full responsibility for that and hope that
> > I've clarified everything now.
> >
> > Best,
> > -Kris
> >
> > ________________________________
> > From: Brian Northan <[log in to unmask]>
> > Sent: Wednesday, April 20, 2022 9:28 AM
> > Subject: Re: 3D Image Analysis Tools and Reproducibility Event -
> > April 27th via Zoom
> >
> > *****
> > To join or leave the confocal microscopy listserv or to change your
> > email address, go to:
> >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=WQfBJ5Y4VW
> f7Nz0nN-nVQLdxVCsinhxexQMO9MQxddA&s=4zdyqyfF4_ngpNFPejYyzhOkqNR1yWohRJ
> hmSXPANB4&e=
> > Post images on
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=WQfBJ5Y4VWf7Nz0nN-nVQLdxVCsinhxexQMO9MQxddA&s=zjccIXaaAFxbaFyr
> DAoVgYOIv77ORrUxympimvjBH1c&e=
> > and include the link in your posting.
> > *****
> >
> > Hi Kris
> >
> > ABRF is marketing this workshop and study as 'trying to promote
> > reproducibility' however in the meantime there is a reproducibility issue
> > in the study itself that has dragged on for several months. This issue
> > influences algorithm choice. Respectfully, isn't it appropriate to
> > push hard to have ABRF solve this issue before doing a
> > reproducibility
> workshop
> > next week? If incorrect information, as to appropriate algorithm
> > choice
> is
> > presented in the workshop or study it will take image processing
> > professionals their time to correct the mistaken impression.
> >
> > I understand people at ABRF are busy and I am absolutely not
> > implying
> that
> > ABRF is trying to hide anything, or has intentionally done anything
> > wrong at all. However, to be honest, I do not think ABRF (and many
> > other microscopy groups) fully appreciate the types of technical
> > skills it
> takes
> > to do proper justice to complex digital image processing problems.
> > Oftentimes image processing workshops put on by microscopy groups
> > are
> done
> > in too simple a manner and have led to a lack of rigor in image
> > processing.
> >
> > Brian
> >
> >
> >
> >
> > On Tue, Apr 19, 2022 at 1:03 PM Kubow, Kristopher E - kubowke <
> > [log in to unmask]> wrote:
> >
> > > *****
> > > To join or leave the confocal microscopy listserv or to change
> > > your
> email
> > > address, go to:
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=WQfBJ5Y4VW
> f7Nz0nN-nVQLdxVCsinhxexQMO9MQxddA&s=4zdyqyfF4_ngpNFPejYyzhOkqNR1yWohRJ
> hmSXPANB4&e=
> > > Post images on
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=WQfBJ5Y4VWf7Nz0nN-nVQLdxVCsinhxexQMO9MQxddA&s=zjccIXaaAFxbaFyr
> DAoVgYOIv77ORrUxympimvjBH1c&e=
> > and include the link in your posting.
> > > *****
> > >
> > > Thanks for your input, Brian - it's much appreciated. And since a
> > > continued discussion of this sort is more appropriate for
> > > image.sc,
> > let's
> > > move the conversation there. It is better to have the full
> > > threaded discussion in one place rather than cross-posted to the listserv.
> > >
> > > Also, I want to emphasize that the study is being conducted by a
> > > group
> of
> > > volunteers in our "spare time." All of us run core facilities -
> > > many of
> > us
> > > completely by ourselves (no staff). Many of us also teach and/or
> > > have
> > other
> > > obligations at our institutions. None of us are paid/compensated
> (either
> > by
> > > our institutions or the ABRF) to run the study. We are in no way
> > > trying
> > to
> > > hide anything. We are trying to do our best within our
> > > constraints. I understand you wish things would move more quickly
> > > (so do we!), but
> even
> > > tasks that might seem to be straightforward/quick/obvious/etc. do
> > > take
> > time
> > > and require coordination among our members. I don't say this to be
> > > confrontational or condescending, but only to emphasize that our
> members
> > > are well-meaning, hard-working individuals who are trying to do
> > > their
> > best
> > > with what they have.
> > >
> > > I look forward to engaging with you further on the deconvolution
> > > issue
> at
> > > image.sc.
