Image analysis platform - How services are provided?

Hello everyone !

We’re starting to think about having an image analysis service here at the IBENS bioinformatics platform (that currently provides genomic data analysis and scientific illustration services).

I have a core question about it, that I think you all could be of great help:

Do you (or your platform) provide an analysis service or an software/workflow that can analyse the data ?

For example: There’s a dataset of images that need cell counting. As far as I can see, there are two approaches:

  1. Develop my workflow and process the data, and give to the client only the results (like a table of counted cells per image).
  2. Create a “friendly” macro/interface that the client can use, and process the data himself (like a Fiji macro or a python notebook that they can load the images and get the cell count by the end).

Option 1 seems easier/faster, as we don’t need to worry about how the client (usually, biologists) could run the algorithm. But I foresee that they are not likely to have all the data ready during the beginning of the project, and to acquire huge amounts of data first, and then later see if it can be processed looks like a recipe for problems. The problem with option 2 is that it would possibly create and endless series of extra requests, like “We changed the resolution of the acquisition, and now the macro doesn’t work properly” or “We’re using a different marker for the cells, and now it doesn’t work” etc etc…

So, what’s the approach of your platform when providing image analysis services ?

Thanks in advance !
– Felipe Delestro

3 Likes

Hi @Delestro

It turns out I am part of a small facility dedicated to image analysis - and not too far from you.

What about we share lunch over your question? -> Pasteur canteen.

For your specific question:
I go to solution 2 if I am sure that

  • there will be little feedback required on my side (question is simple and tool efficient, so the user is unlikely to come back to me discovering new challenges associate to their question);
  • they do not need scientific input from my side (algorithm is simple and well accepted, impact on imaging modality on quantitation is small).

In reality we work with 1) and 2), depending on the project and the user.

4 Likes

Hello, thanks for your reply @tinevez !

Yes, it would be really interesting to have lunch together and better discuss about the possibilities ! Auguste Genovesio, head of the platform here, would also like to join.

When does it works best for you ?

Hello @Delestro

Great!
What about next week?

Let’s discuss this when I am back from holidays because email and calendar by phone = nightmare

Okay, great !

You can reach me on delestro@biologie.ens.fr =)