filter idea for research

[LeroyAlex]

Hi,

It actually comes from a demand of one our users. He wanted to be able to get images from any project he owns when he does a research. Till there, no problem that's what omero does by default if you click on images in the filter choice(ex: images with gfp tags). Then, he asked me if he could select some project he specifically want to search in. I tried every way i could (maybe i didn't think of the good one) but i could not ask the engine to search: let's say all the images gfp tagged in only the 2 project: cancer and histology. He wanted to compare at sight (from thumbnails) the images in both folder. He actually told me there was such a function in a database (evernote) where you choose tags then apply the research to specific "carnets" corresponding to our dataset/project. Did you get any demand of the type from other users? Would it be easy/useful to add it to omero?

Best,

Alex.

[dpwrussell]

Hi Alex,

I understand what he is desiring to achieve, this is a use-case that I have considered for webtagging search. Unfortunately, I have not got to the point of implementing that functionality yet when a project (or even more complicated to implement, projects) can be added to the filter criteria. The main reason for this is that I want to integrate webtagging search into the normal data tree as a filter so that as search terms are added, the display of data reduces. I have been laying the groundwork for this and hopefully we are nearly there. In the new year I hope to make some progress on the search again and I will do what I can to implement your user's requirement before I leave the project in March.

Webtagging link

Cheers,

Douglas

[LeroyAlex]

Hi Douglas,

Thanks for your answer.
We actually already added the webtagging plugin on our omero.web and i and the users find it pretty useful.
Can't wait to try the future features of the plugin.

Best,

Alex.

[dpwrussell]

One potential solution for now. It is by no means optimal and requires extra work up-front, but there's a chance it might work for your user if they really need the search capability.

Say I have a data tree like so:

p1->d1->i1
p1->d2->i2
p2->d3->i3

Each project might have multiple additional datasets and each dataset might have multiple additional images.

Lets say that you were to tag each image with a tag representing the the project and dataset it was in. Yes, this is a duplication of effort/data, but it allows the search to deal exclusively with tags.

Tag representation for images (with tag1 and tag2 being normal tags you've added):
i1: d1, p1, tag1, tag2
i2: d2, p1, tag1, tag2
i3: d3, p2, tag1, tag2

Say the user is then running a webtagging search for some tags.

e.g. Search: 'tag1, tag2'

Tag search will show tag intersections possible including: d1, p1, d2, p2, d3

The user could then restrict the query by entering one of these terms.

e.g. Search 'tag1, tag2, p1'

The results would be i1 and i2.

Tag search is incapable of 'OR'. There are long complicated reasons for this which I will not go into, but that is an obvious limitation if the user wishes to search across several projects. However, it is possible that instead of tagging by project and/or dataset, the user could add meaningful tags to define collections of data. e.g. if he/she is likely to search across p1 and p2, there may be a good reason for that which could be encoded by tagging all images in p1 and p2 with the same tag 'my-common-criteria'.

Another obvious limitation of using tags to represent project/dataset (which might not happen if using a tag describing the data like the 'my-common-criteria' tag), if you move/copy the data, the tags will then be wrong.

[wmoore]

Of course, in the scenario described above, you could also start by filtering on "p1" tag, then use the other tags to filter within that "Project". More like the original workflow that was requested.


Just reading the question again, it occurred to me that the "Thumbnail Figure" could provide a partial solution. In the web client, you can select multiple Datasets (not Projects I'm afraid) and then click the figure icon (grid icon) in the top right of the right panel, and choose "Thumbnail Figure".

Now you will see all the datasets with their thumbnails, and if you choose Tags in the filter field, you'll see the thumbnails organised accordingly.
If some thumbnails contain multiple selected tags, you'll see them nested (see screenshot).

If you want to, you can then run the script to generate the output as an image to download etc.
This script is also available to run in Insight, but without the same preview UI (you just choose tags from a list) so you don't get to see the result of your filtering until you generate and download the figure.

Hope that helps.

Will.

[LeroyAlex]

Hi Russell,

Thanks for your answer. I also thought of the workflow you just described.
It works but i choose to avoid this solution for a few reason:

-first, like you said, you duplicate effort/data (not so much effort if you use the auto tag app though).

-secondly, we try to teach our users to use tags to classify/look for their images, but we want them to be as universal as possible (official nomenclature for strain/mutant for example and limitate the use of more abstract tags such as "good quality image" that not everyone might name like this). The point is to underscore the power of omero when it comes to translational research. By that i mean how users of a same group can look into each others data and are able to make comparaison. Ex: data that could have the same dye and gene tag but different tissues tags underlining a difference of effect related to the tissue. Moreover we tell them that the name of their project/dataset/images is their free space so people even if they can access the generic tags like gfp, they would get confused when they have to choose a project tag cause the project's name of their colleagues might not make any sense for them. In the end they would have to go through every project to get an idea of what it is, then specifically select the few one they want to compare.
Of course it can be used to look into your own data since you know about your own tags.

I apologize if what i said is not clear enough.

Hi wmoore,

Thx for your answer. I did not know you could do that on omero and i ll be sure to inform the users. It might be more useful cause you could make abstraction of what the tag name of project/dataset means just by running the script once with every dataset selected to get an overview of all available tags by dataset, then run it a second time with only the dataset that you want to compare
The only pros is that if there are too many dataset, selecting them all would make the analysis of tags troublesome and if you select them 2 by 2 to simplify this task it might be time consuming. Of course this problem arises just when you look into someone else data since you usually know what kind of images/tags you ll find in your own data.

Best,

Alex.

[wmoore]

Hi Alex,

OK, hopefully some of what we've discussed is helpful. Let us know how you get on with your users.

Regards,

Will.