> > >
> > > Best,
> > > -Kris
> > >
> > > ________________________________
> > > From: Brian Northan <[log in to unmask]>
> > > Sent: Monday, April 18, 2022 8:12 AM
> > > Subject: Re: 3D Image Analysis Tools and Reproducibility Event -
> > > April 27th via Zoom
> > >
> > > *****
> > > To join or leave the confocal microscopy listserv or to change
> > > your
> email
> > > address, go to:
> > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=lMiutr1dy6
> jxfbZfb00NtOJHkGisOXchgySLZr_ltdM&s=qG6OzlR17xLhXVpo66_yZ3GUTqte32xrV4
> awF3aiY3k&e=
> > > Post images on
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=lMiutr1dy6jxfbZfb00NtOJHkGisOXchgySLZr_ltdM&s=tCrWXALdZYlJ-BHr
> rWfqDrabDddXGFqbajpsBYTjJN0&e=
> > > and include the link in your posting.
> > > *****
> > >
> > > Thanks Kris
> > >
> > > Just to re-emphasize why this issue is so important. To get
> > > accurate intensities and segmentation of the nuclei images some
> > > type of image restoration must be done (either deconvolution or AI
> > > based
> restoration).
> > > The segmentations I've seen of the nuclei image, so far, looked
> > > good
> at
> > > first glance (round well separated objects), however are merely
> > > simplifications which fit a spheroid to the data, but do not
> > > capture
> the
> > > true shape or intensities.
> > >
> > > Anyway, I will spend some time this week or next looking over the
> docker
> > > release and see if I can tell what happened with the PSF. Seems
> > > like
> it
> > is
> > > probably a resampling issue, perhaps even as simple as incorrect
> reading
> > of
> > > spacings from the meta-data, but hard to know for sure.
> > >
> > > Brian
> > >
> > > Brian
> > >
> > > On Sun, Apr 17, 2022 at 8:08 PM Kubow, Kristopher E - kubowke <
> > > [log in to unmask]> wrote:
> > >
> > > > *****
> > > > To join or leave the confocal microscopy listserv or to change
> > > > your
> > email
> > > > address, go to:
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=lMiutr1dy6
> jxfbZfb00NtOJHkGisOXchgySLZr_ltdM&s=qG6OzlR17xLhXVpo66_yZ3GUTqte32xrV4
> awF3aiY3k&e=
> > > > Post images on
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=lMiutr1dy6jxfbZfb00NtOJHkGisOXchgySLZr_ltdM&s=tCrWXALdZYlJ-BHr
> rWfqDrabDddXGFqbajpsBYTjJN0&e=
> > > and include the link in your posting.
> > > > *****
> > > >
> > > > Question was cross-posted on image.sc forum please see forum for
> > > response
> > > > and continuation of discussion:
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__forum.image.sc_t_
> 3d-2Dimage-2Danalysis-2Dtools-2Dand-2Dreproducibility-2Devent-2Dapril-
> 2D27th-2Dvia-2Dzoom_64671_5-3Fu-3Dkris-5Fkubow&d=DwIFaQ&c=eLbWYnpnzycB
> Cgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=lMiutr1dy6j
> xfbZfb00NtOJHkGisOXchgySLZr_ltdM&s=-Y-XonaNJ5KjpvnIEH1yhfrJANHJuRnQqU1
> 5_lCOqxs&e=
> > > >
> > > > ________________________________
> > > > From: Brian Northan <[log in to unmask]>
> > > > Sent: Friday, April 15, 2022 6:09 AM
> > > > Subject: Re: 3D Image Analysis Tools and Reproducibility Event -
> April
> > > > 27th via Zoom
> > > >
> > > > *****
> > > > To join or leave the confocal microscopy listserv or to change
> > > > your
> > email
> > > > address, go to:
> > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=hBsapqAzr5
> hAx-G9-qM3LSc9TEwTGgBoh0n4rbwkBks&s=z0i7P1WXTT-6ZCuw6KqHreKuHf-jiYoJTF
> fEJo1rass&e=
> > > > Post images on
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=hBsapqAzr5hAx-G9-qM3LSc9TEwTGgBoh0n4rbwkBks&s=wPbLmuOcWhpfciV0
> CceO3Wq0xE9rQgqBl5U33Gian8E&e=
> > > > and include the link in your posting.
> > > > *****
> > > >
> > > > Hi Kris
> > > >
> > > > Any update on this? Why not just release the forward model code
> > > publically
> > > > and have the community look it over? I don't understand how
> > > > ABRF can
> > > claim
> > > > they are addressing 'reproducibility' when they are working in
> > > > such a closed manner. Isn't openness and communication the
> > > > number one
> > principle
> > > > of reproducibility?
