Software-independent discussion of bio-image analysis solutions

community
forum
Tags: #<Tag:0x00007fb8820d39b8> #<Tag:0x00007fb8820d30d0>

#1

Dear BioImage Analysts,

TL;DR:

I’ve opened a proposal for a dedicated BioImage Analysis forum on StackExchange. This is an experiment to see if there’s a sufficiently large community interested to discuss image analysis solutions across ecosystems (i.e spanning ImageJ/ImgLib2, KNIME, Python, Matlab, CellProfiler and more).

If you like the idea, please support it by following the project proposal and suggesting example questions (one sentence suffices) that define the scope of the new forum.


Background

Following discussions within the NEUBIAS community and on the latest NEUBIAS2020 conference in Lisbon, it was proposed to create a new forum that allows discussion of image analysis solutions independent of the software used, in combination with the @Call4Help initiative announced on another thread.

As the ImageJ forum and its community aim to be open for discussion of solutions independent of the ecosystem, and explicitly welcome discussing and comparing image analysis solutions involving KNIME, CellProfiler, Icy or even projects totally unrelated to ImageJ (such as Python or Matlab), the initial trial was to direct the discussion of problems raised in the call4help sessions to the ImageJ forum.

Despite the efforts to communicate the openness of the ImageJ forum discussions, concerns were raised in the NEUBIAS community, stating that the “branding” of the ImageJ forum and its implicit focus on ImageJ and the SciJava ecosystem would discourage people to suggest solutions based on non-ImageJ software.

In an attempt to create a software-independent discussion platform, I created a proposal for a general BioImage Analysis forum on StackExchange. To make it as easy as possible for the existing communities of the ImageJ, CellProfiler, Icy and KNIME forums to sustain yet another image analysis-related forum, the idea is to cross-link from and to the new forum using the OpenGraph protocol that allows embedding (aka one-boxing) the content across forum instances (both StackExchange and Discourse - the forum backend used by the ImageJ and CellProfiler forums - support this protocol).

How to support the idea

If you agree that an ecosystem-independent forum for general discussion of (bio)image analysis solutions would be useful, please:

  • register on stackexchange.com (you can use your account from stackoverflow.com if you have one)
  • follow the proposal,
  • suggest a few example question that illustrate the type of discussions you’d see fit in this new forum, and
  • vote on the question that have been suggested already.

As stated above, the whole thing is no more than an experiment yet, to see if the community interest is large enough. But if we get enough people to support the project, we might soon have another great resource that connects the various image analysis communities of ImageJ, KNIME, CellProfiler, Icy, Matlab, Python, etc. :slight_smile:


#2

Good idea, Jan - maybe you can also use the Neubias TS2 and TS3 mailing lists to notify people, since the ones who don’t use ImageJ don’t follow what’s going on in this forum. I will notify Ofrah.


#3

I think there are already too many StackExchange sites, and that some of them are already very appropriate for bioimage analysis questions:

  • Cross Validated - statistics, machine learning, data analysis, data mining, and data visualization
  • Signal Processing - the art and science of signal, image and video processing
  • Stack Overflow - targeted technical questions about coding, implementation, etc.

See also Stack Exchange site to ask questions about computer vision? which poses essentially the same question of “where do I ask about image processing on SO?”

If people strongly feel that a new StackExchange site is warranted, I would urge you not to limit it to bioimage analysis, but instead scientific image analysis in general. There are many scientists who use ImageJ who are not in bioimage informatics. It would be best if this new image processing forum included them too, since there is a huge amount of technical overlap in the computational techniques, algorithms, etc.


#4

Of course, the list of all 170 current topic sites on stackexchange is quite overwhelming, and seeing the number of project proposals on area51 that are being closed before reaching their beta phase, I am not too optimistic about the current proposal reaching a healthy state.

However, there are a few types of questions that will be off-topic on the sites you listed, because the people asking the questions…

  • … are struggling to even extract out of their images some “numerical” data suitable for a statistical analysis that would justify asking questions on Cross Validated,
  • … do not (or have the feeling not to) have enough background on maths and signal processing to even know the vocabulary required to accurately phrase a question on Signal Processing, and
  • … did not produce any code that would justify asking on Stack Overflow, because again they’re struggling at a much earlier stage, at even getting something useful out of their image data, not even talking about automation or applying some specifc algorithm.

At least for Stack Overflow, I’ve seen a few questions with the imagej tag that were getting closed and removed after a short period.

For beginners in image analysis, this can be a quite frustrating experience, and I have to say that I’m glad that the ImageJ forum is one of the places where questions like this can be discussed. But as I mentioned above, there were concerns in the community that solutions involving e.g. Icy, Matlab, Imaris, Arivis or similar tools could not be discussed freely here on this forum, as it is primarily targetting ImageJ.


So let’s take this new forum proposal as what it is, merely an experiment if the community can sustain yet another forum… ; my hope is that we can at least define a type of questions that has no appropriate place yet to be asked; and if there are no such questions: even better, then we don’t need to create a new forum :slight_smile:


That’s a good point, we might have to broaden the focus a bit in order to reach a bigger community. As the initiative came from the bio-image analysis community specifically, this was my starting point when creating the proposal.

While I agree that there is quite some technical overlap with other scientific fields, I think the range of specifically biological/biomedical use cases (from “How do I analyze my gel?” to “How can I automatically classify the cells in my huge mutlichannel anisotropic light-sheet dataset?”) is the one that is most lacking an appropriate place for discussion right now.

Happy to discuss this further…


#5

Indeed.

This actually highlights another major concern I have about using StackExchange as a focal point for these sorts of discussions: StackExchange is a Q&A site, not a discussion site. When extended discussions start happening in comments, at minimum, they will get shunted to the StackExchange chat; or worse, the moderation-happy folks in the community will chastise people and delete threads.

Maybe it is fine—I think the vast majority of threads here do end up being Q&A threads. However, there will be some kinds of threads which are not appropriate for StackExchange. This is in fact why Discourse was created: to foster civilized discussion rather than ruthlessly minimizing it.

What if we just renamed this forum to something else, without “ImageJ” in the title? What if we just changed the domain name? What name would convey our focus as well as our generality? Not “SciJava” since it’s not limited to Java. Not “bio” anything, since it’s not limited to bio. Maybe not even “image” since we also care about discussing non-image-specific numerical data science—e.g., machine learning, KNIME, and so on.

Thoughts? Ideas?


#6

One option would be to rename the forum. Another would be to keep the name, but state that ‘imagej’ has evolved and is evolving to a polyglot imaging and data analysis solution, and we support all the tools that imagej interacts with, including KNIME, ICY, MATLAB, Python, CellProfiler, etc.