> > > >
> > > > Brian
> > > >
> > > > On Thu, Mar 24, 2022 at 3:28 PM Kubow, Kristopher E - kubowke <
> > > > [log in to unmask]> wrote:
> > > >
> > > > > *****
> > > > > To join or leave the confocal microscopy listserv or to change
> > > > > your
> > > email
> > > > > address, go to:
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=hBsapqAzr5
> hAx-G9-qM3LSc9TEwTGgBoh0n4rbwkBks&s=z0i7P1WXTT-6ZCuw6KqHreKuHf-jiYoJTF
> fEJo1rass&e=
> > > > > Post images on
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=hBsapqAzr5hAx-G9-qM3LSc9TEwTGgBoh0n4rbwkBks&s=wPbLmuOcWhpfciV0
> CceO3Wq0xE9rQgqBl5U33Gian8E&e=
> > > > and include the link in your posting.
> > > > > *****
> > > > >
> > > > > Thanks Brian and David. We agree that there does seem to be a
> > mismatch
> > > > > between the spacing parameters that were given and what is
> > > > > actually
> > > used
> > > > in
> > > > > the nuclei images. This will impact deconvolution of the
> > > > > images, if
> > > > people
> > > > > elect to do that as part of their segmentation protocol for
> > > > > the
> > study.
> > > > > We’re currently looking into the problem and are in contact
> > > > > with
> the
> > > > group
> > > > > that developed the simulation software. We'll post updates
> > > > > (when we
> > get
> > > > > them) on the image.sc forum:
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__forum.image.sc_t_
> 3d-2Dimage-2Danalysis-2Dtools-2Dand-2Dreproducibility-2Devent-2Dapril-
> 2D27th-2Dvia-2Dzoom_64671&d=DwIFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn
> _5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=hBsapqAzr5hAx-G9-qM3LSc9TEwTGgBo
> h0n4rbwkBks&s=SMse04bgkg546B_NFjehOds-hRMIW4bovk4mbCxMeqc&e=
> > > > >
> > > > > ________________________________
> > > > > From: Brian Northan <[log in to unmask]>
> > > > > Sent: Wednesday, March 23, 2022 9:26 AM
> > > > > Subject: Re: 3D Image Analysis Tools and Reproducibility Event
> > > > > -
> > April
> > > > > 27th via Zoom
> > > > >
> > > > > *****
> > > > > To join or leave the confocal microscopy listserv or to change
> > > > > your
> > > email
> > > > > address, go to:
> > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8EpRKzd
> BsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=-owGyRKzOtmTtqOXx9hHPa81IsG5cCt5iC
> VjWV4ILTg&e=
> > > > > Post images on
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=Bbo8EpRKzdBsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=uQ0i7zYMybxCqH94
> JEtgiMLq5Tqf4RT45jB8kGVrshU&e=
> > > > > and include the link in your posting.
> > > > > *****
> > > > >
> > > > > Interestingly I used similar spacings to those David estimated.
> > After
> > > > > first using the given spacings and getting terrible
> > > > > deconvolution
> > > > results,
> > > > > I simply used the Nyquist spacings and got a much better result.
> > > > > From there I tweaked the spacings a bit and used 0.1 XY and
> > > > > 0.8 Z
> and
> > > > got a
> > > > > pretty good deconvolution.
> > > > >
> > > > > I mentioned this to Kris Kublow, one of the organizers several
> months
> > > ago
> > > > > and he said the group would look into it, but I never heard if
> > > > > they
> > > did.
> > > > >
> > > > > Jessica, did your group ever run tests to validate the PSF,
> > > > > forward
> > > model
> > > > > and spacings? If so, how did you convince yourself the
> > > > > simulation
> > was
> > > > > correct?
> > > > >
> > > > > Brian
> > > > >
> > > > >
> > > > > On Mon, Mar 21, 2022 at 1:09 PM David Biggs
> > > > > <[log in to unmask]>
> > > wrote:
> > > > >
> > > > > > *****
> > > > > > To join or leave the confocal microscopy listserv or to
> > > > > > change
> your
> > > > email
> > > > > > address, go to:
> > > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8EpRKzd
> BsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=-owGyRKzOtmTtqOXx9hHPa81IsG5cCt5iC
> VjWV4ILTg&e=
> > > > > > Post images on
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=Bbo8EpRKzdBsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=uQ0i7zYMybxCqH94
> JEtgiMLq5Tqf4RT45jB8kGVrshU&e=
> > > > > and include the link in your posting.
> > > > > > *****
> > > > > >
> > > > > > In reference to Brian's observations, I also concur that the
> > > parameters
> > > > > > provided for the widefield nuclei dataset do not match the
> > > > > > data
> > > > provided.
> > > > > > For a 0.75 NA air lens, I estimate the Z spacing is closer
> > > > > > to
> > 0.75um,
> > > > and
> > > > > > the XY spacing is closer to 0.1um.
> > > > > > I roughly estimated these from the shape of the missing cone
> > > > > > in
> the
> > > > > Fourier
> > > > > > domain.
> > > > > >
> > > > > > Any clarification would be appreciated.
> > > > > >
> > > > > > Thanks,
> > > > > > David
> > > > > >
> > > > > >
> > > > > >
> > > > > > On Sat, Mar 19, 2022 at 1:44 PM Brian Northan <
> [log in to unmask]>
> > > > > wrote:
> > > > > >
> > > > > > > *****
> > > > > > > To join or leave the confocal microscopy listserv or to
> > > > > > > change
> > your
> > > > > email
> > > > > > > address, go to:
> > > > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8EpRKzd
> BsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=-owGyRKzOtmTtqOXx9hHPa81IsG5cCt5iC
> VjWV4ILTg&e=
> > > > > > > Post images on
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=Bbo8EpRKzdBsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=uQ0i7zYMybxCqH94
> JEtgiMLq5Tqf4RT45jB8kGVrshU&e=
> > > > > and include the link in your
> > > > > > posting.
> > > > > > > *****
> > > > > > >
> > > > > > > Has anyone else tried to deconvolve the nuclei images used
> > > > > > > in
> the
> > > > > > study? I
> > > > > > > was only able to obtain a good deconvolution result after
> > > > introducing a
> > > > > > > fudge factor into the PSF generation. Is anybody else
> > > > > > > running
> > into
> > > > the
> > > > > > > same problem? My concern is that there is a potential
> > > > > > > sampling
> > > issue
> > > > > in
> > > > > > > the PSF used in the simulation, and that in turn could
> (falsely)
> > > make
> > > > > it
> > > > > > > seem as if deconvolution doesn't work. That would be a
> > > > > > > shame, deconvolution is an important tool for processing
> > > > > > > 3D Widefield
> > > > > > fluorescence
> > > > > > > images, and it would be nice if this study could be used,
> > > > > > > to
> > > > > demonstrate
> > > > > > > the benefits of deconvolution.
> > > > > > >
> > > > > > > Contact me off list if you would like. I'd like to verify
> > > > > > > that
> > > these
> > > > > > > images can be deconvolved using standard tools.
> > > > > > >
> > > > > > > Brian
> > > > > > >
> > > > > > > On Fri, Mar 18, 2022 at 9:44 AM Jessica Hornick <
> > > > > > > [log in to unmask]>
> > > > > > > wrote:
> > > > > > >
> > > > > > > > *****
> > > > > > > > To join or leave the confocal microscopy listserv or to
> change
> > > your
> > > > > > email
> > > > > > > > address, go to:
> > > > > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.umn.edu_cgi
> -2Dbin_wa-3FSUBED1-3Dconfocalmicroscopy-26A-3D1&d=DwIFaQ&c=eLbWYnpnzyc
> BCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8EpRKzd
> BsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=-owGyRKzOtmTtqOXx9hHPa81IsG5cCt5iC
> VjWV4ILTg&e=
> > > > > > > > Post images on
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.imgur.com&d=Dw
> IFaQ&c=eLbWYnpnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl
> 0IvoA&m=Bbo8EpRKzdBsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=uQ0i7zYMybxCqH94
> JEtgiMLq5Tqf4RT45jB8kGVrshU&e=
> > > > > and include the link in your
> > > > > > > posting.
> > > > > > > > *****
> > > > > > > >
> > > > > > > > Come learn more about 3D image analysis tools and how
> > > > > > > > they
> can
> > > help
> > > > > > > > improve reproducibility at this two-hour online workshop
> hosted
> > > by
> > > > > the
> > > > > > > > Light Microscopy Research Group of the ABRF!
> > > > > > > >
> > > > > > > > This event will be centered around the LMRG’s ongoing
> > > > > > > > study
> of
> > > > > > > > reproducibility in 3D image analysis in which we seek to
> > > > characterize
> > > > > > and
> > > > > > > > identify sources of (ir)reproducibility in 3D segmentation.
> Try
> > > > your
> > > > > > hand
> > > > > > > > at analyzing the study images and then attend the
> > > > > > > > workshop to
> > see
> > > > how
> > > > > > the
> > > > > > > > pros would do it. Representatives from several major 3D
> > analysis
> > > > > > > platforms
> > > > > > > > will walk attendees through how to segment and analyze
> > > > > > > > the
> > images
> > > > > from
> > > > > > > our
> > > > > > > > study and will discuss how their platforms support
> reproducible
> > > > image
> > > > > > > > analysis. These demonstrations will occur simultaneously
> > > > > > > > in
> > > > breakout
> > > > > > > rooms
> > > > > > > > and attendees may choose up to two to attend. There will
> > > > > > > > also
> > be
> > > an
> > > > > > open
> > > > > > > > session at the conclusion of the program during which
> attendees
> > > may
> > > > > > visit
> > > > > > > > and talk with any representative they like. Note that
> > > > > > > > the
> > > platforms
> > > > > > will
> > > > > > > > not be competing against each other–the purposes of the
> > > > > > > > event
> > are
> > > > > > > > instructional/informational and to promote
> > > > > > > > reproducibility in
> > > image
> > > > > > > > analysis.
> > > > > > > >
> > > > > > > > We will host representatives from the following platforms:
> > > > > ImageJ/FIJI,
> > > > > > > > Napari, CellProfiler, Arivis, and Volocity. We hope to
> > > > > > > > add
> more
> > > to
> > > > > this
> > > > > > > > list before the event.
> > > > > > > >
> > > > > > > > The event will take place on April 27th at 8:00 am
> > > > > > > > Pacific
> > Time /
> > > > > 11:00
> > > > > > > am
> > > > > > > > Eastern Time via Zoom.
> > > > > > > >
> > > > > > > > ** Register for the event here:
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__us02web.zoom.us_m
> eeting_register_tZEsceuvpzIvH9e0kFHEuWdxRMBAgrrpUwzj&d=DwIFaQ&c=eLbWYn
> pnzycBCgmb7vCI4uqNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8E
> pRKzdBsReltSfbg8Udr6_-vbdBTWzo2_VmZUow&s=df5s8P1S7R061gnHC6OXrqAOlhVJ3
> FaJP5Wt8KsC0Qo&e=
> > > > > > > >
> > > > > > > > This event is targeted to people who are interested in
> learning
> > > > more
> > > > > > > about
> > > > > > > > a specific image analysis platform, learning how to use
> > > > > > > > an
> > image
> > > > > > analysis
> > > > > > > > tool better, and/or learning about how reproducibility
> > > > > > > > is
> > > addressed
> > > > > by
> > > > > > > > these platforms.
> > > > > > > >
> > > > > > > > We strongly encourage attendees to participate in the
> > > > > > > > study
> > > before
> > > > > the
> > > > > > > > event so that you are familiar with the image analysis
> problem.
> > > > > > Attendees
> > > > > > > > who participate in the study before the event will be
> > > > > > > > entered
> > in
> > > a
> > > > > $50
> > > > > > > > Amazon gift card drawing. (If you’ve already
> > > > > > > > participated in
> > the
> > > > > study
> > > > > > > and
> > > > > > > > didn’t win in our last drawing, you’re automatically
> entered.)
> > > > Study
> > > > > > info
> > > > > > > > can be found here:
> > > > > > >
> > > > >
> > > >
> > >
> >
> https://urldefense.proofpoint.com/v2/url?u=https-3A__sites.google.com_
> view_lmrg-2Dimage-2Danalysis-2Dstudy&d=DwIFaQ&c=eLbWYnpnzycBCgmb7vCI4u
> qNEB9RSjOdn_5nBEmmeq0&r=wgGHXpNkPeE6PJJsl0IvoA&m=Bbo8EpRKzdBsReltSfbg8
> Udr6_-vbdBTWzo2_VmZUow&s=AvCI1kJRHZzgNZUP-7MfZJdOZZaCKUJOvPZAamulY-0&e
> =
> > > > > > > >
> > > > > > > > If you have questions about the event, please contact
> > > > > > > > Jessica
> > > > > Hornick (
> > > > > > > > [log in to unmask]).
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>
--
________________________________________
Dr. Thomas Pengo
Director of Applications and Services, UMII University of Minnesota, Twin Cities Campus Cancer and Cardiovascular Research Building
2231 6th St SE
Minneapolis, MN 55455
|
